JP4451645B2 - Noise removing apparatus, method and program - Google Patents

Description

The present invention relates to a signal waveform generation technique and a noise removal technique, and more particularly, to a noise removal apparatus, method, and program capable of storing a noise signal waveform and the like used for steady digital noise removal with a small amount of memory.

  In a signal processing apparatus that processes an audio signal, a radio wave signal received by an antenna, etc., noise from a digital system is often mixed in the process of digitizing an analog signal.

  FIG. 15 is a diagram illustrating a system configuration example of a signal processing apparatus according to the related art. In FIG. 15, 2 is an antenna that receives a signal and transmits time-series data as an analog signal to the AD converter, 3 is an AD converter that converts the received analog signal into a digital signal, and 4 is received from the AD converter 3. This is a Fourier transform device that performs Fourier transform on the digital signal and extracts its frequency component.

  When the received analog signal is converted into a digital signal using the system configuration shown in FIG. 15, stationary noise such as a clock waveform of the clock system may leak. For example, if noise similar to the clock waveform shown in FIG. 4B leaks, the noise is mixed with the original clean cosine wave (Cos (2πfk / N)) as shown in FIG. It becomes such a dirty wave.

  When a signal containing a lot of noise is subjected to Fourier transform processing, the accuracy of the original signal is deteriorated due to computation overflow caused by noise. Therefore, in order to fully exploit the capability of the signal processing apparatus, it is necessary to perform signal processing after removing known noise in advance.

  As a first conventional technique, there is a technique that removes noise by holding a noise signal waveform in a memory by the number N of sample points required for signal processing and subtracting the noise signal waveform from an input signal. Further, as a second conventional technique, when outputting a voice or a musical sound, there is a technique in which a signal waveform stored in a waveform memory is read and then an operation is added as necessary.

As a more specific conventional technique, the following Patent Document 1 describes a technique for repeatedly reading out the same sample point from a basic signal waveform for one period and generating a low-frequency signal. The technique described in Patent Document 1 includes a waveform data memory that stores a basic signal waveform for one period as a series of signal values, and controls the signal values in order to sequentially read out the signal values at a predetermined period from the waveform data memory. In accordance with the thinning number set according to the frequency of the test waveform to be given as the target value, the signal value of the same sample point is continuously read repeatedly from the waveform data memory, and the read signal value is corrected according to the thinning number. The basic signal waveform indicated by the series of signal values is restored, and the amplitude of the restored basic signal waveform is adjusted to obtain a test waveform that forms a control target value.
Japanese Patent Laid-Open No. 11-64195

  As described above, when a signal containing a lot of noise is subjected to Fourier transform processing, the accuracy of the original signal is deteriorated due to computation overflow caused by the noise. The reason is as follows. In general, signal processing devices often use fixed-point numbers that can be configured with a small amount of hardware. In the case of arithmetic processing using floating-point numbers, there is little occurrence of arithmetic overflow, but in the case of fixed-point numbers, arithmetic overflow is likely to occur.

  Considering an unsigned 8-bit arithmetic circuit, the integer part can take a value from 0 to 255. In order to use the number of bits effectively, the data is used in the full range as much as possible. For example, if the original signal is a value of 250, the 8-bit accuracy can be fully utilized. If noise having a value of 50 is superimposed on the original signal, the value of 300 must be processed, but cannot be processed in 8 bits. In the case of overflow, there is a method of clipping to the maximum value, but this is a method of forcibly setting the value of 300 to 255 and setting it to 8 bits, so in the process of performing arithmetic processing with other values, etc. Computational noise due to errors occurs.

  For example, when the scaling processing method is used, the data is scaled to ½ and processed. As a result, 300/2 = 150, and processing can be performed with 8 bits. However, since the original signal is also processed with 250/2 = 125, only 7-bit accuracy can be used.

  When the signal waveform of noise mixed in the original signal is known, the signal waveform of the known noise is removed from the input signal and then signal processing is performed to reduce degradation of the accuracy of the original signal. Thus, it is possible to fully exploit the capabilities of the signal processing device. For the purpose of removing such noise, it is not necessary to make the signal waveform of the noise to be synthesized accurate, as long as a certain large noise component can be removed. At this time, it is necessary to store the basic signal waveform of the noise in the memory in advance. Conventionally, the basic signal waveform of the noise is stored with a small amount of memory, and a desired synthesized signal of the noise is generated based on the basic signal waveform. A technique for generating a waveform has not been considered.

  The present invention solves the above-mentioned problems of the prior art and generates a signal waveform having a necessary frequency and the number of sample points from basic signal waveform data consisting of a small number of sample points, thereby achieving the required accuracy with a small amount of memory. The purpose of this invention is to make it possible to synthesize signals, and to further eliminate noise with a small amount of memory by subtracting the synthesized noise signal waveform from the input signal using the signal synthesis technique.

In order to solve the above problems, the present invention uses the following means. First, a technique for generating a composite signal waveform used in the present invention (hereinafter referred to as a related technique) will be described. A basic signal waveform for one period is prepared in a memory as a sample value of the number of sample points S. Based on the sample value of sample point S prepared in the memory, generate a harmonic of F times the required basic signal waveform for one cycle as a composite signal waveform of sample value N (S <N). To do. Thus, by adopting a configuration for generating a composite signal waveform having N sample points from basic signal waveform data for one cycle of sample points S prepared in the memory, the amount of memory necessary for generating a composite signal waveform is obtained. Reduce.

That is, the related technique generates a signal waveform that generates a harmonic of F times the basic signal waveform from a basic signal waveform having sample values of sample points S as a composite signal waveform of sample values of sample points N (S <N). A generator for storing sample values of the basic signal waveform in time series, and an integer of (S * F / N) * k (where k = 0, 1, 2,..., N−1) A calculation means for calculating a remainder when the part is divided by S, a sample value of the basic signal waveform is extracted from a position corresponding to the calculated remainder in the memory, and the extracted sample value is extracted from the synthesized signal waveform And a sample value determining means for setting the k-th sample value.

The present invention inputs a signal mixed with noise, performs Fourier transform, selects the Fourier component of the noise frequency from the result of Fourier transform, and converts the signal waveform resulting from inverse Fourier transform of the Fourier component of the selected noise frequency to the noise component. Is used as a basic signal component to generate a synthesized noise signal waveform by the related technique , and the noise is removed by subtracting the synthesized noise signal waveform from the input sample signal.

Specifically, the present invention provides a receiving unit that receives an input signal, a converting unit that performs a Fourier transform on the input signal, and a Fourier component of the frequency of noise included in the input signal based on the result of the Fourier transform. A selection unit for selecting, a generation unit for performing an inverse Fourier transform on the selected Fourier component to generate a basic signal waveform of the noise, and S sample points extracted from the generated basic signal waveform (S is a positive value) (Integer) sample values are stored in chronological order, and an integer part of (S * F / N) * k (where k = 0, 1, 2,..., N−1) is represented by the sample point S. An arithmetic unit that calculates a remainder when dividing, and a sample value of the basic signal waveform of the noise is extracted from a position in the storage unit corresponding to the calculated remainder, and the extracted sample value is extracted from a synthesized noise signal waveform determine that the k-th sample value, said to have a composition unit for generating a composite noise signal waveform is F-fold harmonic of the fundamental signal waveform, and a subtraction unit for subtracting the synthesized noise signal waveform from the input signal It is characterized by.

FIG. 1 is a diagram illustrating the principle of a technique for generating a composite signal waveform . For example, from the basic signal waveform data for one cycle with the number of sample points S = 32 as shown in FIG. 1A, harmonics that are F times (F = 3) the basic signal waveform for one cycle are converted into N points ( N = 64) is generated as a composite signal waveform of sample values. For this reason, basic signal waveform data for one cycle as shown in FIG. Here, if the basic signal waveform shown in FIG. 1A is simply repeated three times, the basic signal waveform has 96 sample points because it has 32 sample points.

  Since the composite signal waveform to be generated has N = 64 sample points, in order to generate this composite signal waveform, data every 1.5 sample points of a wave that is three times the basic signal waveform is generated. Just plot it. Since there is no data for every 1.5 sample points in the memory, for example, data is plotted every 1 sample point or every 2 sample points. As a result, a composite signal waveform having a waveform three times the basic signal waveform as shown in FIG. 1B can be generated.

The outline of the process of generating synthesis signal waveform will be described as an example the case of synthesizing a steady noise signal waveform. FIG. 2 is a diagram showing noise leakage into the original cosine wave. When noise leaks into the original cosine wave, the signal is mixed with noise as shown by the solid line in FIG.

  When the signal mixed with this noise is subjected to Fourier transform, the spectrum shown in FIG. 3A is detected. As shown in FIG. 3, there is a spectrum with a large amplitude (V) corresponding to a frequency of 4 Hz, and it can be seen that 4 Hz is the original cosine wave frequency. Then, it can be seen that 0 Hz, 3 Hz, 9 Hz, 15 Hz, 21 Hz, and 27 Hz, which are other frequencies from which the spectrum is detected, are frequencies corresponding to noise.

  The present invention, for example, generates a synthesized signal waveform of noise from a spectrum corresponding to noise after performing Fourier transform on a signal mixed with noise and separating it into a spectrum corresponding to the original cosine wave and a spectrum corresponding to noise. . In this case, a basic signal waveform of noise corresponding to one cycle with 32 sample points shown in FIG. 4A is generated and stored in the memory. Here, assuming that the target composite signal waveform is generated from the sample value of three times higher harmonics of the basic signal waveform and sample points N = 64, the basic signal waveform of noise stored in the memory is obtained. The wave of 96 sample points obtained by repeating 3 times is compressed, and the composite signal waveform of 64 sample points shown in FIG. 4B is generated. As described above, the compression to the composite signal waveform is performed by plotting 96 data of the 96-sampled waves as sample values of the composite signal waveform every other one or two.

The determination of the sample value of the noise composite signal waveform is performed as follows. The number of sample points of the composite signal waveform is k, the coefficient (harmonic multiple) indicating how many times the composite signal waveform is the basic signal waveform is F, the number of sample points of the basic signal waveform is S, and the number of sample points of the composite signal waveform is N Then, the data (sample value) of the sample point of the basic signal waveform corresponding to the remainder when the integer part of ( S * F / N) * k is divided by S is the k-th sample value of the composite signal waveform. To do.

  For example, when S = 32, N = 64, and F = 3, the integer part of (S * F / N) * k = (32 * 3/64) * k = 1.5 * k is divided by 32. Let the remainder be the sampling point of the basic signal waveform. In this example, the rounded residue is 0, 2, 3, 5, 6, 8, 9,. This is taken as the k-th sample value of the composite signal waveform while extracting the sample values at the sample positions with an average of 1.5 jumps from the memory by 1 or 2 jumps.

As described above, in the generation of the synthesized signal waveform, for the purpose of signal synthesis, even if the sample point position referenced in the basic signal waveform for one cycle has a value after the decimal point in the calculation, it is corrected to the sample position. Take a sample value. Further, in the calculation, a sample point position to be referred to that exceeds the basic signal waveform for one cycle is converted into a position in the basic signal waveform for one cycle, and a sample value is extracted. In general, this technique cannot generate an accurate synthesized waveform, but can easily generate a pseudo synthesized waveform.

  With this address calculation method, the relationship between the number of sample points S of the basic signal waveform and the number of sample points N of the synthesized signal waveform does not need to be a specific relationship, and can be arbitrarily set, and one cycle of the basic signal An arbitrary harmonic multiple F of the waveform can be set.

In addition, a multiplier is required for the generation of the composite signal waveform. However, when the number N of sample points of the composite signal waveform is large, the basic signal waveform memory for one cycle can be reduced as compared with the conventional case. Even if a device is used, the amount of hardware can be reduced sufficiently. That is, example embodiment, to prepare 32 samples of the noise of the fundamental signal waveform memory, because the number of sample points can be easily generate a composite signal waveform 64 of the noise, the amount of data stored in the memory Can be reduced.

In the present invention, for example, after an analog signal received by an antenna or the like is digitally converted, the signal is decomposed into frequency components by a Fourier transform device, and only a component determined to be noise based on the spectrum distribution is subjected to inverse Fourier transform. Generate a noise waveform. At this time, only one period of the noise component to be removed is set in the memory as a basic signal waveform.

When a square wave such as a clock waveform leaks into the input signal and becomes noise, it is often a noise synthesized from the fundamental frequency of the clock and its odd harmonics. That is, noise can be synthesized from a waveform having a time corresponding to only one period of the fundamental frequency of the clock. Therefore, the present invention uses the above-described technique for generating a synthesized signal waveform to sample a harmonic of F times (for example, F = 3) of the basic signal waveform of noise for one period by sampling points N (for example, N = 64). A value is generated as a synthesized signal waveform, and noise is removed by subtracting the generated synthesized signal waveform from the input sampling signal.

  For example, the basic signal waveform of noise shown in FIG. 4 (A) is prepared in a memory, and the synthesized signal waveform shown in FIG. 4 (B) is generated based on the basic signal waveform. For example, noise is removed by subtracting from a wave (a signal mixed with noise) indicated by a dotted line in FIG. As a result of noise removal, for example, a signal after noise removal as shown by a thick solid line in FIG. 5 is obtained.

  In general, leakage of noise from the clock system or power supply system to the analog signal processing system is a difficult problem to prevent. According to the present invention, in order to use stationary known noise for noise removal from an input signal, it is possible to synthesize a noise waveform with a small amount of hardware and remove the noise while having the required signal waveform accuracy. This makes it possible to extract the performance of the original Fourier transform device.

  In addition, according to the present invention, since a general noise waveform can be synthesized, a necessary signal waveform can be synthesized with a small amount of memory. In particular, a basic waveform for one cycle can be expressed by an arbitrary number of sample points S, and a harmonic multiple F for an arbitrary basic waveform having an arbitrary number of sample points N can be set.

  In the following description, Fix {A} means taking out the integer part of the fixed-point number {A}. Mod [B, S] means that a remainder obtained by dividing B by S is taken out. “*” Indicates multiplication, and “/” indicates division.

Figure 6 is a diagram showing a configuration example of the signal waveform generation unit. In FIG. 6, 10 is a signal waveform generator for generating a composite signal waveform, 11 is a basic signal waveform initial setting unit for initially setting a basic signal waveform for one cycle in a memory 12, and 12 is a memory for setting a basic signal waveform. , 13 is a counter for counting the sample point number k, 14 is a multiplier for calculating k * F (F is a multiple of harmonics), and 15 is a bit shift of k * F (S * F / N) *. k is a shift circuit that outputs k, and 16 selects the remainder of the result of dividing Fix {(S * F / N) * k} by S from the output of the shift circuit 15, and the sample value Q (k ) For selecting a read address.

Figure 7 is a diagram showing an example of a signal waveform generating process flow. It is assumed that a basic signal waveform for one cycle is set in advance in the memory 12 by the basic signal waveform initial setting unit 11. First, the sample point number k is set to k = 0 (step S1), and the following steps S3 to S5 are repeated until k = N (step S2). When k = N, the process is terminated.

  Mod [Fix {(S * F / N) * k}, S] is calculated and set to A (k) (step S3). A (k) indicates a value corresponding to the address of the memory. That is, the remainder when the integer part of (S * F / N) * k is divided by S is A (k). Here, when rounding off when extracting the integer part, a binary number “1” is added to the first decimal place and then the decimal part is discarded.

  8A shows the amplitude value (sample value) for each sample point i of the basic signal waveform shown in FIG. 1A, and FIG. 8B shows the synthesis shown in FIG. Data such as sample values corresponding to each sample point k of the signal waveform is shown. As described above, the composite signal waveform shown in FIG. 1B is three times the basic signal waveform shown in FIG. Further, the number of sample points S of the basic signal waveform is 32, and the number of sample points N of the composite signal waveform is 64. In FIG. 8B, Fix () represents Fix {(S * F / N) * k}, and Mod () represents Mod [Fix {(S * F / N) * k}, S]. .

For example, at the sample point of k = 1 in the composite signal waveform,
(S * F / N) * k = (32 * 3/64) = 1.5
Because
Fix {(S * F / N) * k} = Fix {1.5} = 2
Mod [Fix {(S * F / N) * k}, S] = Mod [2, 32] = 2
It becomes. Therefore, A (k) = 2 is set in step S3.

  The A (k) th sample value M [A (k)] of the basic signal waveform in the memory is set as the kth sample value Q (k) of the synthesized signal waveform (step S4). For example, since A (k) = 2 is set in step S3, 0.9375, which is the sample value of the basic signal waveform in the case of i = 2 in FIG. Output as Q (k). Thereafter, k is incremented (step S5), and the same processing is repeated until k = N.

The operation of the signal waveform generator shown in FIG. 6 will be described in detail. The counter 13 outputs values k = 0, 1, 2,... The number of sample points S of the basic signal waveform for one cycle is a power of 2 (S = 2 ^p ). The number N of synthesized signal waveform sample points is assumed to be a power of 2 (N = 2 ^q ). S / N = 2 ^p / 2 ^q = 2 ^(pq) .

  The multiplier 14 multiplies the given F by k and outputs k * F. The shift circuit 15 shifts k * F by (p−q) bits and outputs a value of (S * F / N) * k. The selection unit 16 outputs the remainder when Fix {(S * F / N) * k} is divided by S by extracting the lower p bits of the integer part. Fix {(S * F / N) * k} can be obtained by discarding the fractional part if it is a rounding-down process, and “1” as a binary number at the first decimal place if rounding-off process. In addition, it can be obtained by discarding the decimal places. The remainder Mod [Fix {(S * F / N) * k}, S] obtained by the selection unit 16 is set as the read position A (k) of the memory 12, and A (k) of the basic signal waveform in the memory 12 The th sample value is output as the k th sample value Q (k) of the combined signal waveform.

  Hereinafter, an example of numerical calculation in the case of k = 5 in the signal waveform generation apparatus 10 will be shown.

S = 32 = 2 ^p = 2 ⁵
N = 64 = ^2q = 2 ⁶
k = 5 = 000101.
F = 3 = 0011.
k * F = 15 = 00001111.
(S * F / N) * k = (k * F) * (1-bit right shift)
= 7.5 = 00000111.1
Rounded off (S * F / N) * k = 8.0 = 0001000.0
A (5) = Mod [8, 32] = 8 = 1000.
Q (5) = M [A (5)] = M [8]
From this result, k = 5th sample value Q (5) of the composite signal waveform is read from the 8th entry M [8] in the memory 12 and output.

Figure 9 is a diagram showing an outline of implementation in the form of a system configuration of the present invention. In FIG. 9, 1 is a noise removing device, 2 is an antenna, 3 is an AD converting device for converting an analog signal into a digital signal, and 4 is a Fourier transforming device. In this onset Ming embodiment, the antenna 2 receives a signal, when output series data as an analog signal. The AD converter 3 converts the analog signal received from the antenna 2 into a digital signal. The noise removal device 1 removes noise from the digital signal received from the AD conversion device 3. The Fourier transform device 4 performs a Fourier transform on the signal wave output from the noise removal device 1 and outputs a frequency component.

Figure 10 is a diagram showing details of a system configuration example of the implementation of the embodiment of the present invention. Reference numeral 17 in the noise removing apparatus 1 shown in FIG. 10 is a subtracting circuit that subtracts the synthesized signal waveform of noise generated by the signal waveform generating apparatus 10 from the digital signal received from the AD converting apparatus 3. The signal waveform generation apparatus 10 is the same as that described with reference to FIG. The basic signal waveform initial setting unit 11 stores in advance the sample value of the noise basic signal waveform for one period in the memory 12.

Figure 11 is a diagram showing an initial setting process flow of a basic signal waveform according to the implementation of the embodiment of the present invention. When setting the sample value of the basic signal waveform of noise in the memory 12, the subtraction circuit 17 is stopped, and the signal input to the subtraction circuit 17 is transmitted to the Fourier transform apparatus 4 as it is without performing subtraction processing.

  First, a signal is received by the antenna 2 (step S11). The received signal is AD converted by the AD converter 3 and transmitted to the Fourier transformer 4 through the subtraction circuit 17 (step S12). The signal input to the Fourier transform device 4 is a signal in which a steady digital noise due to leakage of a clock signal or the like is mixed into the original input signal. The Fourier transform device 4 performs a Fourier transform on this input signal (step S13).

  Next, the basic signal waveform initial setting unit 11 selects a specific noise frequency component from the Fourier components obtained as a result of the Fourier transform (step S14). The selection of the noise frequency component may be performed by displaying the frequency spectrum resulting from the Fourier transform on the display and allowing the user to specify the frequency spectrum using a pointing device or the like. For example, the power of the original input signal is the noise power. If it is apparent that the noise frequency component is larger than the noise frequency component, the noise frequency component may be automatically determined by comparing the power of each frequency.

  Next, the Fourier component of the selected noise frequency is inverse Fourier transformed (step S15). The signal waveform obtained by the inverse Fourier transform is basically a signal waveform of only stationary noise. Therefore, a basic signal waveform for one cycle is cut out from this signal waveform, a sample value of S sample points corresponding to the capacity of the memory 12 is calculated, and the sample value of the basic signal waveform is set in the memory 12 (step) S16).

Figure 12 is a diagram illustrating a noise removal processing flow of implementation of the present invention. First, a signal is received by the antenna 2 (step S21). Next, the received signal is AD converted by the AD converter 3 and transmitted to the subtraction circuit 17 (step S22). The subtracting circuit 17 subtracts the synthesized signal waveform of noise generated by the signal waveform generating device 10 from the received signal to remove noise from the received signal (step S23). The signal after the noise removal is output (step S24), and the process ends. The composite signal waveform generation processing in step S23 is executed according to the same processing as the composite signal waveform generation processing flow shown in FIG.

  Consider an example of leakage noise from the system clock in the AD converter 3. Assume that a system clock of 64 MHz is input to generate a 1024 MHz sampling clock. It is assumed that the data sampled at a sampling rate of 1024 Msamples / second is subjected to a fast Fourier transform of 16384 points (= 16 * 1024 points). The period for Fourier transform is 16384 / 1024M = 16.0 μsec as shown in FIG. One period of the clock waveform of 64 MHz is 1/64 MHz = 0.015625 μsec as shown in FIG. This is 0.0156625 / 16.0 = 1/1024 considering the period of Fourier transform. Since 1/1024 of the number of sample points 16384 to be subjected to the fast Fourier transform is 16, the number of sample points of the basic signal waveform in one cycle is 16. Therefore, a basic signal waveform of one cycle is prepared with 16 sample points, and leakage of noise components from the system clock can be synthesized as a harmonic of F = 1024 times. That is, according to the prior art, when a noise waveform sample value is prepared in a memory, a noise waveform having 16386 sample points is prepared, which requires a memory amount of 1/1024.

  In actual hardware, for example, a memory of a basic signal waveform of one cycle with 1024 sample points, which is 1/16 of 16384 sample points used in the fast Fourier transform, is prepared as hardware. Sets the basic signal waveform of noise for a period. As a result, even when the noise waveform including the noise frequency is changed by changing the AD converter 3 or the like in the operating state, the noise removal function can be easily realized.

  FIG. 14 is a diagram illustrating a configuration example of a signal waveform generation apparatus according to another embodiment of the present invention. In FIG. 14, 20 is a signal waveform generating device, 21 is a counter for counting k, 22 is a multiplier for calculating k * F1, and 23 is bit-shifting k * F1 (S * F1 / N) * k. This is a shift circuit for output. Reference numeral 24 denotes an adder for adding the phase term P1 to (S * F1 / N) * k, and 25 denotes a remainder obtained by dividing Fix {(S * F1 / N) * k + P1} by S. 27 is a selection unit that uses the sample value of the sample point corresponding to the calculated remainder of the first basic signal waveform in 27 as an address to read out, and 26 is a basic unit that initially sets the sample value of the first basic signal waveform in the memory 27. The signal waveform initial setting unit 27 is a memory for storing sample values of the first basic signal waveform, and 28 is a multiplier for multiplying the sample value Q1 (k) read from the memory 27 by the amplitude rate B1. .

  Also, 29 is a multiplier for calculating k * F2, 30 is a shift circuit that bit-shifts k * F2 and outputs (S * F2 / N) * k, and 31 is (S * F2 / N) * k. An adder for adding the phase term P2, 32 calculates the remainder of the result of dividing Fix {(S * F2 / N) * k + P2} by S, and calculates the second basic signal waveform in the memory 34 A selection unit 33, which uses the sample value of the sample point corresponding to the remainder as an address to read, is a basic signal waveform initial setting unit that initially sets the sample value of the second basic signal waveform in the memory 34. Reference numeral 34 denotes a memory for storing sample values of the second basic signal waveform, reference numeral 35 denotes a multiplier for multiplying the sample value Q2 (k) read from the memory 34 by the amplitude rate B2, and reference numeral 36 denotes an output from the multiplier 28. B1 * Q1 (k) and B2 * Q2 (k) output from the multiplier 35 are added to output the final sample value Q (k) of the synthesized signal waveform.

The following embodiment is also possible. First and second basic signal waveforms are prepared with sample values of sample points S, and harmonics F1 times and F2 times of the respective basic signal waveforms are generated as synthesized signal waveforms of sample values of sample points N. A signal obtained by multiplying and adding predetermined amplitude ratios B1 and B2 is used as a final synthesized signal waveform. At this time, with respect to the first basic signal waveform, the sample start point is P1, and the remainder of the integer part when (S * F1 / N) * k + P1 is divided by S is used as the read position in the memory 27 to obtain the sample value. Take out. Also, for the second basic signal waveform, the sample start point is P2, and the sample value is extracted with the remainder of the integer part when (S * F2 / N) * k + P2 is divided by S as the read position in the memory 34. . P1 and P2 are phase terms used for shifting the phase.

  The result obtained by multiplying each sample value by the amplitude rates B1 and B2 is used as the kth sample value of the composite signal waveform, and each sample of the plurality of composite signal waveforms having different sample start points P1 and P2 and amplitude rates B1 and B2 Add the values to obtain the final composite signal waveform.

  The operation in the case of adding two types of synthesized signal waveforms by the signal waveform generating device 20 shown in FIG. 14 will be described. P1 and P2 are set as phase terms, and B1 and B2 are set as amplitude rates. The basic signal waveforms set in the memory 27 and the memory 34 may be different basic signal waveforms or the same.

  In FIG. 14, the multiplier 22 calculates k * F1 from the output k of the counter 21 and a predetermined set value F1, and the shift circuit 23 bit-shifts k * F1 to obtain (S * F1 / N) * k. Output. The adder 24 adds the phase term P1 to the value of (S * F1 / N) * k, and the selection unit 25 takes out the remainder as a result of dividing Fix {(S * F1 / N) * k + P1} by S, The address A1 (k) of the memory 27 is assumed. As a result, the sample value Q1 (k) of the first basic signal waveform stored at the address A1 (k) in the memory 27 is read. The multiplier 28 multiplies Q1 (k) and the amplitude rate B1.

  On the other hand, the multiplier 29 calculates k * F2 from the output k of the counter 21 and a predetermined set value F2, and the shift circuit 30 bit-shifts k * F2 and outputs (S * F2 / N) * k. . The adder 31 adds the phase term P2 to the value of (S * F2 / N) * k, and the selection unit 32 takes out the remainder resulting from dividing Fix {(S * F2 / N) * k + P2} by S, Address A2 (k) of the memory 34 is assumed. As a result, the sample value Q2 (k) of the second basic signal waveform stored at the address A2 (k) in the memory 34 is read out. The multiplier 35 multiplies Q2 (k) and the amplitude rate B2. The adder 36 adds the output B1 * Q1 (k) of the multiplier 28 and the output B2 * Q2 (k) of the multiplier 35, and the result is the sample value Q (k of the final synthesized signal waveform. ).

  By using the signal waveform generation device 20 shown in FIG. 14, for example, a noise composite signal waveform is generated for each type of leaking noise, and the final processing of all the noises is performed by adding the generated composite signal waveforms. A synthetic signal waveform can be generated.

  Specifically, a synthesized signal waveform of noise leaking from the clock system is generated by the counter 21, multiplier 22, shift circuit 23, adder 24, selection unit 25, memory 27, and multiplier 28, for example, around a motor or the like. The counter 21, multiplier 29, shift circuit 30, adder 31, selection unit 32, memory 34, and multiplier 35 generate a synthesized signal waveform of noise generated from the environment of the above, and add these two types of synthesized signal waveforms to the adder. By adding in 36, it is possible to generate a combined signal waveform in which different types of noise are combined.

  The present invention can also be processed separately according to the frequency level. For example, in the spectrum shown in FIG. 3 (A), a synthesized signal waveform of noise corresponding to the low frequencies of 0 Hz and 3 Hz is converted into a counter 21, a multiplier 22, a shift circuit 23, an adder 24, a selection unit 25, Of the spectrum shown in FIG. 3A, which is generated by the memory 27 and the multiplier 28, a synthesized signal waveform of noise corresponding to a high frequency of 9 Hz or 15 Hz is converted into a counter 21, a multiplier 29, a shift circuit 30, The final synthesized signal waveform can be generated by adding the two types of noise waveforms generated by the adder 31, the selector 32, the memory 34, and the multiplier 35, and adding the generated noise waveforms by the adder 36.

  In the example of FIG. 14, an example of a circuit that adds two types of synthesized signal waveforms to obtain a final synthesized signal waveform is shown, but a circuit that adds and synthesizes three or more types of synthesized signal waveforms can also be used. It goes without saying that the same configuration can be adopted.

  In the above-described embodiment of the present invention, the example in which the basic signal waveform of one cycle is prepared in the memory has been described. However, in the present invention, the basic signal waveform prepared in the memory is not limited to that of one cycle. , A signal having a plurality of cycles may be prepared as a basic signal waveform.

  The above signal waveform generation and noise removal processing can be realized by a computer and a software program, and the program can be provided by being recorded on a computer-readable recording medium or via a network. It is.

The features of the technology described above are as follows.

(Supplementary note 1) A signal waveform generation device that generates, from a basic signal waveform having sample values of sample points S, F harmonics of the basic signal waveform as a composite signal waveform of sample values of sample points N (S <N). There,
A memory for accumulating sample values of the basic signal waveform in chronological order;
(S * F / N) * k (where k = 0, 1, 2,..., N−1) calculating means for calculating a remainder when the integer part is divided by S;
Sample value determining means for extracting a sample value of the basic signal waveform from a position corresponding to the calculated remainder in the memory and using the extracted sample value as the k-th sample value of the combined signal waveform; A signal waveform generator characterized by the above.

(Supplementary note 2) Generate a harmonic of F times the basic signal waveform from a noise basic signal waveform having sample values of sample points S as a synthesized noise signal waveform of sample values of sample points N (S <N). A noise removing device that removes noise from the mixed noise signal waveform generated from the mixed signal,
A memory for accumulating sample values of the basic signal waveform of the noise in chronological order;
(S * F / N) * k (where k = 0, 1, 2,..., N−1) calculating means for calculating a remainder when the integer part is divided by S;
A sample value of the basic signal waveform of the noise is extracted from a position corresponding to the calculated remainder in the memory, and the extracted noise value is used as the k-th sample value of the synthesized noise signal waveform. Signal waveform generating means for generating
Subtracting means for subtracting the synthesized noise signal waveform from the signal mixed with the noise.

(Supplementary Note 3) Using a memory that accumulates the sample values of the sample points S of the basic signal waveform in time series order, the harmonics F times the basic signal waveform from the basic signal waveform having the sample values of the sample points S are sampled N A signal waveform generation method for generating a composite signal waveform of sample values of (S <N),
Calculating the remainder when the integer part of (S * F / N) * k (where k = 0, 1, 2,..., N−1) is divided by S;
Extracting a sample value of the basic signal waveform from a position corresponding to the calculated remainder in the memory, and determining the extracted sample value as the k-th sample value of the synthesized signal waveform. A signal waveform generation method.

(Supplementary Note 4) Using a memory that accumulates sample values of sample points S of a noise basic signal waveform in chronological order, from a noise basic signal waveform having sample values of sample points S to F times higher harmonics of the basic signal waveform Is generated as a composite noise signal waveform having a sample value of sample points N (S <N), and noise is removed from a signal mixed with noise using the generated composite noise signal waveform,
Calculating the remainder when the integer part of (S * F / N) * k (where k = 0, 1, 2,..., N−1) is divided by S;
A sample value of the basic signal waveform of the noise is extracted from a position corresponding to the calculated remainder in the memory, and the extracted noise value is used as the k-th sample value of the synthesized noise signal waveform. A step of generating
Subtracting the synthesized noise signal waveform from the signal mixed with the noise.

(Appendix 5) In the noise removal method described in Appendix 4,
Inputting a signal mixed with noise and performing a Fourier transform;
Selecting a Fourier component of the noise frequency from the result of Fourier transform;
An inverse Fourier transform of the Fourier component of the selected noise frequency;
Extracting a sample value of the number S of sample points from the signal waveform resulting from the inverse Fourier transform and storing it in the memory, and initializing the basic signal waveform of the noise.

(Supplementary Note 6) Using a memory that accumulates the sample values of the sample points S of the basic signal waveform in time series order, the harmonics F times the basic signal waveform from the basic signal waveform having the sample values of the sample points S are sampled N A signal waveform generation program for causing a computer to execute a signal waveform generation method for generating a composite signal waveform of sample values of (S <N),
(S * F / N) * k (where k = 0, 1, 2,..., N−1) is calculated by dividing the integer part by S;
Extracting a sample value of the basic signal waveform from a position corresponding to the calculated remainder in the memory, and determining the extracted sample value as the kth sample value of the combined signal waveform;
A signal waveform generation program to be executed by a computer.

(Supplementary note 7) Using a memory that accumulates the sample values of sample points S of the noise basic signal waveform in time series order, from the noise basic signal waveform having the sample values of sample points S to F times higher harmonics of the basic signal waveform Is generated as a synthesized noise signal waveform of sample values of the number of sample points N (S <N), and the computer is caused to execute a noise removal method for removing noise from a signal mixed with noise using the generated synthesized noise signal waveform. A noise elimination program for
(S * F / N) * k (where k = 0, 1, 2,..., N−1) is calculated by dividing the integer part by S;
A sample value of the basic signal waveform of the noise is extracted from a position corresponding to the calculated remainder in the memory, and the extracted noise value is used as the k-th sample value of the synthesized noise signal waveform. Processing to generate
Subtracting the synthesized noise signal waveform from the signal mixed with the noise,
A noise reduction program for running on a computer.

It is a principle explanatory view of synthetic signal waveform generation . It is a figure which shows the leakage of the noise to an original cosine wave. It is a figure which shows the result of the Fourier-transform of the signal which noise mixed. It is a figure which shows the basic signal waveform of noise, and the synthetic | combination signal waveform of noise. It is a figure which shows the signal after noise, the synthetic signal waveform of noise, and the signal after noise removal. It is a diagram illustrating a configuration example of the signal waveform generation unit. It is a figure which shows an example of a signal waveform generation process flow. It is a figure which shows the sample value etc. corresponding to each sample value of a basic signal waveform, and each sample point of a synthetic | combination signal waveform. It is a schematic diagram of a system configuration example of the implementation of the embodiment of the present invention. Is a diagram showing details of a system configuration example of the implementation of the embodiment of the present invention. It is a figure which shows the initialization process flow of a basic signal waveform. It is a figure which shows a noise removal process flow. It is a figure which shows the example of calculation of the basic signal waveform prepared in memory. It is a diagram illustrating another configuration example of the signal waveform generation unit. It is a figure which shows the system structural example of the signal processing apparatus of a prior art.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Noise removal apparatus 2 Antenna 3 AD converter 4 Fourier transform apparatus 10,20 Signal waveform generation apparatus 11,26,33 Basic signal waveform initial setting part 12,27,34 Memory 13,21 Counter 14,22,28,29, 35 Multiplier 15, 23, 30 Shift circuit 16, 25, 32 Selection unit 17 Subtraction circuit 24, 31, 36 Adder

Claims (6)

A receiver for receiving an input signal;
A transform unit for Fourier transforming the input signal;
A selection unit that selects a Fourier component of the frequency of noise included in the input signal based on the result of the Fourier transform;
A generator that performs an inverse Fourier transform on the selected Fourier component and generates a basic signal waveform of the noise;
A storage unit that accumulates sample values of S sample points (S is a positive integer) extracted from the generated basic signal waveform in chronological order;
(S * F / N) * k (where k = 0, 1, 2,..., N−1) an arithmetic part that calculates a remainder when the integer part is divided by the number of sample points S;
Extracting a sample value of the basic signal waveform of the noise from a position in the storage unit corresponding to the calculated remainder, determining the extracted sample value as a k-th sample value of a synthesized noise signal waveform, and A synthesis unit that generates a synthesized noise signal waveform that is an F-fold harmonic of the basic signal waveform;
A subtracting unit that subtracts the synthesized noise signal waveform from the input signal.
The noise removal apparatus characterized by the above-mentioned. The basic signal waveform of the noise is a signal waveform of one cycle of the noise.
The noise removing apparatus according to claim 1, wherein: In the noise removal method of the noise removal device,
A receiving unit included in the noise removing device receives an input signal;
A step of transforming the input signal by a conversion unit included in the noise removing device;
A step of selecting a Fourier component of a frequency of noise included in the input signal based on a result of the Fourier transform by a selection unit included in the noise removing device;
A step of generating a basic signal waveform of the noise by performing an inverse Fourier transform on the selected Fourier component by a generation unit included in the noise removing device;
Accumulating sample values of S sample points (S is a positive integer) extracted from the generated basic signal waveform in a time-series order in the storage unit of the noise removing device;
Remainder when the arithmetic unit of the noise removal device divides the integer part of (S * F / N) * k (where k = 0, 1, 2,..., N−1) by the sample point S. Calculating steps,
A synthesis unit included in the noise removal device extracts a sample value of the basic signal waveform of the noise from a position in the storage unit corresponding to the calculated remainder, and uses the extracted sample value as the kth of the synthesized noise signal waveform. Determining a first sample value and generating a synthesized noise signal waveform that is an F-fold harmonic of the basic signal waveform;
A subtracting unit included in the noise removing device subtracting the synthesized noise signal waveform from the input signal.
The noise removal method characterized by the above-mentioned. The basic signal waveform of the noise is a signal waveform of one cycle of the noise.
The noise removal method according to claim 3. In the noise removal program of the noise removal device,
Computer
Receiving an input signal;
Fourier transforming the input signal;
Selecting a Fourier component of the frequency of noise included in the input signal based on the result of the Fourier transform;
Performing an inverse Fourier transform on the selected Fourier component to generate a basic signal waveform of the noise;
A step of accumulating sample values of S sample points (S is a positive integer) extracted from the generated basic signal waveform in a time series in a storage unit of the computer;
Calculating a remainder when the integer part of (S * F / N) * k (where k = 0, 1, 2,..., N−1) is divided by the number of sample points S;
Extracting a sample value of the basic signal waveform of the noise from a position in the storage unit corresponding to the calculated remainder, determining the extracted sample value as a k-th sample value of a synthesized noise signal waveform, and Generating a composite noise signal waveform that is an F harmonic of the basic signal waveform;
Subtracting the synthesized noise signal waveform from the input signal;
Noise removal program to be executed. The basic signal waveform of the noise is a signal waveform of one cycle of the noise.
The noise removal program according to claim 5, wherein:

Images