JP4294595B2 - Transmission coefficient estimation circuit and transmission coefficient estimation method - Google Patents

Description

  The present invention uses a multiple-input multiple-output (MIMO) channel that realizes broadband mobile communication to multiplex a plurality of different information signal sequences on the same frequency and perform interference cancellation by a receiver. The present invention relates to a transmission coefficient estimation circuit and a transmission coefficient estimation method for a spatially multiplexed signal transmitting / receiving apparatus that realizes high-speed and large-capacity wireless communication for the purpose of achieving an increase in frequency utilization efficiency that is the same number as the number of multiplexed signals.

  In broadband mobile communication, it is essential to improve frequency utilization efficiency in order to increase the capacity using limited frequency resources. As a technique for improving frequency utilization efficiency, a spatial multiplexing channel is formed on the same frequency band using a plurality of transmitting antennas and a plurality of receiving antennas, and a transfer coefficient between the transmitting antenna and the receiving antenna is set on the receiving side. Techniques have been proposed for improving the transmission rate by the same number as the number of transmission antennas by estimating and estimating the transmission signal from the estimated transmission coefficient and received signal sequence.

  In addition, as a countermeasure against frequency selective fading, an OFDM (Orthogonal Frequency Division Multiplexing) method in which a signal is divided into a plurality of mutually orthogonal subcarrier groups and transmitted is used in a wireless LAN system or the like. The OFDM method can flatten the influence of frequency selective fading by narrowing the band of each subcarrier. Further, by adding a guard interval, the influence of inter-symbol interference due to multipath fading is reduced, and a complicated equalization circuit is not required, so that the compatibility with MIMO transmission is high.

  In the OFDM system using the MIMO channel for packet transmission, the transmission coefficient on the receiving side is estimated using a known pilot signal added to the head of the burst. However, when the number of data symbols included in a burst increases, the initial estimated value of the transmission coefficient estimated with the pilot signal and the current propagation path estimated value become more behind the burst due to the time variation of the multipath fading channel. There is a problem that the difference between the true value and the true value becomes larger, the probability that information can be correctly demodulated decreases, and as a result, the error rate increases.

  In order to solve this problem, there has conventionally been proposed a method of inserting a known pilot signal in the middle of a burst and tracking a transfer coefficient (see, for example, Non-Patent Document 1). In this method, since insertion of a pilot signal causes a decrease in transmission efficiency, a transfer coefficient tracking method that does not use a known signal is employed.

An example of an OFDM signal transmitting / receiving apparatus using a MIMO channel that does not use a pilot signal according to the prior art is shown in FIG. 17 (see, for example, Non-Patent Document 2). The prior art shown in FIG. 17 includes an OFDM signal transmitter 1 and an OFDM signal receiver 2. In the OFDM signal transmission apparatus 1, transmission information signal sequences T ₁ , T ₂ ,..., T _{K of} K systems (K is an integer of 2 or more) are K multiplexers 1-2-1 to 1-2K. Is input. The pilot signal generator 1-1 outputs a pilot signal for estimating a transfer coefficient. In the K pilot signal multiplexers 1-2-1 to 1-2K, transmission information signal sequences T ₁ , T ₂ ,..., And T _K and pilot signals for propagation path estimation are on the time axis or Multiplexed on the frequency axis.

  The signal obtained by multiplexing the pilot signal is input to K subcarrier modulators 1-3-1 to 1-3-K, and signal point mapping is performed for each subcarrier. Each subcarrier signal subjected to signal point mapping is converted from a frequency domain signal to a time domain signal sequence by K inverse fast Fourier transforms 1-4-1 to 1-4 -K, and K transmission antennas 1 are transmitted. -5-1 to 1-5-K are transmitted to the space. An example of a burst signal transmitted by the OFDM signal transmission apparatus 1 is shown in FIG.

  In the OFDM signal receiver 2, M received signal sequences (M ≧ 1, M is an integer) transmitted from the transmitting antennas 1-5-1 to 1-5-N and multiplexed in space are M pieces. Received by the receiving antennas 2-1-1-2-1 -M. The received signal sequences received by the M receiving antennas 2-1-1-2-1 -M are subjected to fast Fourier transform by M fast Fourier transformers 2-2-1-2-2 -M, and time The signal in the region is converted into a signal in the frequency region, that is, a subcarrier. The transfer coefficient estimation circuit 2-3 estimates the initial value of the transfer coefficient using the received signal sequence corresponding to the pilot signal in the output signal sequence of the M fast Fourier transformers 2-2-1 to 2-2M. To do. The interference canceller 2-4 cancels interference of the data signal using the estimated initial value, and estimates a signal point for each subcarrier and each transmission antenna from the spatially multiplexed received signal.

  As interference cancellation methods, algorithms such as ZF (Zero-Forcing), MMSE (Minimum Mean-Square Error) linear filtering, MLD (Maximum Likelihood Detection), and BLAST (Bell labs LAyerd Space Time) have been proposed. However, it is known that the QRD-M algorithm combining the QR decomposition on the transfer coefficient matrix and the MLD and K-best methods achieves both a small amount of computation and a good error rate characteristic. The QRD-M algorithm will be described below.

  A transmission signal for subcarrier i transmitted from transmission antenna k (k = 1, 2,..., K) at a certain time (OFDM symbol) is t (i, k), and reception antenna m (m = 1, 2,...). , M) for signals received by y (i, m), thermal noise contained in y (i, m) for n (i, m), transmitting antenna k and receiving antenna m (m = 1, 2,... , M) and h (i, m, k) as an estimated value of the transmission coefficient for the subcarrier i between the transmission signal and the reception signal, the relationship between the transmission signal and the reception signal is expressed as in Equation (1). In this equation, the parameter representing the time is omitted to avoid complication.

  When the transfer coefficient matrix, the transmission signal vector, the thermal noise vector, and the reception signal vector are expressed as H (i), t (i), n (i), and y (i), respectively, Expression 1 is expressed as Expression 2. The

  In the QRD-M algorithm, first, QR decomposition is performed on the transfer coefficient H (i).

Here, Q (i) is an orthogonal matrix of M rows and K columns, and R (i) is an upper triangular matrix of K rows and K columns.
Multiplying the obtained Hermitian conjugate matrix Q ^H (i) of Q (i) from the left on both sides of Equation (2) and substituting Equation (3) yields Equation (4) and Equation (5). .

Where y ′ (t) and n ′ (t) are Q ^H (i) y (i) and Q ^H (i) n (i), respectively, and n ′ (t) is unitary with Q ^H (i). Since it is a matrix, it has the same statistical properties as n (i).

  When each element of the matrix of Expression (5) is expressed, Expression (6) is obtained.

  Here, paying attention to the K-th line in the equation (6), y ′ (i, K) = r (i, K, K) t (i, K) + n ′ (i, K). I understand. That is, in order to perform the maximum likelihood determination of t (i, K), only the replica for t (i, K) needs to be considered. For example, if C-QAM (Quadrature Amplitude Modulation: C is an integer of 2 or more) is used for each subcarrier, there are C types of patterns that can be taken by t (i, K). It is only necessary to select a replica having a minimum distance from the received signal.

Here, Tx represents a set composed of all possible patterns of the transmission signal, and t _MLD (i, k) represents a hard decision value corresponding to the transmission signal t (i, k). If t _MLD (i, K) is used when calculating the likelihood of the (K-1) th row, C calculations may be performed in the maximum likelihood determination of the (K-1) th row.

  Similarly, (m′−1) hard decision results from the Km ′ (m ′ = 1, 2,..., K−1) line to the Km ′ + 1 line are used. And the maximum likelihood determination may be performed.

  With the above method, the determination for all transmission signals of each subcarrier can be obtained by performing C × K operations. However, if the result of the hard decision made in the previous stage is continued to be used in the latter stage, the characteristic deterioration becomes large due to the decision error in the previous stage. Therefore, in the likelihood calculation, the top B pieces having the highest likelihood are selected as decision value candidates. A method is used in which the K-best method is used to improve the characteristics by reflecting it in the determination value of the next stage.

  The interference canceller 2-4 estimates the transmission signal using the QRD-MLD method, and outputs it to the K demodulators 2-5-1 to 2-5-K. K demodulator 2-5-1 to 2-5-K decode the information bit sequence of the transmission signal from the input estimated value of the transmission signal.

  A transfer coefficient tracking circuit using the QRD-MLD method that has been conventionally proposed will be described below. The internal configuration of the transfer coefficient tracking circuit 2-3 is shown in FIG. First, the transmission coefficient of each subcarrier is estimated by the transmission coefficient initial value estimator 2-3-1 using the received signal corresponding to the known pilot signal included in the head of the burst and the known pilot signal pattern. .

  For estimating the initial value of the transfer coefficient, there is a method using a pilot signal at the head of a burst as shown in FIG. While a transmitting antenna is transmitting a pilot signal, all other transmitting antennas do not transmit signals, so that transmission between the antenna transmitting the pilot signal at that time and all receiving antennas is performed. The coefficient can be estimated. By performing the same operation for all K transmission antennas, transfer coefficients between all transmission antennas and reception antennas can be estimated. The estimated transfer coefficient is output and stored in the transfer coefficient estimated value storage circuit 2-3-4 as each delay profile between the transmitting antenna and the receiving antenna.

  In the K fast inverse Fourier transformers 2-3-2-1 to 2-3-2-K, the estimated value of the signal point of the transmission signal, which is the output of the interference canceller 2-4 used as a replica of the transmission signal, is OFDM Convert to a time waveform for each symbol. The time waveform of the replica of the transmission signal generated by the K fast inverse Fourier transformers 2-3-2-1 to 2-3-2-N and stored in the transfer coefficient estimated value storage circuit 2-3-4 The estimated values of the delay profiles for the propagation paths between all the transmitting antennas and the receiving antennas in the past closest to the present time, and the outputs of the M fast Fourier transformers 2-2-1 to 2-2M. Kalman filtering is performed by the Kalman filter 2-3-3 using the current received signal.

  Then, propagation path delay profiles between all transmission antennas and reception antennas at the current time in consideration of time variation of the propagation path during the transition from the past to the present are estimated. The estimated delay profile between the transmitting and receiving antennas is input and stored in the transfer coefficient estimated value storage circuit 2-3-4. The transfer coefficient estimated value storage circuit 2-3-4 receives outputs from both the transfer coefficient initial value estimator 2-3-1 and the Kalman filter 2-3-3, but the initial estimated value at the head of the burst. Immediately after is calculated, the input from the transfer coefficient initial value estimator 2-3-1 is stored, and after the first Kalman filtering is performed, the input from the Kalman filter 2-3-3 is stored. Do.

Delay profile estimation sequences for combinations of K × M propagation paths of all transmission antennas and reception antennas estimated by the Kalman filter 2-3-3 are transmitted to each propagation path by a delay profile estimation value distributor 2-3-5. And is input to K × M fast Fourier transformers 2-3-1-1-1 to 2-3-6 K × M, and K × M fast Fourier transformers 2-3 6-1 to 2-3-6−K × M is converted into a transmission coefficient for each subcarrier and input to the interference canceller 2-4. The interference canceller 2-4 performs a QR-MLD operation using the updated transfer coefficient. As a result, followability due to propagation path fluctuation is improved, and high-quality communication is realized.
Kurosaki, Uchida, Asai, Sugiyama, Umehira “Experimental Study of Broadband SDM-COFDM (2)-Improvement of Transmission Characteristics by Tracking Transmission Coefficient Matrix”-2003 IEICE General Conference Jiang Yue, Kyeong Jin Kim, Tony Reid and Jerry D. Gibson, “Joint Semi-Blind Channel Estimation and Data Detection for MIMO-OFDM”, 2004 IEEE 6th CAS Symposium on Emerging Technologies: Mobile and Wireless Communications)

  In the conventional transfer coefficient estimation method, a Kalman filter is used, but it is necessary to collectively estimate delay profiles of all propagation paths. Therefore, the size of the matrix of the estimation sequence composed of the delay profiles of all the propagation paths to be estimated becomes very large with respect to the increase in the spread of the propagation path delay profile and the increase in the number of transmission and reception antennas. In the Kalman filter, since it is necessary to calculate an inverse matrix of a matrix of the same scale as the estimated series, the amount of calculation becomes very large, and there is a problem that the circuit scale increases exponentially.

  The present invention has been made in consideration of such circumstances, and its purpose is not to obtain the delay profiles of all the propagation paths at once as in the conventional transfer coefficient estimation method, but for each subcarrier. An object of the present invention is to provide a transmission coefficient estimation circuit and a transmission coefficient estimation method that can be realized with a small circuit scale by reducing the amount of calculation by updating the transmission coefficient for each transmission coefficient matrix or transmission coefficient.

In order to solve the above-described problem, the present invention provides an OFDM received signal when a spatially multiplexed OFDM transmission signal sequence transmitted from a plurality of transmitting antennas on the transmitting side is received by a plurality of receiving antennas on the receiving side. Fast Fourier transform is performed on the sequence, and a known pilot signal transmitted from the transmitting side and a received signal after the fast Fourier transform corresponding to the pilot signal are used to generate a sub-carrier from the spatially multiplexed received signal. A transmission coefficient estimation circuit for estimating a transmission coefficient between antennas for outputting a signal point of the signal, remodulating a transmission data sequence that is a demodulation output for the signal point for each subcarrier, and transmitting a signal for each subcarrier. A remodulator that generates a replica, and a replica matrix that includes transmission signal replicas generated by the remodulator as sub-elements. Elements are a replica matrix generator to be generated, a pseudo inverse matrix calculator for obtaining an inverse matrix or a pseudo inverse matrix of the replica matrix for each subcarrier, and a reception signal output for each subcarrier after the fast Fourier transform A reception signal matrix generator for generating a reception signal matrix for each subcarrier, and multiplying the reception signal matrix by an inverse matrix or a pseudo inverse matrix of a replica matrix that is an output of the pseudo inverse matrix calculator, Estimate the initial value of the transfer coefficient matrix using a matrix multiplier that outputs as a replica estimated transfer coefficient matrix, and a known pilot signal transmitted from the transmitting side and the received signal after the Fast Fourier Transform corresponding to the pilot signal A transfer coefficient matrix initial estimation circuit that outputs the initial transfer coefficient estimation matrix, the initial transfer coefficient estimation matrix, and the replica estimated transfer coefficient row Weighted synthesized bets, comprising a transmission coefficient estimation matrix updating circuit for outputting a current transfer coefficient estimate, the transmission coefficient estimation matrix update circuit, a matrix weight calculating for calculating a weighting factor for the replica estimated transfer coefficient matrix Circuit and a matrix weighting synthesis circuit for weighting and synthesizing the initial transmission coefficient estimation matrix and the replica estimation transmission coefficient estimation matrix for each subcarrier using the weighting coefficient calculated by the matrix weight calculation circuit. The matrix weight calculation circuit uses the weight coefficient for the replica estimated transfer coefficient matrix estimated in the past and the present for each component of the pseudo inverse matrix of the replica matrix used when deriving each replica estimated transfer coefficient matrix. The weighting factor is inversely proportional to the sum of squares of absolute values.

  According to the present invention, in the above invention, the pseudo inverse matrix calculator has a function of generating and outputting an inverse matrix of a replica matrix for each subcarrier.

  According to the present invention, in the above invention, the transfer coefficient estimation matrix update circuit adds the weight coefficients corresponding to a plurality of replica transfer coefficient estimation matrices estimated at different times for each subcarrier for each same subcarrier. A transfer coefficient estimation matrix update determination that outputs an update command for the subcarrier to the transfer coefficient estimation matrix update circuit when the addition result by the weight adder and the matrix weight adder exceeds a predetermined threshold value The matrix weighting / synthesizing circuit for the subcarriers notified of the update command from the transfer coefficient estimation matrix update determination circuit, the past and current replica estimated transfer coefficient matrices and the past transfer coefficient estimates. Each matrix is weighted and combined and output as the current transfer coefficient estimation matrix. And outputs the transmission coefficient estimation matrix.

  According to the present invention, in the above invention, the transfer coefficient estimation matrix update circuit updates the time and current time when the past transfer coefficient estimation matrix is updated for each subcarrier based on the determination result by the transfer coefficient estimation matrix update determination circuit. The transfer coefficient estimation matrix non-update period calculation circuit that calculates the difference from the time and outputs it as the non-update period of the transfer coefficient estimation matrix, and exceeds a predetermined threshold for the non-update period of the transfer coefficient estimation matrix A transfer coefficient matrix history information control circuit for outputting a history erasure command, and a part of a past replica estimated transfer coefficient matrix for a subcarrier to which an erasure command is notified by the transfer coefficient history information control circuit. Alternatively, the entire history is erased, and the history of matrix weights corresponding to the erased past replica estimated transmission coefficients is erased.

  According to the present invention, in the above invention, based on the transfer coefficient estimation matrix and the replica estimated transfer coefficient matrix, whether or not update processing is valid or invalid is determined for each component of the transfer coefficient estimation matrix of each subcarrier. And a replica estimated transfer coefficient matrix filter that outputs only the subcarriers determined to be valid based on the determination result as a replica estimated transfer coefficient matrix.

  In the present invention, the replica estimated transfer coefficient matrix filter may calculate a difference between the transfer coefficient estimated matrix estimated in the past and the replica estimated transfer coefficient matrix output from the matrix multiplier as a subcarrier. For each transmission coefficient difference matrix for each subcarrier, the sum of all the geometric distances of the elements constituting the matrix is calculated for each subcarrier and the transmission coefficient difference matrix. If the matrix geometric distance calculation circuit to output as a value and the transfer coefficient difference value is below a predetermined threshold, the input replica estimated transfer coefficient matrix is output, and the default threshold is set. In the case of waiting for the top, it is constituted by a transfer coefficient estimation matrix selection circuit that does not output the input replica estimated transfer coefficient matrix.

  According to the present invention, in the above invention, the geometric distance calculation circuit uses an absolute value of elements constituting the matrix or a square value of the absolute value as a geometric distance of each element of the transfer coefficient difference matrix. It is characterized by that.

  According to the present invention, in the above invention, the geometric distance calculation circuit, as the geometric distance of each element of the transfer coefficient difference matrix, is an absolute value of the real part and an absolute value of the imaginary part of the elements constituting the matrix. Or the square of the value is used.

  According to the present invention, in the above invention, based on the transmission coefficient estimation matrix estimated in the past and the replica estimated transmission coefficient matrix, a determination for selecting an output is performed for each subcarrier, and the determination result is It has a transfer coefficient estimation matrix filter that outputs a replica estimated transfer coefficient matrix only for subcarriers determined to be valid.

  According to the present invention, in the above invention, the transfer coefficient estimation matrix filter calculates, for each subcarrier, a difference between the transfer coefficient estimation matrix estimated at the current time and the transfer coefficient estimation matrix estimated in the past. A matrix subtractor that outputs as a transfer coefficient difference matrix and a matrix that calculates the sum of geometric distances for all the elements constituting the matrix for each subcarrier transfer coefficient difference matrix and outputs as a transfer coefficient difference value When the geometric distance calculation circuit and the transfer coefficient difference value are below a predetermined threshold value, the transfer coefficient estimation matrix estimated at the current time is output to the interference canceller, and the predetermined threshold value is output. When waiting for an increase in value, the transmission coefficient estimation matrix selection circuit outputs the transfer coefficient estimation matrix estimated in the past to the interference canceller. To.

  According to the present invention, in the above invention, the matrix geometric distance calculation circuit calculates an absolute value of elements constituting the matrix or a square value of the absolute value as a geometric distance of each element of the transfer coefficient difference matrix. It is characterized by using.

  According to the present invention, in the above invention, the matrix geometric distance calculation circuit is configured so that the absolute value of the real part and the absolute value of the imaginary part of the elements constituting the matrix are used as the geometric distance of each element of the transfer coefficient difference matrix. A sum of values or a square value of the values is used.

Further, in order to solve the above-described problem, the present invention provides a method for receiving a spatially multiplexed OFDM transmission signal sequence transmitted from a plurality of transmission antennas on the transmission side by using a plurality of reception antennas on the reception side. Fast Fourier transform is performed on the received signal sequence, and the received signal that has been transmitted from the transmitting side and the received signal that has been subjected to the fast Fourier transform corresponding to the pilot signal are subtracted from the spatially multiplexed received signal. A transmission coefficient estimation circuit for estimating a transmission coefficient between antennas for outputting a signal point for each carrier, remodulating a transmission data sequence, which is a demodulation output for the signal point for each subcarrier, for each subcarrier A remodulator that generates a transmission signal replica and a replica matrix that includes the transmission signal replica generated by the remodulator as sub-elements. A replica matrix generator generated for each sub-carrier, a pseudo-inverse matrix calculator for obtaining an inverse matrix or pseudo-inverse matrix of the replica matrix for each subcarrier, and a reception signal output for each subcarrier after the fast Fourier transform. A reception signal matrix generator for generating a reception signal matrix as an element for each subcarrier, and multiplying the reception signal matrix by an inverse matrix or a pseudo inverse matrix of a replica matrix that is an output of the pseudo inverse matrix calculator, A matrix multiplier that outputs a multiplication result as a replica estimated transfer coefficient matrix, and an initial transfer coefficient matrix using a known pilot signal transmitted from the transmitting side and the received signal after the fast Fourier transform corresponding to the pilot signal. A transfer coefficient matrix initial estimation circuit that estimates a value and outputs it as an initial transfer coefficient estimation matrix, the initial transfer coefficient estimation matrix, and the replica estimated transfer And the number matrix weighted synthesized, comprising a transmission coefficient estimation matrix updating circuit for outputting a current transfer coefficient estimate, the transmission coefficient estimation matrix updating circuit, each time using the output of the pseudo-inverse matrix calculator A weight estimation circuit for calculating a weight coefficient for each column vector of the replica estimated transfer coefficient matrix for each subcarrier in the subcarrier, the currently estimated replica estimated transfer coefficient matrix and the previously estimated transfer coefficient estimated matrix, and the weight The replica estimated transfer coefficient matrix and the past transfer coefficient estimated matrix are each composed of a matrix weighting synthesis circuit that performs weighting synthesis for each column vector using a coefficient.

  According to the present invention, in the above invention, the vector weight calculation circuit uses a weight coefficient for a column vector of the replica estimated transfer coefficient matrix estimated in the past and present when deriving each replica estimated transfer coefficient matrix. The weight coefficient is inversely proportional to the sum of squares of the absolute values of the components of the inverse matrix of the replica matrix or the row vector of the pseudo inverse matrix.

  According to the present invention, in the above invention, the transfer coefficient estimated value update circuit adds a weight coefficient corresponding to a plurality of replica transfer coefficient estimation matrices estimated at different times for each subcarrier for each same column vector. A weight adder, a transfer coefficient estimation matrix column vector update determination circuit that outputs an update command for the column vector when the addition result of the matrix weight adder exceeds a predetermined threshold, and the transfer coefficient For the column vector for which the update command is notified from the estimation matrix update determination circuit, the current transfer coefficient is obtained by weighting and synthesizing each column vector of the past and current replica estimated transfer coefficient matrix and the past transfer coefficient estimation matrix. For a column vector that is output as a column vector of an estimation matrix and for which an update command is not notified, the replica estimated transmission coefficient matrix or the past transmission matrix And having a transmission coefficient estimation matrix column vector updating circuit for outputting a coefficient estimation matrix intact.

  According to the present invention, in the above invention, the transfer coefficient estimation matrix update circuit updates the past transfer coefficient estimation matrix for each column vector based on a determination result by the transfer coefficient estimation matrix update determination circuit and a current A transfer coefficient estimation matrix sequence vector non-update period calculation circuit for calculating a time difference and outputting as a transfer coefficient estimation matrix sequence vector non-update period, and a predetermined threshold for the transfer coefficient estimation matrix sequence vector non-update period If there is a transfer coefficient matrix history information control circuit that outputs a history erasure command, a column vector that has been notified of the erasure command by the transfer coefficient history information control circuit, the relevant replica estimation transfer coefficient matrix The history of a part or the whole of the column vector is deleted, and the matrix weight corresponding to the column vector of the deleted past estimated replica transfer coefficient is deleted. It characterized in that it erased the gravel.

  According to the present invention, in the above invention, based on the transfer coefficient estimation matrix and the replica estimated transfer coefficient matrix, update processing is enabled or disabled for each component of the column vector of the transfer coefficient estimation matrix of each subcarrier. It has a replica estimated transfer coefficient matrix column vector filter that makes a determination and outputs only the column vector that is determined to be valid as a result of the determination and outputs the column vector of the replica estimated transfer coefficient estimation matrix.

  According to the present invention, in the above invention, the replica estimated transfer coefficient matrix column vector filter calculates a difference between a transfer coefficient estimated matrix estimated in the past and the replica estimated transfer coefficient matrix for each column vector. A vector subtractor that outputs as a vector and a vector geometric that calculates the sum of the squares of all geometric distances of the elements that make up the vector for each transfer coefficient difference vector for each subcarrier and outputs it as a vector difference value When the distance calculation circuit and the vector difference value are below a predetermined threshold, the column vector of the replica estimated transfer coefficient matrix is output, and when waiting for the predetermined threshold above And a transfer coefficient matrix column vector estimation accuracy determination circuit that does not output the column vector of the replica estimated transfer coefficient matrix.

  According to the present invention, in the above invention, the vector geometric distance calculation circuit, as a geometric distance of each element of the transfer coefficient difference vector, is an absolute value of an element constituting the vector or a square value of the absolute value. It is characterized by using.

  According to the present invention, in the above invention, the vector geometric distance calculation circuit calculates the absolute value of the real part and the imaginary part of the elements constituting the vector as the geometric distance of each element of the transfer coefficient difference vector. A sum of absolute values or a square value of the values is used.

  According to the present invention, in the above invention, based on the transfer coefficient estimation matrix estimated in the past and the replica estimated transfer coefficient matrix, a determination for selecting an output is performed for each column vector. Only the column vector judged to be valid outputs the column vector of the replica estimated transfer coefficient matrix in the past, and for the column vector not judged valid, the column vector of the transfer coefficient estimation matrix estimated in the past is outputted. And a transfer coefficient estimation matrix sequence vector filter.

  According to the present invention, in the above invention, the transfer coefficient estimation matrix column vector filter calculates a difference between the transfer coefficient estimation matrix estimated at a current time and the transfer coefficient estimation matrix estimated in the past for each column vector. A matrix subtractor that outputs as a difference vector, a vector geometric distance calculation circuit that calculates a square sum of all geometric distances of elements constituting a column vector for each difference vector for each subcarrier, and the vector If the output of the geometric distance calculation circuit is below the predetermined threshold, the column vector of the transfer coefficient estimation matrix estimated at the current time is output to the interference canceller, and the predetermined threshold is set. If so, a transfer coefficient estimation matrix column vector selection circuit that outputs the column vector of the transfer coefficient estimation matrix estimated in the past to the interference canceller, etc. Characterized in that it is configured.

  According to the present invention, in the above invention, the vector geometric distance calculation circuit, as a geometric distance of each element of the transfer coefficient difference vector, is an absolute value of an element constituting the vector or a square value of the absolute value. It is characterized by using.

  According to the present invention, in the above invention, the vector geometric distance calculation circuit calculates the absolute value of the real part and the imaginary part of the elements constituting the vector as the geometric distance of each element of the transfer coefficient difference vector. A sum of absolute values or a square value of the values is used.

Further, in order to solve the above-described problem, the present invention provides a method for receiving a spatially multiplexed OFDM transmission signal sequence transmitted from a plurality of transmission antennas on the transmission side by using a plurality of reception antennas on the reception side. Fast Fourier transform is performed on the received signal sequence, and the received signal that has been transmitted from the transmitting side and the received signal that has been subjected to the fast Fourier transform corresponding to the pilot signal are subtracted from the spatially multiplexed received signal. A transfer coefficient estimation method used in a transfer coefficient estimation circuit for estimating a transfer coefficient between antennas for outputting a signal point for each carrier , wherein the remodulator is a demodulated output for the signal point for each subcarrier. re modulates transmission data sequence, perform the remodulation procedure that generates a transmission signal replica for each subcarrier, a replica matrix generator is pre Run the replica matrix generation step of generating a replica matrix to a transmission signal replica generated by the re-modulator elements for each subcarrier, pseudo inverse matrix calculator is inverse or pseudo replica matrix for each of the sub-carrier The reception signal matrix generator executes a pseudo inverse matrix calculation procedure for obtaining an inverse matrix, and the reception signal matrix generator generates a reception signal matrix for each subcarrier, the reception signal being output for each subcarrier after the fast Fourier transform. A signal matrix generation procedure is executed, and the matrix multiplier multiplies the reception signal matrix by an inverse matrix or a pseudo inverse matrix of a replica matrix that is an output of the pseudo inverse matrix calculator, and the multiplication result is a replica estimated transfer coefficient. perform the matrix multiplication procedure for outputting as a matrix, transfer coefficient matrix initial estimation circuit, known pilot signal and the pilot signal transmitted from the transmitting side By using the received signal after the fast Fourier transformation corresponding estimates the initial value of the transfer coefficient matrix, perform the transfer coefficient matrix initial estimation procedure for outputting an initial transmission coefficient estimation matrix, transmission coefficient estimation matrix updating circuit The transmission coefficient estimation matrix update procedure for weighting and synthesizing the initial transmission coefficient estimation matrix and the replica estimation transmission coefficient matrix and outputting as a current transmission coefficient estimation value is executed . In the transmission coefficient estimation matrix update procedure, The coefficient estimation matrix update circuit uses the matrix weight calculation procedure for calculating a weight coefficient for the replica estimated transfer coefficient matrix and the weight coefficient calculated by the matrix weight calculation circuit, and the initial transfer coefficient estimation matrix and the replica estimation Matrix weighting synthesis procedure for weighting and synthesizing the transfer coefficient estimation matrix for each subcarrier , and the matrix weight calculation procedure In the above, the weight coefficient for the replica estimated transfer coefficient matrix estimated in the past and the present is inversely proportional to the sum of squares of the absolute value of each component of the pseudo inverse matrix of the replica matrix used in deriving each replica estimated transfer coefficient matrix It is characterized in that the weighting factor is set .

Further, in order to solve the above-described problem, the present invention provides a method for receiving a spatially multiplexed OFDM transmission signal sequence transmitted from a plurality of transmission antennas on the transmission side by using a plurality of reception antennas on the reception side. Fast Fourier transform is performed on the received signal sequence, and the received signal that has been transmitted from the transmitting side and the received signal that has been subjected to the fast Fourier transform corresponding to the pilot signal are subtracted from the spatially multiplexed received signal. A transfer coefficient estimation method used in a transfer coefficient estimation circuit for estimating a transfer coefficient between antennas for outputting a signal point for each carrier , wherein the remodulator is a demodulated output for the signal point for each subcarrier. re modulates transmission data sequence, perform the remodulation procedure for generating a transmission signal replica for each subcarrier, a replica matrix generator is pre Run the replica matrix generation step of generating a replica matrix to a transmission signal replica generated by the re-modulator elements for each subcarrier, pseudo inverse matrix calculator is inverse or pseudo replica matrix for each of the sub-carrier The reception signal matrix generator executes a pseudo inverse matrix calculation procedure for obtaining an inverse matrix , and the reception signal matrix generator generates a reception signal matrix for each subcarrier, the reception signal being output for each subcarrier after the fast Fourier transform. A signal matrix generation procedure is executed, and the matrix multiplier multiplies the reception signal matrix by an inverse matrix or a pseudo inverse matrix of a replica matrix that is an output of the pseudo inverse matrix calculator, and the multiplication result is a replica estimated transfer coefficient. perform the matrix multiplication procedure for outputting as a matrix, transfer coefficient matrix initial estimation circuit, known pilot signal and the pilot signal transmitted from the transmitting side By using the received signal after the fast Fourier transformation corresponding estimates the initial value of the transfer coefficient matrix, perform the transfer coefficient matrix initial estimation procedure for outputting an initial transmission coefficient estimation matrix, transmission coefficient estimation matrix updating circuit The transmission coefficient estimation matrix update procedure for weighting and synthesizing the initial transmission coefficient estimation matrix and the replica estimation transmission coefficient matrix and outputting as a current transmission coefficient estimation value is executed. In the transmission coefficient estimation matrix update procedure, The coefficient estimation matrix update circuit uses the output of the pseudo inverse matrix calculator to calculate a replica estimated transfer coefficient matrix for each subcarrier at each time, the currently estimated replica estimated transfer coefficient matrix, and the previously estimated transfer coefficient A vector weight calculation procedure for calculating a weighting factor for each column vector of the estimation matrix, and the replica estimation using the weighting factor A matrix weighting synthesis procedure for performing matrix weighting for weighting and synthesizing the transfer coefficient matrix and the past transfer coefficient estimation matrix for each column vector is performed.

  According to the present invention, the remodulator remodulates the transmission data sequence, which is the demodulated output for the signal point for each subcarrier, to generate a transmission signal replica for each subcarrier, and the replica matrix generator generates the retransmission signal. A replica matrix having a transmission signal replica generated by the modulator as an element is generated for each subcarrier, and an inverse matrix or a pseudo inverse matrix of the replica matrix for each subcarrier is obtained by a pseudo inverse matrix calculator, and a received signal matrix is obtained. A generator generates a reception signal matrix having a reception signal output for each subcarrier after the fast Fourier transform as an element for each subcarrier, and a matrix multiplier generates the pseudo inverse matrix for the reception signal matrix. Multiply the inverse matrix or pseudo inverse matrix of the replica matrix that is the output of the arithmetic unit, and output the multiplication result as a replica estimated transfer coefficient matrix. A matrix initial estimation circuit estimates an initial transmission coefficient estimation matrix using a known pilot signal transmitted from the transmission side and the received signal after the fast Fourier transform corresponding to the pilot signal, and a transmission coefficient estimation matrix update circuit Thus, the initial transmission coefficient estimation matrix and the replica estimation transmission coefficient matrix are weighted and synthesized, and output as a current transmission coefficient estimation value. Therefore, by performing transmission coefficient estimation for each subcarrier using a pseudo inverse matrix of a transmission signal replica, it is possible to reduce the amount of calculation required for transmission coefficient estimation compared to the case of using a conventional Kalman filter. The advantage is obtained.

  According to the present invention, the inverse matrix of the replica matrix is generated for each subcarrier by the pseudo inverse matrix calculator. Therefore, by performing transmission coefficient estimation for each subcarrier using an inverse matrix of a transmission signal replica, it is possible to reduce the amount of computation required for transmission coefficient estimation compared to the case of using a conventional Kalman filter. Benefits are gained.

  According to the present invention, the transfer coefficient estimation matrix update circuit calculates a weight coefficient for the replica estimated transfer coefficient matrix by a matrix weight calculation circuit, and calculates by the matrix weight calculation circuit by a matrix weighting synthesis circuit. The initial transmission coefficient estimation matrix and the replica estimated transmission coefficient estimation matrix are weighted and synthesized for each subcarrier using the weighting coefficient. Therefore, an advantage that the transfer coefficient estimation accuracy can be improved by weighting and combining the transfer coefficient estimation matrix and the replica estimated transfer coefficient matrix is obtained.

  Further, according to the present invention, the weight of the estimated replica transfer coefficient matrix estimated in the past and the present by the matrix weight calculation circuit is used to simulate the replica matrix used in deriving each estimated replica transfer coefficient matrix. The weighting coefficient is inversely proportional to the sum of squares of absolute values of each component of the inverse matrix. Therefore, when the transfer coefficient estimation matrix and the replica estimated transfer coefficient matrix are weighted and synthesized, the transfer coefficient estimation accuracy can be improved by adaptively changing the weight according to the estimation accuracy of the replica estimated transfer coefficient matrix. The advantage that it can be obtained.

  Further, according to the present invention, in the transfer coefficient estimation matrix update circuit, the weight coefficients corresponding to a plurality of replica transfer coefficient estimation matrices estimated at different times for each subcarrier by the matrix weight adder are set to the same subcarrier. When the addition result by the matrix weight adder exceeds a predetermined threshold by the transfer coefficient estimation matrix update determination circuit, an update command for the subcarrier is transmitted to the transfer coefficient estimation matrix update circuit. For the subcarriers for which the update command is notified by the matrix weighting synthesis circuit, the past and current replica estimated transmission coefficient matrix and the past transmission coefficient estimation matrix are weighted and synthesized, and the current transmission is performed. Output a past transfer coefficient estimation matrix for subcarriers that are output as a coefficient estimation matrix and are not notified of an update command. It was. Therefore, when the transfer coefficient estimation matrix and the replica estimated transfer coefficient matrix are weighted and combined, the estimation accuracy after the weighted combination of a plurality of replica estimated transfer coefficient matrices can be maintained at a certain level or more, and the transfer coefficient estimation accuracy can be increased. The advantage that it can be improved is obtained.

  Further, according to the present invention, in the transfer coefficient estimation matrix update circuit, a transfer coefficient estimation matrix non-update period calculation circuit causes a past transfer for each subcarrier based on a determination result by the transfer coefficient estimation matrix update determination circuit. The difference between the time when the coefficient estimation matrix is updated and the current time is calculated as the non-update period of the transfer coefficient estimation matrix, and the transfer coefficient matrix history information control circuit determines the difference for the non-update period of the transfer coefficient estimation matrix. If the threshold value of the transfer coefficient history information control circuit is exceeded, a history erasure command is output. And the history of matrix weights corresponding to the deleted past estimated replica transfer coefficients are deleted. Therefore, it is possible to avoid the weighted synthesis of the replica estimated transfer coefficient matrix in the past for a certain period of time and to improve the transfer coefficient estimation accuracy.

  Further, according to the present invention, the replica estimation transfer coefficient matrix filter performs update processing on each component of the transfer coefficient estimation matrix of each subcarrier based on the transfer coefficient estimation matrix and the replica estimated transfer coefficient matrix. Validity or invalidity is determined, and only the subcarriers determined to be effective based on the determination result are output as the replica estimated transmission coefficient matrix. Therefore, since it is possible to avoid using a replica estimated transmission coefficient matrix with an extremely low estimation accuracy for updating the transmission coefficient estimation matrix, there is an advantage that the transmission coefficient estimation accuracy can be improved.

  Further, according to the present invention, in the replica estimated transfer coefficient matrix filter, a matrix subtracter calculates a transfer coefficient estimated matrix estimated in the past and a replica estimated transfer coefficient matrix that is an output from the matrix multiplier. The difference is calculated for each subcarrier to obtain a transfer coefficient difference matrix, and for each transfer coefficient difference matrix for each subcarrier, the matrix geometric distance calculation circuit calculates all the geometric distances of the elements constituting the matrix. The transfer coefficient difference value is calculated by calculating the sum, and if the transfer coefficient difference value is below a predetermined threshold, the input replica estimated transfer coefficient matrix is output by the transfer coefficient estimation matrix selection circuit. When waiting for a predetermined threshold value, the input replica estimated transfer coefficient matrix is not output. Therefore, since it is possible to avoid using a replica estimated transmission coefficient matrix with an extremely low estimation accuracy for updating the transmission coefficient estimation matrix, there is an advantage that the transmission coefficient estimation accuracy can be improved.

  According to the invention, the geometric distance calculation circuit uses an absolute value of elements constituting the matrix or a square value of the absolute value as a geometric distance of each element of the transfer coefficient difference matrix. I made it. Therefore, since it is possible to avoid using a replica estimated transmission coefficient matrix with an extremely low estimation accuracy for updating the transmission coefficient estimation matrix, there is an advantage that the transmission coefficient estimation accuracy can be improved.

  According to the invention, in the geometric distance calculation circuit, the absolute value of the real part and the absolute value of the imaginary part of the elements constituting the matrix are used as the geometric distance of each element of the transfer coefficient difference matrix. The sum or the square of the value was used. Therefore, by using the square sum of the Manhattan distance of the difference matrix, there is an advantage that the amount of calculation when detecting the replica estimated transfer coefficient matrix with excessively low estimation accuracy can be reduced.

  Further, according to the present invention, the transfer coefficient estimation matrix filter makes a determination for selecting an output for each subcarrier based on the previously estimated transfer coefficient estimation matrix and the replica estimated transfer coefficient matrix. The replica estimated transfer coefficient matrix is output only for the subcarriers determined to be valid based on the determination result. Therefore, it is possible to eliminate a transfer coefficient estimation matrix having an excessively low estimation accuracy, thereby obtaining an advantage that the transfer coefficient estimation accuracy can be improved.

  According to the present invention, in the transfer coefficient estimation matrix filter, a difference between the transfer coefficient estimation matrix estimated at the current time and the transfer coefficient estimation matrix estimated in the past is calculated for each subcarrier by a matrix subtractor. To calculate a transfer coefficient difference matrix, and for each transfer coefficient difference matrix for each subcarrier, calculate a sum of geometric distances for all of the elements constituting the matrix by a matrix geometric distance calculation circuit. A transfer coefficient difference value is obtained, and when the transfer coefficient difference value is below a predetermined threshold by the transfer coefficient estimation matrix selection circuit, the transfer coefficient estimation matrix estimated at the current time is transferred to the interference canceller. In the case of outputting and waiting for a predetermined threshold value, the transfer coefficient estimation matrix estimated in the past is output to the interference canceller. Therefore, it is possible to eliminate a transfer coefficient estimation matrix having an excessively low estimation accuracy, thereby obtaining an advantage that the transfer coefficient estimation accuracy can be improved.

  According to the invention, in the matrix geometric distance calculation circuit, an absolute value of elements constituting the matrix or a square value of the absolute value is used as a geometric distance of each element of the transfer coefficient difference matrix. I did it. Therefore, it is possible to eliminate a transfer coefficient estimation matrix having an excessively low estimation accuracy, thereby obtaining an advantage that the transfer coefficient estimation accuracy can be improved.

  According to the present invention, in the matrix geometric distance calculation circuit, the absolute value of the real part and the imaginary part of the elements constituting the matrix are used as the geometric distance of each element of the transfer coefficient difference matrix. Or the square value of the value. Therefore, by using the square sum of the Manhattan distance of the difference matrix, there is an advantage that it is possible to reduce the amount of calculation when detecting the transfer coefficient estimation matrix with extremely low estimation accuracy.

  Further, according to the present invention, in the transfer coefficient estimation matrix update circuit, the vector weight calculation circuit uses the output of the pseudo inverse matrix calculator, the replica estimated transfer coefficient matrix for each subcarrier at each time, the current estimation A weighted coefficient for each column vector of the estimated replica transfer coefficient matrix and the previously estimated transfer coefficient estimation matrix, and using the weight coefficient by a matrix weighting synthesis circuit, the replica estimated transfer coefficient matrix and the The past transfer coefficient estimation matrix is weighted and synthesized for each column vector. Therefore, it is possible to improve the transfer coefficient estimation accuracy by weighting and combining the transfer coefficient estimation matrix and the replica estimated transfer coefficient matrix in units of column vectors.

  Further, according to the present invention, in the vector weight calculation circuit, the replica used in deriving each replica estimated transfer coefficient matrix is the weight coefficient for the column vector of the replica estimated transfer coefficient matrix estimated in the past and the present. The weighting coefficient is inversely proportional to the sum of squares of the absolute values of the components of the inverse matrix of the matrix or the row vector of the pseudo inverse matrix. Therefore, when the transfer coefficient estimation matrix and the replica estimated transfer coefficient matrix are weighted and combined in units of column vectors, the weight is adaptively changed according to the estimation accuracy for each column vector of the replica estimated transfer coefficient matrix. The advantage that the coefficient estimation accuracy can be improved is obtained.

  Further, according to the present invention, in the transfer coefficient estimated value update circuit, weight vector coefficients corresponding to a plurality of replica transfer coefficient estimation matrices estimated at different times for each subcarrier by a vector weight adder are stored in the same column vector. When the addition result of the matrix weight adder exceeds a predetermined threshold by the transfer coefficient estimation matrix column vector update determination circuit, an update command for the column vector is output and the transfer coefficient For the column vector for which the update command is notified by the estimation matrix column vector update circuit, the column vectors of the past and current replica estimated transfer coefficient matrix and the past transfer coefficient estimation matrix are weighted and combined to obtain the current For a column vector that is output as a column vector of a transfer coefficient estimation matrix and for which an update command is not notified, the replica estimated transfer coefficient matrix or the above The transfer coefficient estimation matrix Sa was allowed to be outputted. Therefore, when the transfer coefficient estimation matrix and the replica estimated transfer coefficient matrix are weighted and combined in units of column vectors, the estimation accuracy after weighted combining for each column vector of a plurality of replica estimated transfer coefficient matrices must be maintained at a certain level or higher. Thus, there is an advantage that the transfer coefficient estimation accuracy can be improved.

  Further, according to the present invention, in the transfer coefficient estimation matrix update circuit, the transfer coefficient estimation matrix column vector non-update period calculation circuit performs past data for each column vector based on the determination result by the transfer coefficient estimation matrix update determination circuit. The difference between the time at which the transmission coefficient estimation matrix of the update was updated and the current time is calculated and output as a transmission coefficient estimation matrix sequence vector non-update period, and the transfer coefficient estimation matrix sequence vector is not updated by the transfer coefficient matrix history information control circuit When a predetermined threshold value is exceeded for the period, a history erasure command is output, and a column vector for which the erasure command is notified is partly or entirely of the column vector of the past replica estimated transfer coefficient matrix. The history is erased, and the history of the matrix weight corresponding to the column vector of the erased past replica estimated transmission coefficient is erased. Accordingly, it is possible to avoid using the column vector of the replica estimated transmission coefficient matrix having an excessively low estimation accuracy for updating the transmission coefficient estimation matrix, so that there is an advantage that the transmission coefficient estimation accuracy can be improved. .

  Further, according to the present invention, each component of the column vector of the transfer coefficient estimation matrix of each subcarrier based on the transfer coefficient estimation matrix and the replica estimated transfer coefficient matrix by the replica estimated transfer coefficient matrix column vector filter The update processing is determined to be valid or invalid, and only the column vector determined to be valid based on the determination result is output as the column vector of the replica estimated transfer coefficient estimation matrix. Accordingly, it is possible to avoid using the column vector of the replica estimated transmission coefficient matrix having an excessively low estimation accuracy for updating the transmission coefficient estimation matrix, so that there is an advantage that the transmission coefficient estimation accuracy can be improved. .

  Further, according to the present invention, in the replica estimated transfer coefficient matrix column vector filter, a vector subtracter calculates a difference between a transfer coefficient estimated matrix estimated in the past and the replica estimated transfer coefficient matrix for each column vector. The transmission coefficient difference vector is calculated, and the vector geometric distance calculation circuit calculates the sum of the squares of all geometric distances of the elements constituting the vector for each transmission coefficient difference vector for each subcarrier. When the vector difference value is below a predetermined threshold by the transfer coefficient matrix column vector estimation accuracy judgment circuit, the column vector of the replica estimated transfer coefficient matrix is output and the default value is determined. When the threshold value is waited for, the column vector of the replica estimated transfer coefficient matrix is not output. Accordingly, it is possible to avoid using the column vector of the replica estimated transmission coefficient matrix having an excessively low estimation accuracy for updating the transmission coefficient estimation matrix, so that there is an advantage that the transmission coefficient estimation accuracy can be improved. .

  According to the invention, in the vector geometric distance calculation circuit, the absolute value of the elements constituting the vector or the square value of the absolute value is used as the geometric distance of each element of the transfer coefficient difference vector. I used it. Therefore, by using the square sum of the Manhattan distance of the difference matrix, there is an advantage that it is possible to reduce the amount of calculation when detecting the column vector of the replica estimation transfer coefficient matrix with excessively low estimation accuracy. .

  According to the present invention, in the vector geometric distance calculation circuit, the absolute value of the real part and the absolute value of the imaginary part of the elements constituting the vector are used as the geometric distance of each element of the transfer coefficient difference vector. The sum of the values or the square value of the values was used. Therefore, it is possible to eliminate the column vector of the transfer coefficient estimation matrix that has an excessively low estimation accuracy, thereby obtaining the advantage that the transfer coefficient estimation accuracy can be improved.

  Further, according to the present invention, for the selection of the output for each column vector based on the transfer coefficient estimation matrix estimated in the past and the replica estimated transfer coefficient matrix by the transfer coefficient estimation matrix column vector filter. Only the column vector judged to be valid as a result of the judgment is output in the past for the column vector of the replica estimated transfer coefficient matrix, and the transfer coefficient estimated in the past for the column vector not determined to be valid The column vector of the estimation matrix is output. Therefore, it is possible to eliminate the column vector of the transfer coefficient estimation matrix that has an excessively low estimation accuracy, thereby obtaining the advantage that the transfer coefficient estimation accuracy can be improved.

  Further, according to the present invention, in the transfer coefficient estimation matrix column vector filter, a matrix subtractor calculates a difference between the transfer coefficient estimation matrix estimated at the current time and the transfer coefficient estimation matrix estimated in the past as a column vector. The difference vector is calculated every time and the sum of squares of all the geometric distances of the elements constituting the column vector is calculated for each difference vector for each subcarrier by the vector geometric distance calculation circuit. When the output of the vector geometric distance calculation circuit is below a predetermined threshold by the estimation matrix column vector selection circuit, the column vector of the transfer coefficient estimation matrix estimated at the current time is used as the interference canceller. If the value exceeds the predetermined threshold, the column vector of the transfer coefficient estimation matrix estimated in the past is output to the interference canceller. Was Unishi. Therefore, it is possible to eliminate the column vector of the transfer coefficient estimation matrix that has an excessively low estimation accuracy, thereby obtaining the advantage that the transfer coefficient estimation accuracy can be improved.

  According to the invention, in the vector geometric distance calculation circuit, the absolute value of the elements constituting the vector or the square value of the absolute value is used as the geometric distance of each element of the transfer coefficient difference vector. I used it. Therefore, it is possible to eliminate the column vector of the transfer coefficient estimation matrix that has an excessively low estimation accuracy, thereby obtaining the advantage that the transfer coefficient estimation accuracy can be improved.

  According to the present invention, in the vector geometric distance calculation circuit, the absolute value of the real part and the absolute value of the imaginary part of the elements constituting the vector are used as the geometric distance of each element of the transfer coefficient difference vector. The sum of the values or the square value of the values was used. Therefore, by using the square sum of the Manhattan distance of the column vector of the difference matrix, there is an advantage that it is possible to reduce the amount of calculation when detecting the column vector of the transfer coefficient estimation matrix with excessively low estimation accuracy. .

Hereinafter, a transfer coefficient estimation circuit according to an embodiment of the present invention will be described with reference to the drawings.
A. First Embodiment (Claim 1)
FIG. 1 is a block diagram showing an OFDM signal transmission / reception apparatus including a transfer coefficient estimation circuit according to a first embodiment of the present invention. In the figure, an OFDM signal transmitting apparatus 3 includes a pilot signal generator 3-1, K multiplexers 3-2-1-2-3-K, and K subcarrier modulators 3-3-1 to 3-3-1. -3-K, composed of K inverse fast Fourier transformers 3-4-1 to 3-4-K and K transmission antennas 3-5-1 to 3-5-K.

The pilot signal generator 3-1 outputs a pilot signal for estimating a transfer coefficient. K pilot signal multiplexers 3-2-1-2-2-K transmit information signal sequences T ₁ , T _2, ... T _K and propagation path estimation pilot signals on a time axis or a frequency axis. Multiplex above. The signal obtained by multiplexing the pilot signal is input to K subcarrier modulators 3-3-1 to 3-3-3K, and signal point mapping is performed for each subcarrier. Each subcarrier signal subjected to signal point mapping is converted from a frequency domain signal to a time domain signal series by K inverse fast Fourier transforms 3-4-1 to 3-4 -K, and K transmission antennas 3 are transmitted. -5-1 to 3-5-K are transmitted in the same frequency band.

  The OFDM signal receiving apparatus 4 includes M receiving antennas 4-1-1 to 4-1-M, M fast Fourier transformers 4-2-1 to 4-2M, an initial transfer coefficient estimating circuit 4- 3, a transfer coefficient estimation circuit 4-4, an interference canceller 4-5, and demodulators 4-6-1 to 4-6-K.

  Signal sequences transmitted from the K transmission antennas 3-5-1 to 3-5 -N and multiplexed in space are received by M reception antennas 4-1-1-4-1 -M. M received signal sequences received by the M receiving antennas 4-1-1 to 4-1 -M are fast Fourier transformed by M fast Fourier transformers 4-2-1 to 4-2 -M. Is converted from a signal in the time domain into a signal in the frequency domain, that is, for each subcarrier.

  The initial transfer coefficient estimation circuit 4-3 uses the received signal sequence corresponding to the pilot signal in the output signal sequence of the M fast Fourier transformers 4-2-1 to 4-2M to use the initial value of the transfer coefficient matrix. Is estimated. The transfer coefficient estimation circuit 4-4 estimates a transfer coefficient using a transmission signal replica generated from the decoded received signal sequence and a transfer coefficient matrix history. The interference canceller 4-5 performs interference cancellation of the data signal using the estimated initial value, and estimates signal points for each subcarrier and each transmission antenna from the spatially multiplexed received signal. The estimated K system signal points are input to K demodulators 4-6-1 to 4-6-K, respectively, and decoded into bit sequences corresponding to the signal points. The decoded signal sequence is input to the transfer coefficient estimation circuit 4-4 to create a replica of the transmission signal.

  Next, FIG. 2 is a block diagram showing a configuration of the transfer coefficient estimation circuit 4-4 in the first embodiment. The transfer coefficient estimation circuit 4-4 includes K remodulators 4-4-1-1 to 4-4-1-K, a replica matrix generator 4-4-2, and a pseudo inverse matrix calculator 4-4- 3, a reception signal matrix generator 4-4-4, a matrix multiplier 4-4-5, a transfer coefficient estimation matrix storage circuit 4-4-6, and a transfer coefficient estimation matrix update circuit 4-4-7.

  For the signal series demodulated by the K demodulators 4-6-1 to 4-6-K, a replica of the transmission signal for each subcarrier is generated using the same operation as that on the OFDM signal transmitter side. That is, although not included in the block diagram of the first embodiment, a function to perform the same operation when operations such as error correction coding and interleaving are included in the first embodiment. It shall have. In addition, when the radio section is a time division multiplex system, there is no need to transmit a signal in the middle of processing a received signal. Therefore, in a transceiver-integrated system, this remodulator is an OFDM signal transmitter. A similar block included in can be substituted. Further, the signal point of the interference canceller output can be directly used as a replica of the transmission signal as in the prior art. In this case, the K demodulators 4-6-1 to 4-6-K It becomes unnecessary.

The replicas of the subcarriers generated in the K remodulators 4-4-1-1 to 4-4-1-K are input to the replica matrix generator 4-4-2. The replica corresponding to the transmission signal transmitted from the subcarrier i at time d + 1 by the transmission antenna 3-5-k is t _rep (i, k, d + 1), and the transmission signal vector at time d + 1, subcarrier i is t _rep ( i, d + 1)

It can be expressed as Here, () ^T represents transposition of a matrix. The replica matrix generator 4-3-2 groups transmission signal vectors of L columns (L is an integer equal to or greater than M) for each subcarrier, and generates a replica matrix T _rep (i).

The pseudo inverse matrix calculator 4-4-3 calculates (T _rep (i)) ⁺ the pseudo inverse matrix of the replica matrix.

_{Here, t 'rep (i, l} , k) is set to _(T rep ^{_(i)) +} a l rows k-th component pseudo inverse matrix _(T rep ^{_(i)) +} is the right pseudo inverse matrix,
(T _rep (i)) ^H · (T _rep (i) · (T _rep (i)) ^{H −1} . However, (A) ^H represents Hermitian transformation of matrix A, and (A) ⁺ represents an inverse matrix of matrix A.

The output of the fast Fourier transform of the received signal is input to the received signal matrix generator 4-3-4. The fast Fourier transform output of the received signal received from the subcarrier i by the receiving antenna 4-1-k at time d + 1 is y (i, k, d + 1), and the received signal vector at time d + 1, subcarrier i is y _rep. If (i, d + 1),

  It can be expressed as The reception signal matrix generator 4-4-4 groups transmission signal vectors of L columns (L is an integer greater than or equal to M) for each subcarrier, and generates a reception signal matrix Y (i) as shown in the following equation. .

  The relationship between the reception signal and the transmission signal is expressed as Equation (2). If the replica matrix is correct, the transmission signal vector and the reception signal vector are replaced with the replica matrix and the reception signal matrix, respectively. 15).

Here, N (i) represents a noise matrix of M rows and L columns.
The matrix multiplier 4-4-5 multiplies the pseudo-inverse matrix of the replica matrix from the right side of both sides of Equation (15), thereby replica estimated transfer coefficient matrix H ′ _rep (the estimated value of the transfer coefficient matrix at time d + L). i).

The obtained replica estimated transfer coefficient matrix H ′ _rep (i) is input and stored in the transfer coefficient estimated matrix storage circuit 4-4-6. In addition, the transfer coefficient matrix storage circuit 4-4-6 receives the initial estimation result of the transfer coefficient matrix performed by using the pilot signal added at the head of the burst in the initial transfer coefficient estimation circuit 4-3.

  The transfer coefficient estimation matrix update circuit 4-4-7 uses the replica estimated transfer coefficient matrix stored in the transfer coefficient estimation matrix storage circuit 4-4-6 and the transfer coefficient estimation matrix estimated in the past to Are determined and output to the interference canceller 4-5.

As an operation example of the transfer coefficient estimation matrix update circuit 4-4-7, a transfer coefficient estimation method using a forgetting coefficient using only the currently estimated replica estimated transfer coefficient will be described. Assuming that the forgetting factor is μ (μ takes a value between 0 and 1), the past transfer coefficient estimation matrix stored in the transfer coefficient estimation matrix storage circuit 4-4-6 is updated to H ′ _OLD (i). Assuming that the transfer coefficient estimation matrix is H ′ _NEW (i), it is given by Equation (17).

  The value of the forgetting factor μ may be an optimum value for the fluctuation speed of the propagation path. If μ = 0, since the past transfer coefficient estimation matrix is not reflected at all, it is not necessary to store the transfer coefficient estimation matrix estimated in the past in the transfer coefficient estimation matrix storage circuit 4-4-6.

  The updated transfer coefficient estimation matrix is output to the interference canceller 4-5 and the transfer coefficient estimation matrix storage circuit 4-4-6.

  Even in the above configuration, although the inverse matrix operation is included as in the conventional technique, since the inverse matrix operation is performed for each subcarrier, the size of each matrix is smaller than that in the conventional technique. The amount of calculation can be reduced as compared with the prior art, and the circuit scale can be reduced.

B. Second Embodiment (Claim 2)
Next, a second embodiment of the present invention will be described. In the second embodiment, in the first embodiment, the pseudo inverse matrix calculator 4-4-3 is provided with a function of obtaining an inverse matrix of a matrix of K rows and K columns. Compared with the case of using a pseudo-inverse matrix of a replica matrix, when using an inverse matrix, the number of replica matrices that can be generated per unit period can be increased, and the followability to propagation path fluctuations can be improved. become. Further, the amount of calculation can be reduced in the inverse matrix operation of K rows and K columns compared to the pseudo inverse matrix operation of K rows and L columns.

C. Third Embodiment (Claim 3)
Next, a third embodiment of the present invention will be described. In the third embodiment, the transfer coefficient estimated value update circuit 4-4-7 is changed from the first and second embodiments described above. Here, FIG. 3 is a block diagram showing a configuration of a transfer coefficient estimated value update circuit according to the third embodiment. In the figure, the transfer coefficient estimation matrix update circuit 5 includes a matrix weight calculation circuit 5-1, a weight coefficient storage circuit 5-2, and a matrix weight synthesis circuit 5-3.

  In the first and second embodiments, the weighted composition of the two matrices of the past transfer coefficient estimation matrix and the current replica estimated transfer coefficient matrix is performed using the default forgetting coefficient μ. The estimation accuracy of the matrix varies depending on the inverse matrix of the replica matrix used when the matrix is derived and the pattern of the received signal matrix.

  Therefore, in the third embodiment, the square sum of each element of the inverse matrix of the replica matrix is used in addition to the forgetting factor μ when performing weighted synthesis. Considering the noise component included in the replica estimated transfer coefficient matrix, the matrix term of the noise component is given by the pseudo inverse matrix of the replica matrix or the product of the inverse matrix and the noise matrix from Equation (16). The larger the sum of squares of the elements included in the coefficient matrix, the more the noise is increased. Conversely, the smaller the sum is, the more the noise is suppressed. Therefore, the matrix weighting synthesis circuit 5-3 updates the transfer coefficient estimation matrix using the following equation instead of the equation (17).

Here, p _imat (i) is the pseudo-inverse of the replica matrix or the sum of squares of all elements of the inverse matrix, and is given by the following equation.

() ^* Represents a complex conjugate. When an inverse matrix operation circuit is used instead of the pseudo inverse matrix operation circuit 4-4-3, L = K.

In addition, w (p _imat (i)) in Expression (18) is a function that gives a weight to the input p _imat (i). In general, when p _imat (i) is large, noise tends to be amplified, so that the estimation accuracy of the replica estimation transfer coefficient matrix is low, and therefore the reliability of the updated transfer coefficient estimation matrix is low. Set smaller. Conversely, when p _imat (i) is small, noise tends to be suppressed, so a large weight is set. p _imat (i) and μw (p _imat (i)) are calculated in the matrix weight calculation circuit 5-1.

The calculated matrix weight μw (p _imat (i)) is output to the weight coefficient storage circuit 5-2, and is used when the weighting synthesis of the transfer coefficient estimation matrix is performed in the matrix weighting synthesis circuit 5-3.

  According to the third embodiment described above, since the transfer coefficient matrix update using the replica estimated transfer coefficient matrix is executed by reflecting different accuracy for each subcarrier, it is possible to realize characteristic improvement.

D. Fourth Embodiment (Claim 4)
Next, a fourth embodiment of the present invention will be described. In the fourth embodiment, the weight coefficient for the replica estimated transfer coefficient matrix calculated in the matrix weight calculation circuit 5-1 is proportional to the reciprocal of p _imat (i), that is, with respect to predetermined positive constants C and α. The coefficient is given by C × p _imat (i) ^−α .

E. Fifth embodiment (Claim 5)
Next, a fifth embodiment of the present invention will be described. FIG. 4 is a block diagram showing the configuration of the transfer coefficient estimation matrix update circuit 6 according to the fifth embodiment. The transfer coefficient estimated value update circuit 6 includes a matrix weight calculation circuit 6-1, a weight coefficient storage circuit 6-2, a matrix weight synthesis circuit 6-3, a matrix weight adder 6-4, a transfer coefficient estimation matrix update determination circuit 6- It is composed of five. The transmission coefficient estimated value update circuit 6 differs from the transmission coefficient estimation matrix update circuit 5 of the third embodiment described above in that it includes a matrix weight adder 6-4 and a transmission coefficient estimation matrix update determination circuit 6-5. .

The matrix weight adder 6-4 stores the replica estimated transfer coefficient estimation for each subcarrier from the last update of the transfer coefficient estimation matrix stored in the weight coefficient storage circuit 4-3-9-2. Calculate the sum of weight coefficient history p _{inv_mat_sum} (i) for the matrix.

Here, p _imat (i, b) is p _imat (i) at the time of derivation of the replica estimated transfer coefficient matrix estimated b times before for the subcarrier i (b = 0 is currently estimated) B (i) represents the number of replica estimated transfer coefficient matrices estimated from the last update of the transfer coefficient estimation matrix to the present.

When the fluctuation of the propagation path is sufficiently gradual, it is possible to improve the accuracy of the combined transmission coefficient estimation matrix by weighted synthesis of a plurality of successive replica estimation transmission coefficient matrices. p _{inv_imat_sum} (i) is an amount that reflects the average SNR of the entire matrix after combining all the past replica estimated transfer coefficient matrices. _Therefore, in order to maintain the SNR after the weighted combining at a predetermined level, p _{Inv_imat_sum} (i) determines whether or not (i) exceeds a predetermined threshold value in transfer coefficient estimation matrix update determination circuit 4-3-9-5, and outputs the determination result to matrix weighting synthesis circuit 6-5. .

In the weighting synthesis circuit 4-3-9-3, a replica is determined for a subcarrier that is determined by the transfer coefficient estimation matrix update determination circuit 6-5 that p _{inv_imat_sum} (i) exceeds a predetermined threshold. The estimated transfer coefficient matrix and the transfer coefficient estimation matrix used immediately before are weighted and combined and output to the interference canceller. For _{subcarriers for which} p _{inv_imat_sum} (i) is determined not to exceed a predetermined threshold, a past transfer coefficient estimation matrix input from the transfer coefficient estimation matrix storage circuit 4-4-6 is output. .

  As an example of weighting synthesis, there is a method of weighting and synthesizing the replica estimated transmission coefficient matrix and then weighting and combining the weighted synthesis result and the immediately preceding transmission coefficient estimation matrix using a forgetting coefficient.

Here, H ′ _rep (i, b) is a replica estimated transfer coefficient estimation matrix estimated b times before.
Incidentally, equations (20) and Equation 21 is an example, as shown in description of the fourth _{embodiment, p} IMAT (i, b) respectively may be replaced _p IMAT (i, b) ^alpha and.

F. Sixth embodiment (Claim 6)
Next, a sixth embodiment of the present invention will be described. FIG. 5 is a block diagram showing a configuration of the transfer coefficient estimation matrix update circuit 7 according to the sixth embodiment. The transfer coefficient estimated value update circuit 7 includes a matrix weight calculation circuit 7-1, a weight coefficient storage circuit 7-2, a matrix weight synthesis circuit 7-3, a matrix weight adder 7-4, and a transfer coefficient estimation matrix update determination circuit 7-. 5, a transfer coefficient estimation matrix non-update period calculation circuit 7-6, and a transfer coefficient matrix history information control circuit 7-7. The transfer coefficient estimated value update circuit 7 is different from the transfer coefficient estimated value update circuit 6 of the fifth embodiment described above in that the transfer coefficient estimated value non-update period calculation circuit 7-6 and the transfer coefficient matrix history information control circuit 7-7. It is different in having.

  The determination result of the transfer coefficient matrix update determination circuit is input to the transfer coefficient estimated value non-update period calculation circuit 7-6. The transmission coefficient estimated value non-update period calculation circuit 7-6 calculates a period from the time when the previous transmission coefficient estimation matrix is updated to the current time by referring to the determination result for each subcarrier. For example, in the case of OFDM-based processing, it is a simple method to count the period in which the transfer coefficient estimation matrix is not updated in units of OFDM symbols.

  The transmission coefficient matrix history information control circuit 4-3-9-7 determines whether or not the non-update period of the transmission coefficient estimation matrix for all subcarriers exceeds a predetermined threshold value. For subcarriers that exceed the predetermined threshold, it is assumed that all or part of the information of the replica estimated transfer coefficient matrix to be weighted and combined does not sufficiently reflect the current transfer coefficient matrix. An instruction to delete the history information of the replica estimated transfer coefficient matrix is output to the transfer coefficient estimated matrix storage circuit 4-4-6 and the weighting synthesis circuit 7-2.

  The replica estimated transfer coefficient matrix storage circuit 7-6 and the weighting composition circuit 7-2, which have received the history information erasure instruction from the transfer coefficient matrix history information control circuit 7-7, store the replica estimated transfer coefficient estimation matrix and weight stored therein. Erase part or all of a coefficient.

  As a result, it is possible to reduce the influence of the replica estimated transfer coefficient matrix that does not sufficiently reflect the current transfer coefficient information, and to improve the transfer coefficient estimation accuracy.

G. Seventh embodiment (Claims 7, 8, and 9)
Next, a seventh embodiment of the present invention will be described. FIG. 6 is a block diagram showing a configuration of the transfer coefficient estimation circuit 8 according to the seventh embodiment. The transfer coefficient estimation circuit 8 includes K remodulators 8-1-1 to 8-1 -K, a replica matrix generator 8-2, a pseudo inverse matrix calculator 8-3, and a received signal matrix generator 8-4. , A matrix multiplier 8-5, a transfer coefficient estimation matrix storage circuit 8-6, a transfer coefficient estimation matrix update circuit 8-7, and a replica estimated transfer coefficient matrix filter 8-8. The transfer coefficient estimation circuit 8 differs from the first to sixth embodiments in that it includes a replica estimated transfer coefficient matrix filter 8-8.

FIG. 7 is a block diagram showing an internal configuration of the replica estimated transfer coefficient matrix filter 8-8. The replica estimated transfer coefficient matrix filter 8-8 includes a matrix subtractor 8-8-1, a matrix Euclidean distance calculation circuit 8-8-2, and a replica estimated transfer coefficient matrix selection circuit 8-8-3. In the matrix subtractor 8-8-1, the currently estimated replica estimated transfer coefficient matrix that is the output from the matrix multiplier 4-4-5 and the output immediately before the output of the transfer coefficient estimated matrix storage circuit 4-4-6 The difference from the updated transfer coefficient estimation matrix, that is, H ′ _OLD (i) −H ′ _rep (i) is calculated for each subcarrier and output as a difference matrix.

  The matrix Euclidean distance calculation circuit 8-8-2 calculates the squares of the Euclidean distances of all the elements constituting the difference matrix, and outputs them to the replica estimated transfer coefficient matrix selection circuit 8-8-3. In the replica estimated transfer coefficient matrix selection circuit 8-8-3, for the subcarrier whose sum of squares of the input Euclidean distance exceeds a predetermined threshold, the accuracy of the replica estimated transfer coefficient matrix of the subcarrier is determined. Is not output to the transfer coefficient estimation matrix storage circuit 4-4-6, and when the square sum of the input Euclidean distance is equal to or less than a predetermined threshold value, the replica estimated transfer coefficient of the subcarrier It is determined that the accuracy of the matrix is sufficiently obtained, and the matrix is output to the replica estimated transfer coefficient matrix storage circuit 4-4-6.

  Depending on noise and the pseudo inverse matrix of the replica matrix, the calculated replica estimated transfer coefficient matrix may not have sufficient accuracy, and the estimation accuracy of the transfer coefficient estimation matrix due to the subsequent weighted synthesis result may decrease. By using the seventh embodiment, it is possible to prevent the replica estimated transfer coefficient estimation matrix with low estimation accuracy from being weighted and synthesized in advance.

  In this example, the matrix Euclidean distance calculation circuit calculates the squares of the Euclidean distances of all the elements constituting the difference matrix. However, in general, a geometric distance other than the Euclidean distance may be used. I do not care. For example, it may be a power of Euclidean distance. Furthermore, in the next eighth embodiment, the Manhattan distance is used instead of the Euclidean distance.

H. Eighth Embodiment (Claim 10)
Next, an eighth embodiment of the present invention will be described. In the eighth embodiment, a matrix Manhattan distance calculation circuit is used instead of the matrix Euclidean distance calculation circuit 8-8-2 in the seventh embodiment. In the matrix Manhattan distance calculation circuit, the entire difference matrix of Manhattan distances (sum of absolute values of the real part and the imaginary part) of all elements constituting the input difference matrix H ′ _OLD (i) −H ′ _rep (i) And is output to the replica estimated transfer coefficient matrix selection circuit 4-3-10-3. The replica estimated transfer coefficient matrix selection circuit 4-3-10-3 determines whether the threshold is the same as the sum of the Euclidean distances in the seventh embodiment described above with respect to the sum of Manhattan distances of the input difference matrix. And whether or not to output the replica estimated transfer coefficient is selected.

  According to the eighth embodiment, compared to the above-described seventh embodiment, since the Manhattan distance is used instead of the Euclidean distance, the calculation for comparing the sum of the scalar amounts of all the elements of the difference matrix is simplified. Thus, the circuit can be simplified.

I. Ninth Embodiment (Claims 11, 12, 13)
Next, a ninth embodiment of the present invention will be described. FIG. 8 is a block diagram showing the configuration of the transfer coefficient estimation circuit 9 according to the ninth embodiment. The transfer coefficient estimation circuit 9 includes K remodulators 9-1-1 to 9-1-K, a replica matrix generator 9-2, a pseudo inverse matrix calculator 9-3, and a received signal matrix generator 9-4. , Matrix multiplier 9-5, transfer coefficient estimation matrix storage circuit 9-6, transfer coefficient estimation matrix update circuit 4-3-7, replica estimation transfer coefficient estimation matrix filter 9-8, transfer coefficient estimation matrix filter 9-9 Composed. The transfer coefficient estimation circuit 9 differs from the above-described seventh embodiment in that it includes a replica estimated transfer coefficient matrix filter 9-9.

  FIG. 9 is a block diagram showing a configuration of a transfer coefficient matrix filter 9-9 according to the ninth embodiment. The transfer coefficient matrix filter 9-9 includes a matrix subtractor 9-9-1, a matrix Euclidean distance calculation circuit 9-9-2, and a transfer coefficient estimation matrix selection circuit 9-9-3. In the seventh embodiment described above, the filtering is performed on the replica estimated transfer coefficient matrix. However, the purpose of the ninth embodiment is to perform the filtering on the updated transfer coefficient estimating matrix.

In the matrix multiplier 4-3-11-1, the updated transfer coefficient estimation matrix that is an output from the transfer coefficient matrix update circuit 9-7 and the previous transfer that is the output of the transfer coefficient estimation matrix storage circuit 9-6. The difference from the coefficient estimation matrix, that is, H ′ _OLD (i) −H ′ _NEW (i) is calculated for each subcarrier and output as a difference matrix. The matrix Euclidean distance calculation circuit 9-9-2 calculates the squares of the Euclidean distances of all elements constituting the difference matrix and outputs the squares to the transfer coefficient estimation matrix selection circuit 9-9-3.

In the transfer coefficient estimation matrix selection circuit 9-9-3, for the subcarrier whose sum of squares of the input Euclidean distance exceeds a predetermined threshold, the transfer coefficient estimation matrix updated in the subcarrier When it is determined that the accuracy is low, and the interference coefficient canceller outputs a previously estimated transfer coefficient estimation matrix H ′ _OLD (i), and the square sum of the input Euclidean distance is less than a predetermined threshold Then, it is determined that the accuracy of the transmission coefficient estimation matrix of the subcarrier is sufficiently obtained, and the transmission coefficient estimation matrix H ′ _NEW (i) updated at the current time is output to the interference canceller.

  Depending on noise or the pseudo inverse matrix of the replica matrix, the calculated replica estimation transfer coefficient matrix may not have sufficient accuracy, and the estimation accuracy of the transfer coefficient estimation matrix may decrease. Thus, it is possible to prevent a transfer coefficient estimation matrix with low estimation accuracy from being output.

J. et al. Tenth Embodiment (Claim 14)
Next, a tenth embodiment of the present invention will be described. In the tenth embodiment, a matrix Manhattan distance calculation circuit is used instead of the matrix Euclidean distance calculation circuit 9-9-2 in the ninth embodiment. In the matrix Manhattan distance calculation circuit, the entire difference matrix of Manhattan distances (sum of absolute values of real part and imaginary part) of all elements constituting the input difference matrix H ′ _OLD (i) −H ′ _NEW (i) And output to the transfer coefficient estimation matrix selection circuit 9-9-3. The replica estimation transfer coefficient matrix selection circuit determines the magnitude of the threshold with respect to the sum of Manhattan distances of the input difference matrix in the same manner as the sum of Euclidean distances in the ninth embodiment, and updates the transfer coefficient estimation matrix currently updated. Or a transfer coefficient estimation matrix estimated in the past is selected.

  In the tenth embodiment, the Manhattan distance is used instead of the Euclidean distance as compared with the ninth embodiment described above, so that the calculation for comparing the sum of the scalar amounts of all the elements of the difference matrix becomes simple. The circuit can be simplified.

  Although the ninth embodiment has been described as the Euclidean distance or the Manhattan distance, other geometric distances may be used. Specifically, it may be a power of Euclidean distance or a power of Manhattan distance.

K. Eleventh embodiment (Claim 15)
Next, an eleventh embodiment of the present invention will be described. FIG. 10 is a block diagram showing a configuration of the transfer coefficient estimation matrix update circuit 10 according to the eleventh embodiment. The transfer coefficient estimation matrix update circuit 10 includes a vector weight calculation circuit 10-1, a weight coefficient storage circuit 10-2, and a vector weight synthesis circuit 10-3. The transfer coefficient estimation matrix update circuit 10 replaces the transfer coefficient estimation matrix update circuit 4-4-7 described above.

In the third embodiment described above, an example in which the sum of squares of each element of the inverse matrix of the replica matrix is used in addition to the forgetting factor μ when performing weighting synthesis is described. However, a noise component included in the replica estimated transfer coefficient matrix is considered. Since the matrix term of the noise component is given by the pseudo inverse matrix of the replica matrix or the product of the inverse matrix and the noise matrix from Equation (16), the larger the square sum of the elements included in the replica estimated transfer coefficient matrix, the larger As the noise increases, the smaller the noise, the more the noise is suppressed. Here, if a column vector constituting the k-th column (k is an integer of 1 or more and K or less) of the pseudo inverse matrix of the replica matrix is expressed as t ′ _rep (i, k), the pseudo inverse matrix of the replica matrix is It is expressed as Equation (22).

  Also, the noise matrix N (i) is defined as in Equation (23).

  Using the definitions of Equation (22) and Equation (23), the noise term in Equation (16) is expressed as:

It becomes. The component of m rows and k columns of Equation (24) is

It becomes. Assuming that each noise component has the same variance σ ² , the variance of the component of m rows and k columns of Equation (24) is

It becomes. From Equation (26), it can be seen that the variance of the noise component of m rows and k columns in Equation (24) depends only on k. That is, it can be seen that the variance of the noise term included in each element of the replica estimated transfer coefficient matrix in Equation (16) is the same even if the rows are different if the columns are the same.

  In the eleventh embodiment, the transfer coefficient estimation matrix is not updated for the entire matrix as in the third and fourth embodiments described above, but individually for each column vector constituting the transfer coefficient estimation matrix of each subcarrier. Take the technique to do. As an example of weighting synthesis, consider the following equation (27).

Here, h ′ _OLD (i, k) is the column vector of the kth column of _H′OLD (i), _h′OLD (i, k) is the column vector of the kth column of _H′OLD (i), h ' _rep (i, k) represents a column vector of the k-th column of H' _rep (i).

_First , the vector weight calculation circuit 10-1 calculates the square sum p _{imat_col} (i, k) for each column vector of the k-th column constituting the pseudo inverse matrix (or inverse matrix) of the replica matrix.

Obtained _{p imat_col} (i, k) with respect to the weight calculated for the k-th column vector of subcarrier i by using the default function _{w col (p imat_col (i,} k)), the weighting coefficient storage circuit Output to 4-3-9-2. Since p _{imat_col} (i, k) is a term proportional to the variance of noise contained in each element of the k-th column of the replica estimated transfer coefficient matrix, when p _{imat_col} (i, k) is small, replica estimated transfer Since the coefficient matrix has high reliability, w _col (p _{imat_col} (i, k)) is large, and when p _{imat_col} (i, k) is large, the reliability of the replica estimated transfer coefficient matrix is low, so w _col (P _{imat_col} (i, k)) is reduced.

  The vector weighting synthesis circuit 10-3 weights and synthesizes the past transfer coefficient estimation matrix and the replica estimated transfer coefficient matrix as the current transfer coefficient estimation matrix, and the interference canceller 4-4 and the transfer coefficient estimation matrix storage circuit 4-4. Output to -6. In the vector weighting synthesis circuit, the weighting coefficient input from the weighting coefficient storage circuit 10-2 is used to perform weighting synthesis for each column vector of the transmission coefficient estimation matrix based on Equation (27), and the transmission coefficient estimation matrix is updated. I do.

  As a result, in the eleventh embodiment, weighting synthesis can be performed in consideration of different transmission coefficient estimation accuracy for each vector, so that a more accurate transmission coefficient estimation matrix can be obtained.

L. Twelfth embodiment (Claim 16)
Next, a twelfth embodiment of the present invention will be described. In the twelfth embodiment, the weight coefficient for each column vector of the replica estimated transfer coefficient matrix calculated in the vector weight calculation circuit 10-1 in the eleventh embodiment described above is proportional to the reciprocal of p _{imat_col} (i), for example, predetermined For the positive constants C and α, a weighting coefficient is given by C × p _{imat_col} (i) ^−α .

M.M. Thirteenth Embodiment (Claim 17)
Next, a thirteenth embodiment of the present invention is described. FIG. 11 is a block diagram showing a configuration of a transfer coefficient estimated value update circuit in the thirteenth embodiment. The transfer coefficient estimated value update circuit 11 includes a vector weight calculation circuit 11-1, a weight coefficient storage circuit 11-2, a vector weight synthesis circuit 11-3, a vector weight adder 11-4, a transfer coefficient estimation matrix sequence vector update determination circuit. 4-3-9-5. The thirteenth embodiment differs from the twelfth embodiment described above in that it includes a vector weight adder 11-4 and a transfer coefficient estimation matrix update determination circuit 11-5. The vector weight adder 11-4 stores the weight for each column vector of the replica estimated transfer coefficient estimation matrix from the last update of the transfer coefficient estimation matrix stored in the weight coefficient storage circuit 11-2 to the present. The sum of coefficient histories p _{inv_imat_col_sum} (i, k) is calculated.

Here, p _{imat_col} (i, k, b) is p _{imat_col} at the time of derivation of the replica estimated transfer coefficient matrix estimated b times before the column vector of the kth column of the replica estimated transfer coefficient matrix of subcarrier i. (I, k) (b = 0 represents the currently estimated replica estimated transfer coefficient matrix), and B (i, k) is the current value after the column vector of the transfer coefficient estimated matrix is updated last time. It is the number of column vectors of the replica estimated transfer coefficient matrix estimated so far.

When the fluctuation of the propagation path is sufficiently gradual, it is possible to improve the accuracy of the combined transmission coefficient estimation matrix by weighted synthesis of a plurality of successive replica estimation transmission coefficient matrices. p _{inv_imat_col_sum} (i, k) is an amount proportional to the SNR after the maximum ratio combining of the column vectors of all past replica estimated transfer coefficient matrices, and thus maintains the SNR after the weighted combining at a predetermined level. _Therefore , it is determined in the transfer coefficient estimation matrix update determination circuit 11-5 whether or not p _{inv_imat_col_sum} (i, k) exceeds a predetermined threshold, and the determination result is sent to the vector weighting synthesis circuit 11-3. Output.

In the weighting synthesis circuit 11-3, a replica is determined for a subcarrier that has been determined by the transfer coefficient estimation matrix sequence vector update determination circuit 11-5 that p _{inv_imat_col_sum} (i, k) exceeds a predetermined threshold. The estimated transfer coefficient matrix and the column vector of the transfer coefficient estimation matrix used immediately before are weighted and combined and output to the interference canceller. For the column vector determined that p _{inv_imat_col_sum} (i, k) does not exceed the predetermined threshold value, the input from the transfer coefficient estimation matrix storage circuit 4-4-6 is output as it is.

  As an example of weighting synthesis, there is a method of performing weighting synthesis using a forgetting coefficient on the weighted synthesis result and the immediately preceding transmission coefficient estimation matrix after synthesizing the replica estimated transmission coefficient matrix at the maximum ratio.

Here, h ′ _rep (i, k, b) is the column vector of the replica estimated transfer coefficient estimation matrix estimated b times before.

The determination result in the transfer coefficient matrix sequence vector update determination circuit 11-5 is also output to the transfer coefficient estimation matrix storage circuit 4-4-6 and the weight coefficient storage circuit 11-2, and p _{inv_imat_col_sum} (i, k For a column vector that is determined to exceed a predetermined threshold value, the history of the replica estimated transfer coefficient and the weight coefficient regarding the column vector is erased, thereby realizing efficient use of the storage circuit.

In addition, numerical formula (29) and numerical formula (30) are one example, and p _{imat_col} (i, k, b) is changed to p _{imat_col} (i, k, b) as shown in the description of the twelfth embodiment. ^It may be replaced with ^α .

N. Fourteenth embodiment (Claim 18)
Next, a fourteenth embodiment of the present invention is described. FIG. 12 is a block diagram showing a configuration of a transfer coefficient estimated value update circuit in the fourteenth embodiment. The transfer coefficient estimated value update circuit 12 includes a vector weight calculation circuit 12-1, a weight coefficient storage circuit 12-2, a vector weight synthesis circuit 12-3, a vector weight adder 12-4, a transfer coefficient estimation matrix sequence vector update determination circuit. 12-5, a transfer coefficient estimation matrix sequence vector non-update period calculation circuit 12-6, and a transfer coefficient matrix history information control circuit 12-7. The fourteenth embodiment is different from the eleventh embodiment in that a transmission coefficient estimation matrix sequence vector non-update period calculation circuit 12-6 and a transmission coefficient matrix history information control circuit 12-7 are provided.

  The determination result of the transfer coefficient matrix column vector update determination circuit is input to the transfer coefficient estimation matrix column vector non-update period calculation circuit 12-6. The transfer coefficient estimation matrix column vector non-update period calculation circuit 12-6 calculates a period from the time when the transfer coefficient estimation matrix was last updated to the current time by referring to the determination result for each column vector. For example, in the case of OFDM-based processing, it is a simple method to count the period in which the transfer coefficient estimation matrix is not updated in units of OFDM symbols.

  The transfer coefficient matrix history information control circuit 12-7 determines whether or not the non-update period exceeds a predetermined threshold value for each column vector of the transfer coefficient estimation matrix for all subcarriers. For column vectors that exceed the default threshold, it is assumed that all or part of the information of the replica estimated transfer coefficient matrix to be weighted and combined does not sufficiently reflect the current transfer coefficient matrix. An instruction to delete the history information of the replica estimated transfer coefficient matrix is output to the replica estimated transfer coefficient matrix storage circuit 4-4-6 and the weight coefficient storage circuit 12-2.

  The transfer coefficient estimation matrix storage circuit 4-4-6 and the weight coefficient storage circuit 12-2, which have received the history information erasure instruction from the transfer coefficient matrix history information control circuit 12-7, The stored replica estimation transfer coefficient estimation matrix and part or all of the weight coefficients are deleted.

  According to the fourteenth embodiment, it is possible to reduce the influence of the replica estimated transfer coefficient matrix that does not sufficiently reflect the current transfer coefficient information in units of column vectors, and to improve the transfer coefficient estimation accuracy.

O. Fifteenth embodiment (claims 19, 20, 21)
Next, a fifteenth embodiment of the present invention is described. FIG. 13 is a block diagram showing the configuration of the transfer coefficient estimation circuit in the fifteenth embodiment. The transfer coefficient estimation circuit 13 includes K remodulators 13-1-1-1 to 13-1-K, a replica matrix generator 13-2, a pseudo inverse matrix calculator 13-3, and a received signal matrix generator 13-4. , A matrix multiplier 13-5, a transmission coefficient estimation matrix storage circuit 13-6, a transmission coefficient estimation matrix update circuit 13-7, and a replica estimation transmission coefficient estimation matrix column vector filter 13-8. The fifteenth embodiment differs from the fourteenth embodiment described above in that it includes a replica estimated transfer coefficient matrix column vector filter 13-8.

FIG. 14 is a block diagram showing an internal configuration of the replica estimated transfer coefficient matrix sequence vector filter 13-8. The replica estimated transfer coefficient matrix column vector filter 13-8 includes a matrix subtractor 13-8-1, a vector Euclidean distance calculation circuit 13-8-2, and a replica estimated transfer coefficient matrix string vector selection circuit 13-8-3. The In the matrix subtractor 13-8-1, the replica estimated transfer coefficient matrix that is the output from the matrix multiplier 13-5 and the transfer coefficient estimated matrix that has been updated immediately before that is the output of the transfer coefficient estimated matrix storage circuit 13-6. , That is, H ′ _OLD (i) −H ′ _rep (i) is calculated for each subcarrier and output as a difference matrix.

  In the matrix Euclidean distance calculation circuit 13-8-2, for each column vector of the difference matrix, the square sum of the Euclidean distances of all elements constituting the column vector is calculated, and the replica estimated transfer coefficient matrix column vector selection circuit 13-8 is calculated. To -3. In the replica estimated transfer coefficient matrix column vector selection circuit 13-8-3, the replica estimated transfer coefficient matrix of the subcarrier is determined for a column vector in which the square sum of the input Euclidean distance exceeds a predetermined threshold value. If the accuracy of the column vector is determined to be low and is not output to the replica estimated transfer coefficient matrix storage circuit, and the square sum of the input Euclidean distance is less than or equal to the predetermined threshold, the replica of the subcarrier It is determined that the accuracy of the column vector of the estimated transfer coefficient matrix is sufficiently obtained, and is output to the replica estimated transfer coefficient matrix storage circuit.

  Depending on noise and pseudo-inverse of the replica matrix, the calculated replica estimated transfer coefficient matrix may not be able to obtain sufficient accuracy, and the transfer coefficient estimation matrix estimation accuracy due to the weighted synthesis result at the latter stage may be reduced. By using the fifteenth embodiment, it is possible to prevent a replica estimated transfer coefficient estimation matrix with low estimation accuracy from being weighted and synthesized in units of column vectors.

P. Sixteenth Embodiment (Claim 22)
Next, a sixteenth embodiment of the present invention will be described. In the sixteenth embodiment, a vector Manhattan distance calculation circuit is used instead of the vector Euclidean distance calculation circuit 13-8-2 in the fifteenth embodiment. In the vector Manhattan distance calculation circuit, for every column vector of the input difference matrix H ′ _OLD (i) −H ′ _rep (i), the Manhattan distance (absolute real part and imaginary part) of all elements constituting the column vector (Sum of values) is calculated and output to the replica estimated transfer coefficient matrix column vector selection circuit 13-8-3. The replica estimated transfer coefficient matrix selection circuit 13-8-3 determines whether the threshold value is the same as the sum of the Euclidean distances in the embodiments of claims 19 to 21 with respect to the sum of the Manhattan distances of the input difference matrix. And whether or not to output the replica estimated transfer coefficient is selected.

  Compared with the fifteenth embodiment described above, the sixteenth embodiment uses the Manhattan distance instead of the Euclidean distance, and thus the number of multipliers can be reduced, and the circuit can be simplified. .

  Although the fifteenth embodiment has been described as the Euclidean distance or the Manhattan distance, other geometric distances may be used. Specifically, it may be a power of Euclidean distance or a power of Manhattan distance.

Q. Seventeenth embodiment (Claims 23, 24, 25)
Next, a seventeenth embodiment of the present invention is described. FIG. 15 is a block diagram showing the configuration of the transfer coefficient estimation circuit in the seventeenth embodiment. The transfer coefficient estimation circuit 14 includes K remodulators 14-1-1-1 to 14-1-K, a replica matrix generator 14-2, a pseudo inverse matrix calculator 14-3, and a received signal matrix generator 14-4. , Matrix multiplier 14-5, transfer coefficient estimation matrix storage circuit 14-6, transfer coefficient estimation matrix update circuit 14-7, replica estimation transfer coefficient estimation matrix string vector filter 14-8, transfer coefficient estimation matrix string vector filter 14- It is composed of nine. The seventeenth embodiment is different from the first embodiment described above in that a replica estimated transfer coefficient matrix filter 14-9 is included.

  FIG. 16 is a block diagram showing the configuration of the transfer coefficient estimation matrix sequence vector filter 14-9. The transfer coefficient matrix filter 14-9 includes a matrix subtractor 14-9-1, a vector Euclidean distance calculation circuit 14-9-2, and a transfer coefficient estimation matrix sequence vector selection circuit 14-9-3. In the thirteenth to fifteenth embodiments, filtering for each column vector is performed on the replica estimated transfer coefficient matrix, but in the seventeenth embodiment, filtering on the updated transfer coefficient estimation matrix is performed on a column vector basis. The purpose is to do.

In the matrix subtractor 14-9-1, the updated transfer coefficient estimation matrix output from the transfer coefficient matrix update circuit 14-7 and the output of the transfer coefficient estimation matrix storage circuit 4-4-6, The difference from the transfer coefficient estimation matrix, that is, H ′ _OLD (i) −H ′ _NEW (i) is calculated for each subcarrier and output as a difference matrix. The vector Euclidean distance calculation circuit 14-9-2 calculates the sum of the squares of the Euclidean distances of all elements constituting the column vector of the difference matrix and outputs the sum to the transfer coefficient estimation matrix column vector selection circuit 14-9-3.

In the transfer coefficient estimation matrix column vector selection circuit 14-9-3, for a column vector whose square sum of input Euclidean distances exceeds a predetermined threshold, transfer coefficient estimation updated in the column vector is performed. It is determined that the accuracy of the matrix is low, and the column vector of the transfer coefficient estimation matrix H ′ _OLD (i) estimated in the past is output to the interference canceller 4-5 and input to the interference canceller 4-5. If the sum of squares of the Euclidean distance is equal to or less than a predetermined threshold value, it is determined that the accuracy of the transfer coefficient estimation matrix of the subcarrier is sufficiently obtained, and the transfer coefficient estimation matrix H ′ updated at the current time is determined. _The column vector of _NEW (i) is output to the interference canceller 4-5.

  Depending on noise or the like and the pseudo inverse matrix of the replica matrix, the calculated replica estimation transfer coefficient matrix may not have sufficient accuracy, and the estimation accuracy of the transfer coefficient estimation matrix may decrease, but this 17th embodiment is used. Thus, it is possible to prevent a transfer coefficient estimation matrix with low estimation accuracy from being output in units of column vectors.

R. Eighteenth embodiment (Claim 26)
Next, an eighteenth embodiment of the present invention will be described. In the eighteenth embodiment, a vector Manhattan distance calculation circuit is used instead of the vector Euclidean distance calculation circuit 14-9-2 in the seventeenth embodiment. In the matrix Manhattan distance calculation circuit, the Manhattan distance (sum of absolute values of the real part and the imaginary part) of all elements constituting the column vector of the input difference matrix H ′ _OLD (i) −H ′ _NEW (i) The sum is calculated and output to the transfer coefficient estimation matrix sequence vector selection circuit 14-9-3. The transfer coefficient estimation matrix column vector selection circuit 14-9-3 determines the magnitude of the threshold value for the sum of the Manhattan distances of the input difference matrix as in the fifteenth embodiment, and updates the current value. The transmission coefficient estimation matrix or the transmission coefficient estimation matrix estimated in the past is output.

  Compared to the fifteenth embodiment, the eighteenth embodiment uses the Manhattan distance instead of the Euclidean distance, so that the calculation for comparing the sum of the scalar quantities of the column vectors of the difference matrix is simplified. Simplification is possible.

  In addition, although the 17th embodiment has been described as the Euclidean distance or the Manhattan distance, other geometric distances may be used. Specifically, it may be a power of Euclidean distance or a power of Manhattan distance.

  In the above-described embodiment, the initial transfer coefficient estimation circuit 4-3, the transfer coefficient estimation circuit 4-4, and the like are executed in a computer system. A series of processing steps by the initial transfer coefficient estimation circuit 4-3 and the transfer coefficient estimation circuit 4-4 are stored in a computer-readable recording medium in the form of a program, and the computer reads the program. The above processing is performed by executing the above. That is, each processing means and processing unit in the initial transfer coefficient estimation circuit 4-3 and the transfer coefficient estimation circuit 4-4 is read by the central processing unit such as a CPU into the main storage device such as a ROM or RAM. This is realized by executing information processing / arithmetic processing.

  Here, the computer-readable recording medium means a magnetic disk, a magneto-optical disk, a CD-ROM, a DVD-ROM, a semiconductor memory, or the like. Alternatively, the computer program may be distributed to the computer via a communication line, and the computer that has received the distribution may execute the program.

1 is a block diagram showing an OFDM signal transmitting / receiving apparatus including a transfer coefficient estimation circuit according to a first embodiment of the present invention. It is a block diagram which shows the structure of the transmission coefficient estimation circuit 4-4 in this 1st Embodiment. It is a block diagram which shows the structure of the transmission coefficient estimated value update circuit by 3rd Embodiment of this invention. It is a block diagram which shows the structure of the transmission coefficient estimated value update circuit 6 by 5th Embodiment of this invention. It is a block diagram which shows the structure of the transmission coefficient estimation matrix update circuit 7 by 6th Embodiment of this invention. It is a block diagram which shows the structure of the transfer coefficient estimation circuit 8 by 7th Embodiment of this invention. It is a block diagram which shows the internal structure of the replica estimation transmission coefficient estimation matrix filter 8-8 of this 7th Embodiment. It is a block diagram which shows the structure of the transmission coefficient estimation circuit 9 by 9th Embodiment of this invention. It is a block diagram which shows the structure of the transmission coefficient matrix filter 9-9 by this 9th Embodiment. It is a block diagram which shows the structure of the transmission coefficient estimation matrix update circuit 10 by 11th Embodiment of this invention. It is a block diagram which shows the structure of the transmission coefficient estimated value update circuit in 13th Embodiment of this invention. It is a block diagram which shows the structure of the transmission coefficient estimated value update circuit in 14th Embodiment of this invention. It is a block diagram which shows the structure of the transmission coefficient estimation circuit in 15th Embodiment of this invention. It is a block diagram which shows the internal structure of the replica estimation transmission coefficient matrix sequence vector filter 13-8 of this 15th Embodiment. It is a block diagram which shows the structure of the transmission coefficient estimation circuit in 17th Embodiment of this invention. It is a block diagram which shows the structure of the transmission coefficient estimation matrix sequence vector filter 14-9 of this 17th Embodiment. It is a block diagram which shows the structural example of the OFDM signal transmission / reception apparatus by a prior art. It is a figure which shows an example of the burst signal in a prior art. It is a block diagram which shows the structure of the transfer coefficient estimation circuit by a prior art.

Explanation of symbols

4 OFDM signal receiving apparatus 4-1-1 to 4-1-K receiving antenna 4-2-1 to 4-2-2K fast Fourier transformer 4-3 initial transmission coefficient estimation circuit 4-4, 8 transmission coefficient estimation circuit 4-5 Interference canceller 4-6-1 to 4-6-K demodulator 4-4-1-1 to 4-4-1-K, 8-1-1 to 8-1-K, 9-1- 1-9-1-K, 13-1-1 to 13-1-K, 14-1-1 to 14-1-K Remodulator 4-4-2, 8-2, 9-2, 13- 2, 14-2 Replica matrix generator 4-4-3, 8-3, 9-3, 13-3, 14-3 Pseudo inverse matrix calculator 4-4-4, 8-4, 9-4, 13 -4, 14-4 Received signal matrix generator 4-4-5, 8-5, 9-5, 13-5, 14-5 Matrix multiplier 4-4-6, 8-6, 9-6, 13 -6, 14-6 Transfer coefficient estimation matrix storage Circuit 4-4-7, 8-7, 9-7, 13-7, 14-7 Transfer coefficient estimation matrix update circuit 5, 6, 7, 10, 11, 12 Transfer coefficient estimation circuit 5-1, 6-1 7-1 Matrix weight calculation circuit 5-2, 6-2, 7-2 Weight coefficient storage circuit 5-3, 6-3, 7-3 Matrix weight synthesis circuit 6-4, 7-4 Matrix weight adder 6 -5, 6-5 Transfer coefficient estimation matrix update determination circuit 7-6 Transfer coefficient estimation matrix non-update period calculation circuit 7-7 Transfer coefficient matrix history information control circuit 8-8, 9-8, 14-8 Replica estimated transfer coefficient Matrix filter 8-8-1, 9-9-1, 13-8-1, 14-9-1 Matrix subtractor 8-8-2, 9-9-2 Matrix Euclidean distance calculation circuit 8-8-3 Replica Estimated transfer coefficient matrix selection circuit 9-9, 14-9 Transfer coefficient estimation matrix filter 9-9-3 Transfer coefficient Column selection circuit 10-1, 11-1, 12-1 Vector weight calculation circuit 10-2, 11-2, 12-2 Weight coefficient storage circuit 10-3, 11-3, 12-3 Vector weighting synthesis circuit 11- 4, 12-4 Vector weight adder 11-5, 12-5 Transfer coefficient estimation matrix sequence vector update determination circuit 12-6 Transfer coefficient estimation matrix sequence vector non-update period calculation circuit 12-7 Transfer coefficient matrix history information control circuit 13 -8 Replica estimated transfer coefficient matrix string vector filter 13-8-2, 14-9-2 Vector Euclidean distance calculation circuit 13-8-3 Replica estimated transfer coefficient matrix string vector selection circuit 14-9-3 Transfer coefficient matrix string vector Selection circuit

Claims (26)

When a spatially multiplexed OFDM transmission signal sequence transmitted from a plurality of transmission antennas on the transmission side is received by a plurality of reception antennas on the reception side, a fast Fourier transform is performed on the OFDM reception signal sequence, and the transmission side Transfer coefficient between antennas for outputting a signal point for each subcarrier from a spatially multiplexed received signal using a known pilot signal to be transmitted and a received signal after fast Fourier transform corresponding to the pilot signal A transfer coefficient estimation circuit for estimating
A remodulator that remodulates a transmission data sequence that is a demodulated output for a signal point for each subcarrier, and generates a transmission signal replica for each subcarrier;
A replica matrix generator that generates, for each subcarrier, a replica matrix whose elements are transmission signal replicas generated by the remodulator;
A pseudo inverse matrix calculator for obtaining an inverse matrix or a pseudo inverse matrix of the replica matrix for each subcarrier;
A reception signal matrix generator for generating a reception signal matrix having a reception signal output for each subcarrier after the fast Fourier transform as an element;
A matrix multiplier that multiplies the received signal matrix by an inverse matrix or a pseudo inverse matrix of a replica matrix that is an output of the pseudo inverse matrix calculator, and outputs a multiplication result as a replica estimated transfer coefficient matrix;
A transmission coefficient matrix for estimating an initial value of a transmission coefficient matrix using a known pilot signal transmitted from the transmission side and the received signal after the fast Fourier transform corresponding to the pilot signal, and outputting as an initial transmission coefficient estimation matrix An initial estimation circuit;
A transmission coefficient estimation matrix update circuit that weights and synthesizes the initial transmission coefficient estimation matrix and the replica estimation transmission coefficient matrix and outputs the current transmission coefficient estimation value ;
The transfer coefficient estimation matrix update circuit includes:
A matrix weight calculation circuit for calculating a weight coefficient for the replica estimated transfer coefficient matrix;
A matrix weighting synthesis circuit for weighting and synthesizing the initial transmission coefficient estimation matrix and the replica estimation transmission coefficient estimation matrix for each subcarrier, using the weighting coefficient calculated by the matrix weight calculation circuit;
Consists of
The matrix weight calculation circuit includes:
Weights inversely proportional to the sum of squares of the absolute values of the components of the pseudo-inverse matrix of the replica matrix used to derive the replica estimated transfer coefficient matrix for the weight coefficient for the replica estimated transfer coefficient matrix estimated in the past and present Coefficient
A transfer coefficient estimation circuit characterized by the above.   The transfer coefficient estimation circuit according to claim 1, wherein the pseudo inverse matrix calculator has a function of generating and outputting an inverse matrix of a replica matrix for each subcarrier. The transfer coefficient estimation matrix update circuit includes:
A matrix weight adder for adding the weight coefficients corresponding to a plurality of replica transfer coefficient estimation matrices estimated at different times for each subcarrier for each same subcarrier;
A transfer coefficient estimation matrix update determination circuit that outputs an update command for the subcarrier to the transfer coefficient estimation matrix update circuit when the addition result by the matrix weight adder exceeds a predetermined threshold value; ,
The matrix weighting synthesis circuit weights and synthesizes the past and current replica estimated transmission coefficient matrixes and the past transmission coefficient estimation matrices for the subcarriers for which an update command is notified from the transmission coefficient estimation matrix update determination circuit. 3. The transfer coefficient according to claim 1 , wherein a past transfer coefficient estimation matrix is output for a subcarrier that is output as a current transfer coefficient estimation matrix and for which an update command is not notified. Estimating circuit. The transfer coefficient estimation matrix update circuit includes:
Based on the determination result by the transfer coefficient estimation matrix update determination circuit, the difference between the time when the past transfer coefficient estimation matrix is updated for each subcarrier and the current time is calculated and output as a non-update period of the transfer coefficient estimation matrix A transfer coefficient estimation matrix non-update period calculation circuit to perform,
A transfer coefficient matrix history information control circuit that outputs a history deletion command when a predetermined threshold value is exceeded for a non-update period of the transfer coefficient estimation matrix;
For a subcarrier for which an erasure command from the transmission coefficient history information control circuit has been notified, the history of a part or the whole of the past replica estimated transmission coefficient matrix is erased, and the erased past replica estimated transmission coefficient is 4. The transfer coefficient estimation circuit according to claim 3, wherein a history of corresponding matrix weights is deleted. Based on the transfer coefficient estimation matrix and the replica estimated transfer coefficient matrix, it is determined whether update processing is valid or invalid for each component of the transfer coefficient estimation matrix of each subcarrier, and the determination result is determined to be valid. 5. The transfer coefficient estimation circuit according to claim 1, further comprising a replica estimated transfer coefficient matrix filter that outputs only the subcarrier as a replica estimated transfer coefficient matrix. The replica estimated transfer coefficient matrix filter is:
A matrix subtractor that calculates a difference between the previously estimated transfer coefficient estimation matrix and a replica estimated transfer coefficient matrix that is an output from the matrix multiplier for each subcarrier, and outputs the difference as a transfer coefficient difference matrix;
A matrix geometric distance calculation circuit that calculates the sum of all geometric distances of elements constituting the matrix for each transmission coefficient difference matrix for each subcarrier, and outputs the result as a transmission coefficient difference value;
When the transfer coefficient difference value is below a predetermined threshold value, the input replica estimation transfer coefficient matrix is output, and when waiting for the predetermined threshold value to be exceeded, the input replica estimation value is output. 6. The transfer coefficient estimation circuit according to claim 5, comprising a transfer coefficient estimation matrix selection circuit that does not output a transfer coefficient matrix. The geometric distance calculation circuit includes:
7. The transfer coefficient estimation circuit according to claim 6, wherein an absolute value of elements constituting the matrix or a square value of the absolute value is used as a geometric distance of each element of the transfer coefficient difference matrix. The geometric distance calculation circuit calculates, as a geometric distance of each element of the transfer coefficient difference matrix, a sum of absolute values of real parts and imaginary parts of elements constituting the matrix or a square value of the values. 7. The transfer coefficient estimating circuit according to claim 6 , wherein the transfer coefficient estimating circuit is used. Based on the transfer coefficient estimation matrix estimated in the past and the replica estimated transfer coefficient matrix, a determination is made for output selection for each subcarrier, and only subcarriers determined to be valid based on the determination result are replicas. 9. The transfer coefficient estimation circuit according to claim 5, further comprising a transfer coefficient estimation matrix filter that outputs an estimated transfer coefficient matrix. The transfer coefficient estimation matrix filter is:
A matrix subtractor that calculates a difference between the transfer coefficient estimation matrix estimated at the current time and the transfer coefficient estimation matrix estimated in the past for each subcarrier, and outputs the difference as a transfer coefficient difference matrix;
A matrix geometric distance calculation circuit that calculates the sum of geometric distances for all of the elements constituting the matrix for each transmission coefficient difference matrix for each subcarrier, and outputs the result as a transmission coefficient difference value;
When the transfer coefficient difference value is below a predetermined threshold value, the transfer coefficient estimation matrix estimated at the current time is output to the interference canceller, and when waiting for the predetermined threshold value to be exceeded The transmission coefficient estimation circuit according to claim 9 , further comprising: a transmission coefficient estimation matrix selection circuit that outputs the transmission coefficient estimation matrix estimated in the past to the interference canceller. The matrix geometric distance calculation circuit includes:
11. The transfer coefficient estimation circuit according to claim 10, wherein an absolute value of elements constituting the matrix or a square value of the absolute value is used as a geometric distance of each element of the transfer coefficient difference matrix. The matrix geometric distance calculation circuit, as the geometric distance of each element of the transfer coefficient difference matrix, the sum of the absolute value of the real part and the absolute value of the imaginary part of the elements constituting the matrix or the square value of the value The transfer coefficient estimation circuit according to claim 10, wherein: When a spatially multiplexed OFDM transmission signal sequence transmitted from a plurality of transmission antennas on the transmission side is received by a plurality of reception antennas on the reception side, a fast Fourier transform is performed on the OFDM reception signal sequence, and the transmission side Transfer coefficient between antennas for outputting a signal point for each subcarrier from a spatially multiplexed received signal using a known pilot signal to be transmitted and a received signal after fast Fourier transform corresponding to the pilot signal A transfer coefficient estimation circuit for estimating
A remodulator that remodulates a transmission data sequence that is a demodulated output for a signal point for each subcarrier, and generates a transmission signal replica for each subcarrier;
A replica matrix generator that generates, for each subcarrier, a replica matrix whose elements are transmission signal replicas generated by the remodulator;
A pseudo inverse matrix calculator for obtaining an inverse matrix or a pseudo inverse matrix of the replica matrix for each subcarrier;
A reception signal matrix generator for generating a reception signal matrix having a reception signal output for each subcarrier after the fast Fourier transform as an element;
A matrix multiplier that multiplies the received signal matrix by an inverse matrix or a pseudo inverse matrix of a replica matrix that is an output of the pseudo inverse matrix calculator, and outputs a multiplication result as a replica estimated transfer coefficient matrix;
A transmission coefficient matrix for estimating an initial value of a transmission coefficient matrix using a known pilot signal transmitted from the transmission side and the received signal after the fast Fourier transform corresponding to the pilot signal, and outputting as an initial transmission coefficient estimation matrix An initial estimation circuit;
A transmission coefficient estimation matrix update circuit that weights and synthesizes the initial transmission coefficient estimation matrix and the replica estimation transmission coefficient matrix and outputs the current transmission coefficient estimation value ;
The transfer coefficient estimation matrix update circuit includes:
For each column vector of the replica estimated transfer coefficient matrix for each subcarrier at each time using the output of the pseudo inverse matrix calculator, the currently estimated replica estimated transfer coefficient matrix, and the previously estimated transfer coefficient estimated matrix A vector weight calculation circuit for calculating a weight coefficient;
A matrix weighting synthesis circuit for weighting and synthesizing the replica estimated transmission coefficient matrix and the past transmission coefficient estimation matrix for each column vector using the weighting coefficient;
A transfer coefficient estimation circuit comprising: The vector weight calculation circuit includes:
The weight coefficient for the column vector of the replica estimated transfer coefficient matrix estimated in the past and the present is used for each component of the row matrix of the inverse matrix of the replica matrix or the pseudo inverse matrix used in deriving each replica estimated transfer coefficient matrix. 14. The transfer coefficient estimation circuit according to claim 13, wherein the weight coefficient is inversely proportional to the sum of squares of absolute values. The transfer coefficient estimated value update circuit includes:
A vector weight adder for adding weight coefficients corresponding to a plurality of replica transfer coefficient estimation matrices estimated at different times for each subcarrier for each same column vector;
If the addition result of the matrix weight adder exceeds a predetermined threshold value, a transfer coefficient estimation matrix column vector update determination circuit that outputs an update command for the column vector;
For the column vector for which the update instruction is notified from the transfer coefficient estimation matrix update determination circuit, the column vectors of the past and the current replica estimated transfer coefficient matrix and the past transfer coefficient estimation matrix are weighted and synthesized. A transfer coefficient estimation matrix column vector update circuit that outputs the replica estimation transfer coefficient matrix or the past transfer coefficient estimation matrix as it is for a column vector that is output as a column vector of the transfer coefficient estimation matrix The transfer coefficient estimation circuit according to claim 14, further comprising: The transfer coefficient estimation matrix update circuit includes:
Based on the determination result by the transfer coefficient estimation matrix update determination circuit, the difference between the time when the past transfer coefficient estimation matrix is updated for each column vector and the current time is calculated and output as a transfer coefficient estimation matrix column vector non-update period Transfer coefficient estimation matrix column vector non-update period calculation circuit,
A transmission coefficient matrix history information control circuit that outputs a history erasure command when a predetermined threshold is exceeded for the transmission coefficient estimation matrix column vector non-update period, the erasure command by the transmission coefficient history information control circuit is For the notified column vector, a part or the entire history of the column estimated past transfer coefficient matrix of the past replica estimated transfer coefficient matrix is deleted, and a matrix corresponding to the column vector of the deleted past replica estimated transfer coefficient 16. The transfer coefficient estimation circuit according to claim 15, wherein the weight history is deleted. Based on the transfer coefficient estimation matrix and the replica estimated transfer coefficient matrix, it is determined whether update processing is valid or invalid for each component of the column vector of the transfer coefficient estimation matrix of each subcarrier, and the determination result is valid. 17. The transfer coefficient estimation circuit according to claim 13, further comprising: a replica estimated transfer coefficient matrix column vector filter that outputs only the column vector determined to be the column vector of the replica estimated transfer coefficient estimation matrix. . The replica estimated transfer coefficient matrix column vector filter is:
A vector subtractor that calculates a difference between the transfer coefficient estimation matrix estimated in the past and the replica estimated transfer coefficient matrix for each column vector, and outputs the difference as a transfer coefficient difference vector;
For each transfer coefficient difference vector for each subcarrier, a vector geometric distance calculation circuit that calculates the sum of the squares of all the geometric distances of the elements constituting the vector and outputs as a vector difference value;
When the vector difference value is below a predetermined threshold value, the column vector of the replica estimated transfer coefficient matrix is output, and when waiting for the predetermined threshold value to be exceeded, the replica estimated transfer value is output. 18. The transfer coefficient estimation circuit according to claim 17, further comprising: a transfer coefficient matrix column vector estimation accuracy determination circuit that does not output the column vector of the coefficient matrix. The vector geometric distance calculation circuit includes:
19. The transfer coefficient estimation circuit according to claim 18 , wherein an absolute value of elements constituting the vector or a square value of the absolute value is used as a geometric distance of each element of the transfer coefficient difference vector. The vector geometric distance calculation circuit calculates the geometric distance of each element of the transfer coefficient difference vector as the sum of the absolute value of the real part and the absolute value of the imaginary part of the element constituting the vector or the square of the value. 19. The transfer coefficient estimation circuit according to claim 18 , wherein a value is used. Based on the transfer coefficient estimation matrix estimated in the past and the replica estimated transfer coefficient matrix, a determination for selecting an output is performed for each column vector, and only column vectors determined to be valid based on the determination result are stored in the past. Output the column vector of the replica estimated transfer coefficient matrix to the column vector that has not been determined to be valid and outputs the column vector of the transfer coefficient estimation matrix estimated in the past. 21. The transfer coefficient estimation circuit according to claim 1 , wherein the transfer coefficient estimation circuit is provided. The transfer coefficient estimation matrix column vector filter is:
A matrix subtractor that calculates a difference between the transfer coefficient estimation matrix estimated at the current time and the transfer coefficient estimation matrix estimated in the past for each column vector, and outputs the difference vector;
A vector geometric distance calculation circuit for calculating a sum of squares of all geometric distances of elements constituting the column vector for each difference vector for each subcarrier;
If the output of the vector geometric distance calculation circuit is below a predetermined threshold, the column vector of the transfer coefficient estimation matrix estimated at the current time is output to the interference canceller, and the predetermined threshold is set. If exceeds the value of claim 21, characterized in that it is configured the column vector of the estimated transfer coefficient estimation matrix in the past and a transmission coefficient estimation matrix column vector selection circuit for outputting the interference canceller Transfer coefficient estimation circuit. The vector geometric distance calculation circuit includes:
23. The transfer coefficient estimation circuit according to claim 22 , wherein an absolute value of elements constituting the vector or a square value of the absolute value is used as a geometric distance of each element of the transfer coefficient difference vector. The vector geometric distance calculation circuit includes:
Claims, characterized in that the use of the as the geometric distance of the respective elements of the transfer coefficient difference vectors, the square value of the sum or said value of the absolute value of the absolute value and the imaginary part of the real part of the elements constituting the vector 22. A transfer coefficient estimation circuit according to 22 . When a spatially multiplexed OFDM transmission signal sequence transmitted from a plurality of transmission antennas on the transmission side is received by a plurality of reception antennas on the reception side, a fast Fourier transform is performed on the OFDM reception signal sequence, and the transmission side Transfer coefficient between antennas for outputting a signal point for each subcarrier from a spatially multiplexed received signal using a known pilot signal to be transmitted and a received signal after fast Fourier transform corresponding to the pilot signal A transfer coefficient estimation method used in a transfer coefficient estimation circuit for estimating
Remodulator, the transmission data sequence is demodulated output for the signal points of each of the sub-carrier re-modulated, perform the remodulation procedure that generates a transmission signal replica for each subcarrier,
The replica matrix generator executes a replica matrix generation procedure for generating, for each subcarrier, a replica matrix having the transmission signal replica generated by the remodulator as an element ,
The pseudo inverse matrix calculator performs a pseudo inverse matrix calculation procedure for obtaining an inverse matrix or a pseudo inverse matrix of a replica matrix for each subcarrier ,
The reception signal matrix generator executes a reception signal matrix generation procedure for generating, for each subcarrier, a reception signal matrix whose element is a reception signal output for each subcarrier after the Fast Fourier Transform,
Matrix multiplier, the matrix multiplication steps where the inverse matrix or the pseudo inverse matrix of replica matrices the which is the output of the pseudo-inverse matrix calculator multiplying the received signal matrix, and outputs the multiplication result as a replica estimated transfer coefficient matrix Run ,
The transfer coefficient matrix initial estimation circuit estimates an initial value of the transfer coefficient matrix using a known pilot signal transmitted from the transmission side and the received signal after the fast Fourier transform corresponding to the pilot signal, and an initial transfer coefficient Perform transfer coefficient matrix initial estimation procedure to output as estimation matrix ,
The transfer coefficient estimation matrix update circuit weights and synthesizes the initial transfer coefficient estimation matrix and the replica estimated transfer coefficient matrix, and executes a transfer coefficient estimation matrix update procedure for outputting as a current transfer coefficient estimation value,
In the transfer coefficient estimation matrix update procedure, the transfer coefficient estimation matrix update circuit includes:
A matrix weight calculation procedure for calculating a weight coefficient for the replica estimated transfer coefficient matrix ;
A matrix weighting synthesis procedure for weighting and synthesizing the initial transmission coefficient estimation matrix and the replica estimated transmission coefficient estimation matrix for each subcarrier using the weighting coefficient calculated by the matrix weight calculation circuit ;
Run
In the matrix weight calculation procedure ,
Weights inversely proportional to the sum of squares of the absolute values of the components of the pseudo-inverse matrix of the replica matrix used to derive the replica estimated transfer coefficient matrix for the weight coefficient for the replica estimated transfer coefficient matrix estimated in the past and present Coefficient
A transfer coefficient estimation method characterized by the above. When a spatially multiplexed OFDM transmission signal sequence transmitted from a plurality of transmission antennas on the transmission side is received by a plurality of reception antennas on the reception side, a fast Fourier transform is performed on the OFDM reception signal sequence, and the transmission side Transfer coefficient between antennas for outputting a signal point for each subcarrier from a spatially multiplexed received signal using a known pilot signal to be transmitted and a received signal after fast Fourier transform corresponding to the pilot signal A transfer coefficient estimation method used in a transfer coefficient estimation circuit for estimating
The remodulator remodulates a transmission data sequence which is a demodulated output for the signal point for each subcarrier, and executes a remodulation procedure for generating a transmission signal replica for each subcarrier,
The replica matrix generator executes a replica matrix generation procedure for generating, for each subcarrier, a replica matrix having the transmission signal replica generated by the remodulator as an element,
The pseudo inverse matrix calculator performs a pseudo inverse matrix calculation procedure for obtaining an inverse matrix or a pseudo inverse matrix of a replica matrix for each subcarrier ,
The reception signal matrix generator executes a reception signal matrix generation procedure for generating, for each subcarrier, a reception signal matrix whose element is a reception signal output for each subcarrier after the Fast Fourier Transform ,
Matrix multiplier, the matrix multiplication steps where the inverse matrix or the pseudo inverse matrix of replica matrices the which is the output of the pseudo-inverse matrix calculator multiplying the received signal matrix, and outputs the multiplication result as a replica estimated transfer coefficient matrix Run ,
The transfer coefficient matrix initial estimation circuit estimates an initial value of the transfer coefficient matrix using a known pilot signal transmitted from the transmission side and the received signal after the fast Fourier transform corresponding to the pilot signal, and an initial transfer coefficient Perform transfer coefficient matrix initial estimation procedure to output as estimation matrix ,
The transfer coefficient estimation matrix update circuit weights and synthesizes the initial transfer coefficient estimation matrix and the replica estimated transfer coefficient matrix, and executes a transfer coefficient estimation matrix update procedure for outputting as a current transfer coefficient estimation value ,
In the transfer coefficient estimation matrix update procedure, the transfer coefficient estimation matrix update circuit includes:
For each column vector of the replica estimated transfer coefficient matrix for each subcarrier at each time using the output of the pseudo inverse matrix calculator, the currently estimated replica estimated transfer coefficient matrix, and the previously estimated transfer coefficient estimated matrix A vector weight calculation procedure for calculating a weight coefficient ;
A matrix weighting synthesis procedure for weighting and synthesizing the replica estimated transfer coefficient matrix and the past transfer coefficient estimation matrix for each column vector using the weighting coefficient ;
The transfer coefficient estimation method characterized by performing.

Images