JPH0744500B2 - Encoding / decoding method and device suitable for variable rate transmission - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an encoding / decoding system and apparatus suitable for compressing and transmitting information such as voice / music signals in a transmission network having a variable transmission rate.

[0002]

2. Description of the Related Art In order to efficiently transmit information contained in a signal such as voice / music by using a line having a limited transmission capacity, reducing the information is called band compression.
An adaptive differential pulse code modulation method [ADPCM] method (Digital coding of Waveforms (Dig) by Giant et al.
ital coding of waveform
s), Prentice Hall (Prentice-H)
all), 1984, p. 308; hereinafter referred to as "reference 1"), and adaptive transform coding [ATC] (Zelinski et al.), which is coding using linear transform, and approaches to adaptive transformer. Forms, Speech, Coding, At Law,
Bit rate (Approaches to Ada)
ptive Transfoam Speech Co
ing at Low Bit Rate, IEEE TRANSACTIONS ON A
SSP) Vol. 27, No. 1, 1979, p. 895;
[Reference 2] below, and a method of dividing into multiple bands in the frequency domain and adaptively assigning the number of quantization bits to each band signal and coding (Yves Francis (YvesFran)).
A Musicum Source Codec for Digital Audio Broadcasting and Storage (A. Cois Dehery)
MUSICAM Source Code for
Digital Audio Broadcasti
ng and Storage), Proceedings of ICA SSP 91
(Proceedings of ICASSP'9
1) vol. 5, 1991, pages 3605-3608; hereinafter referred to as "Reference 3"). ADPC
The M method is a standardized method for encoding a narrow band (7 kHz) signal, but it is a wide band (15 to 20 kHz).
z) ATC or band division type coding is more effective for high-quality coding of signals. On the other hand, in a multiplexed transmission network that multiplexes and transmits some channels, there is a network in which the data amount of each channel is variable according to the fluctuation of traffic in the network. An example of such a transmission network in which the transmission rate for each channel is variable is an ATM network in which some data is discarded. In the ATM network, the information of each channel is collected in a unit called a cell of a predetermined format and sent to the network, but cell discard occurs in the network according to traffic. In order to transmit voice / music, etc. in a network in which the transmission rate is variable in this way, an encoding method that causes little deterioration in quality even when data is discarded is required. Conventionally, in the ADPCM system, a method called embedded in which lower bits of data are deleted when data is discarded and the lower bits lost on the decoding side are decoded as zero (Goodman, Embedded DCM for four) Variable Bit Rate Transform (Embedded D)
PCM for Variable Bit Rate
Transform, IEE Transactions on Communications (IEEE T)
RANSACTIONS ON COMMUNICAT
IONS) Volume 28, No. 7, 1980, 1040-104
An embedded ADPCM system adapted to page 6; hereinafter, "Reference 4") has been known.

[0003]

When the transmission data is reduced in the encoding method effective for the high quality encoding of the wide band signal described in the conventional example, the received signal cannot be decoded and the quality is remarkably increased. There is a problem that it is deteriorated and cannot be applied to a transmission network of variable rate substantially. It is an object of the present invention to provide a coding / decoding system and apparatus that can obtain good coding quality even when transmission data is reduced and that can reduce adaptive quantized bit allocation information.

[0004]

According to the encoding / decoding method of the present invention, on the encoding side, when the quantum bit is adaptively assigned to each band signal divided into a plurality of bands, the transmission rate is α and Two types of adaptive bit allocation such that β (α> β) are performed, the band signal is quantized by the number of quantized bits allocated with the transmission rate α, and each quantized signal of the band signal is Separated into the lower bits and the upper bits corresponding to the difference bit number between the two kinds of quantization bit numbers assigned at the two kinds of transmission rates, and collecting the separated lower bit signals as a second separated signal. Collecting the remaining high-order bit signals as a first separation signal, the first and second separation signals, the signal of the number of difference bits between the two kinds of quantization bit allocation numbers, and the transmission rate assigned according to β And the signal of the number of quantization bits is transmitted, and on the decoding side, the first separated signal is distributed to the upper bits of each band signal according to the information of the number of quantization bits allocated according to the transmission rate β. With
According to the difference signal between the two types of quantized bit allocation numbers, the transmission rate is α according to the determination signal that informs whether the transmission rate at the time of transmission was α or β.
If it is, the second separated signal is distributed. If it is β, an arbitrary signal is distributed to the lower bits of each band signal instead of the second separated signal, and the second signal is distributed for each band. A high-order bit and a low-order bit are combined to generate a combined signal, and the combined signal has a difference bit number between the two kinds of quantization bit allocation numbers and a quantization bit number allocated according to the transmission rate β. It is characterized in that inverse quantization is performed with the number of quantization bits obtained by adding. Further, the encoding device of the present invention, on the encoding side, when quantizing each band signal divided into a plurality of frequency bands by the band division circuit, inputs each band signal, and the transmission rate is α A first adaptive bit allocation circuit that allocates the number of quantization bits to each band signal so that the band rate signals are input to each band signal so that the transmission rate becomes β (α> β). A second adaptive bit allocation circuit for allocating a quantization bit number; a subtracter for subtracting an output of the second adaptive bit allocation circuit from an output of the first adaptive bit allocation circuit to obtain a difference bit number between the two; Input each band signal,
A quantizer that quantizes each band signal according to the output of the first adaptive bit allocation circuit, each quantized signal of the band signal that is the output of the quantizer, and a difference bit that is the output of the subtractor. Using the number as an input, each quantized signal in the previous period is separated into each lower bit and each remaining upper bit corresponding to each difference bit number, and each separated lower bit is collected to generate a second separated signal, and And a separation circuit for collecting the respective upper bits to generate a first separation signal, and supplying the output of the subtracter and the output of the second adaptive bit allocation circuit to the multiplexer. Characterize. Furthermore, the decoding device of the present invention, on the decoding side,
A signal having a difference bit number between two types of quantization bit numbers assigned so that the transmission rate becomes α and β and the number of quantization bits assigned to each band signal so that the transmission rate becomes β. An adder for adding the signal of No. 1, a signal of the number of differential bits, a signal of the number of quantized bits allocated in accordance with the transmission rate β, a first separated signal and a second separated signal, When the transmission rate is α, the first separation signal is output, and when the transmission rate is β, the first separation signal is output according to the determination signal indicating whether the transmission rate is α or β. Of the separated signals are distributed to the upper bits of the quantized signal of each band signal according to the signal of the quantized bit number allocated according to the transmission rate β, and the second component is divided according to the signal of the differential bit number. A signal is distributed to the lower bits of the quantized signal of each band signal, and a combining circuit that combines the distributed upper bits and lower bits for each band signal, and each band signal that is the output of the combining circuit An inverse quantizer for inversely quantizing the quantized signal according to the output of the adder, and a band synthesizing circuit for inputting and synthesizing each band signal which is the output of the inverse quantum below. To do.

[0005]

According to the present invention, different kinds of quantized bits having different transmission rates are assigned to each band signal, and the difference bit number between both assigned bits is the quantized bit assigned at the maximum transmission rate. Separate from the lower bit side of the quantized signal of each band signal quantized by a number, delete this separated signal when the transmission rate is compressed, and at the same time insert any signal into the separated signal on the decoding side to combine, By inverse quantization, it is possible to easily cope with the switching of the transmission rate, and the quality deterioration when data deletion occurs is only the deterioration of the quantization accuracy due to the loss of the low-order few bits. It is possible to maintain the quality of conversion.

Further, according to the present invention, when transmitting two kinds of quantized bit allocation information having different transmission rates to the decoding side,
By transmitting the information of the number of quantized bits allocated at the lower transmission rate and the difference information of the two types of allocated bits, it is possible to reduce the information amount of the allocated number of bits.

[0007]

1 shows a first embodiment of the system according to the invention. The encoder comprises an input terminal 1, a band division circuit 2, a quantizer 3, a multiplexer 5, a first adaptive bit allocation circuit 6, a second adaptive bit allocation circuit 7, a separation circuit 8 and a subtractor 9. The input signal supplied from the input terminal 1 at the sampling rate f is divided into N frequency bands (M is set to be an integral multiple of N) in the band division circuit 2 in units of a predetermined M samples. It The band division circuit 2 generally includes a plurality of QMFs (Quadrature M).
error filter) (Reference 3). The N band signals output from the band division circuit 2 are
Quantizer 3, first adaptive bit allocation circuit 6 and second
It is supplied to the adaptive bit allocation circuit 7. The first adaptive bit allocation circuit 6 assigns the number of quantization bits to each band signal so that the sum of the number of quantization bits assigned to each band signal becomes equal to α according to the power of each band signal. It is assigned and supplied to the quantizer 3. The quantizer 3 includes a number of quantizers equal to the number of division bands N, and each band signal is quantized by the corresponding quantizer according to the output of the first applied bit allocation circuit 6. , Each quantized signal is supplied to the separation circuit 8. On the other hand, in the second adaptive bit allocation circuit 7, the number of quantized bits corresponding to the power of each band signal is calculated so that the transmission rate becomes β in the same manner as in the first adaptive bit allocation circuit 6, and each band is calculated. Output the number of quantization bits assigned to the signal. The outputs of the first adaptive bit allocating circuit 6 and the second adaptive bit allocating circuit 7 are supplied to the subtracter 9, the difference bit number between the two in each band signal is calculated, and then supplied to the separation circuit 8. At the same time, the output signal 11 of the second adaptive bit allocation circuit 7 and the output signal 10 of the subtractor 9 are supplied to the multiplexer 5. Separation circuit 8
Then, the N quantized signals output from the quantizer 3 are separated into upper bits and lower bits by a method described later,
The upper bits of the N quantized signals are combined into the first separated signal 23 ₁ , and the lower bits of the N quantized signals are combined into the second separated signal 23 ₁ .
The separated signal 23 ₂ is supplied to the multiplexer 5. In the multiplexer 5, the first separation signal 23 ₁ and the second separation signal 23 ₂ supplied from the separation circuit 8, the output signal 11 of the second adaptive bit allocation circuit 7, and the output signal 10 of the subtractor 9
And are multiplexed and sent to the transmission line 12.

The decoder comprises a separating device 13, a synthesizing circuit 14,
It includes an adder 17, an inverse quantizer 18, a band synthesizing circuit 19, and an output terminal 20. The multiplexed signal from the transmission line 12 is demultiplexed by the demultiplexing device 13, and it is determined whether or not the first demultiplexed signal 25 ₁ , the second demultiplexed signal 25 ₂ and the second demultiplexed signal 25 ₂ are deleted due to the change of the transmission rate. The transmission rate determination signal 29 that is represented, the difference signal 15 of the quantized bit allocation that matches the transmission rate α and the quantized bit allocation that matches the transmission rate β, and the quantized bit allocation signal 16 that matches the transmission rate β are combined. It is supplied to the circuit 14. The synthesizing circuit 14 obtains the quantized bit allocation difference signal 15 and the quantized bit allocation signal 16 matched to the transmission rate β, and uses the method described later to generate the first separated signal 25 ₁ and the second separated signal 2
The quantized signals of N band signals from 5 ₂ are combined and supplied to the inverse quantizer 18. The dequantizer 18 includes N dequantizers therein similarly to the quantizer 3, and each dequantizer quantizes in accordance with the quantized bit allocation difference signal 15 and the transmission rate β. Dequantize each quantized signal according to the output of the adder 17 for obtaining the sum with the bit allocation signal 16, that is, the quantized bit allocation according to the transmission rate α,
It is supplied to the band synthesis circuit 19. In the band synthesis circuit 19,
The coding side synthesizes the signals divided into N bands and supplies the reproduced signal to the output terminal 20.

In the first and second bit allocation circuits, the number of quantization bits is allocated according to the power of each band signal, and the psychological auditory sense of human beings is used to reduce the number of difficult-to-hear signals and make them easily audible. In some cases, a method of allocating a large number of quantization bits to a signal is used (Reference 3).

Next, the details of the separation circuit 8 shown in FIG. 1 will be described with reference to FIG. First adaptive bit allocation circuit 6
, The number of quantized bits assigned to each of the N band signals is expressed as a vector P = h ₁₁ ,, h ₁₂ , .., h _1n (h ₁₁ + h ₁₂ + .. + h _1n = Mα /
Expressed in fs, also the vector Q = h ₂₁ the output of the second adaptive bit allocation circuit 7, h _22, ······, expressed in h _2n, further vector P-Q vector R = k _1, k ₂ , ..., K _n (n = number of band divisions N). Then, the first separator 21 ₁ , the second separator 21 ₂ , and the nth separator 21 _{n of FIG.}
H _11, h _12, respectively, ······, h _1n-bit data is to be input from the quantizer 3 in. Then, according to the vector R that is the output of the subtractor 9, the first separator 21 ₁
Of the ₁₁ bits, the lower k ₁ bit is separated, and the remaining upper (h ₁₁ −k ₁ ) bit is separated into the first transmission signal generator 22 ₂ .
The signal is supplied to the transmission signal generator 22 ₁ . Similarly, the second separator 21 ₂ to the second transmission signal generator 22 ₂ to separate the lower k ₂ bits of the bit, remaining upper (h ₁₂ -k ₂₎ bits of the first transmission signal generator 22 ₁ To the n-th separator 21 _n
Supplies the lower k _n bits of the h _1n bits to the second transmission signal generator 22 ₂ and the remaining upper (h _1n −k _n ) bits to the first transmission signal generator 22 ₁ . First transmission signal generator 22 ₁
Is a first separator 23 that combines the data supplied from the first separator 21 ₁ , the second separator 21 ₂ , and the nth separator 21 _n.
1 is generated, and the second transmission signal generator 22 ₂ is used as the first separator 2
The data supplied from 1 ₁ , the second demultiplexer 21 ₂ and the nth demultiplexer 21 _n are combined to generate a second demultiplexed signal 23 ₂ which is supplied to the multiplexer 5. In FIG. 2B, the first reception signal separator 24 ₁ and the second reception signal separator 24 ₂ are
The first separated signal 25 ₁ and the second separated signal received through the transmission line
The separation signal 25 ₂ is input. First received signal separator 24
₁ is the first separated signal 25 ₁ according to the quantized bit allocation signal 16 that matches the transmission rate β, that is, the vector Q.
Is separated into h ₂₁ , _{n 22} , ..., h _2n- bit data,
The first synthesizer 26 ₁ , the second synthesizer 26 ₂ , ...
... Supply to the nth combiner 26 _n . On the other hand, the second separated signal 25 ₂ may be deleted in accordance with the change in the transmission rate, and when deleted, an operation of inserting an arbitrary signal in place of the second separated signal 25 ₂ is required. For this reason,
The second received signal separator 24 ₂ separates the second separated signal 25 ₂ into k ₁ , k ₂ , ..., K _n bits according to the difference signal 15 of the quantized bit allocation, that is, the vector R. The signal is supplied to the selector 27, and at the same time, the signal generator 28 outputs k ₁ , k ₂ ,
...... separates the k _n bits supplied to the selector 27, the signal generating噐28 k _1, k ₂ simultaneously, ......, k _n
An arbitrary bit signal is supplied to the selector 27. At this time, an example of the signal generated by the signal generator 28 will be shown later. The selector 27 receives the transmission rate judgment signal 29 from the demultiplexing device 13 which indicates whether or not the second separated signal 25 ₂ has been deleted due to the change in the transmission rate, and if it has not been deleted, the second received signal separator 28. Select the output of the first
Synthesizer 26 ₁ , second synthesizer 26 ₂ , ..., Nth
It is supplied to the synthesizer 26 _n . The first combiner 26 ₁ adds the k ₁ bit supplied from the selector 27 to the lower part of the h ₂₁ bit, and the second combiner 26 ₂ adds the k ₂ bit to the lower part of the h ₂₂ bit and the nth combiner. 26 _n adds k _n bits to h _2n bits, synthesizes quantized signals of n band signals, and supplies the resulting signals to the inverse quantizer 18. A zero signal is considered as the signal generated by the signal generation circuit 28. Further, as a method of reducing the error, if it is represented by a two's complement expression, for example, when the lower 2 bits are deleted, "01" is given if the upper bit is positive, and "10" if it is negative. "Generate a,
Similarly, when the k bits are deleted, if the upper bit is positive, the signal with the largest k bits to be deleted is 0 and the other is 1, and if the higher bits are negative, the largest k bits is 1 and the others are A method of generating a signal with 0 as 0 can be considered. A second embodiment according to the method of the present invention is shown in FIG. Functional blocks given the same numbers as in FIG. 1 have the same functions. The difference from FIG. 1 is that a linear conversion circuit 31 and a linear inverse conversion circuit 32 are provided. The linear conversion circuit 31 obtains an M-sample discrete signal input from the input terminal 1 and performs M-point discrete linear conversion. Fast Fourier transform (FFT) or the like is used as the linear transform. In this embodiment, finer frequency spectrum analysis can be performed by performing linear transformation, and adaptive bit allocation (reference 3) to each band signal using human auditory characteristics can be performed using the obtained transformation coefficient. It will be easy. The linear inverse transform circuit 32 transforms each transform coefficient dequantized on the composite side into a time domain signal.

[0011]

As described above in detail, according to the present invention, the quality deterioration caused by the deletion of data in the variable rate transmission network is caused by the loss of the lower several bits. Since it is only deteriorated, it is possible to provide an encoding / decoding system and apparatus that can maintain good encoding quality and reduce adaptive quantized bit allocation information for transmission.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of an encoding / decoding device that is a first embodiment of the present invention.

FIG. 2 is a block diagram showing details of a separation circuit and a synthesis circuit in the encoding / decoding device of FIG.

FIG. 3 is a block diagram showing the configuration of an encoding / decoding device that is a second embodiment of the present invention.

[Explanation of symbols]

 1 Input Terminal 2 Band Division Circuit 3 Quantizer 5 Multiplexing Device 6,7 Adaptive Bit Allocation Circuit 8 Separation Circuit 9 Subtractor 13 Separation Device 14 Synthesis Circuit 15 Inverse Quantizer 17 Adder 19 Band Synthesis Circuit 20 Output Terminal 29 Transmission rate determination signal 31 Linear conversion circuit 32 Linear inverse conversion circuit

Claims (3)

[Claims] 1. On the encoding side, when a quantum bit is adaptively assigned to each band signal divided into a plurality of bands,
Two kinds of adaptive bit allocation such that the transmission rates are α and β (α> β) are performed, and the band signal is quantized by the number of quantization bits allocated by the transmission rate α, and each of the band signals is quantized. The quantized signal is separated into the lower bits and the remaining upper bits corresponding to the difference bit number between the two kinds of quantization bit numbers allocated at the two kinds of transmission rates, and the separated lower bit signals are collected, 2 separate signals,
The remaining high-order bit signals are collected and used as a first separation signal, the first and second separation signals, the signal of the number of difference bits between the two kinds of quantized bit allocation numbers, and the transmission rate β. And the signal of the number of quantized bits allocated according to the above, and on the decoding side, according to the information of the number of quantized bits allocated according to the transmission rate β, the first separation signal of each of the band signals is transmitted. The transmission rate was α according to the difference signal between the two kinds of quantized bit allocation numbers by the determination signal that distributed to the upper bits and notified whether the transmission rate at the time of transmission was α or β. In the case of β, in the case of β, an arbitrary signal is distributed to the lower bits of each of the band signals instead of the second separated signal, and the upper bits are divided into the upper bits of each band. under A quantum that is obtained by combining bits to generate a combined signal, and adding the combined signal to the difference bit number between the two kinds of quantized bit allocation numbers and the quantized bit number allocated according to the transmission rate β. An encoding / decoding method characterized by performing inverse quantization with the number of encoded bits. 2. On the encoding side, when quantizing each band signal divided into a plurality of frequency bands by a band dividing circuit, each band signal is input, and each band signal is input so that the transmission rate becomes α. First to allocate the number of quantization bits to the band signal
A second adaptive bit allocation circuit for allocating a quantization bit number to each band signal so that the transmission rate becomes β (α> β), A subtracter that subtracts the output of the second adaptive bit allocation circuit from the output of the first adaptive bit allocation circuit to obtain the difference bit number between the two, and the respective band signals as input, A quantizer that quantizes each band signal according to an output, each quantized signal of the band signal that is the output of the quantizer, and the number of difference bits that is the output of the subtractor, as input The signal is separated into lower bits and remaining upper bits corresponding to the respective difference bit numbers, the separated lower bits are collected to generate a second separated signal, and the remaining upper bits are collected to form a first Separation of No. at least and a separation circuit for generating the output of the subtractor, the encoding apparatus and supplying an output of said second adaptive bit allocation circuit multiplexer. 3. On the decoding side, the transmission rates are α and β.
The signal of the number of difference bits between the two types of quantization bit numbers assigned so that the signal of the number of quantization bits assigned to each band signal so that the transmission rate is β is added. The adder, the signal of the difference bit number, the signal of the quantization bit number allocated according to the transmission rate β, the first separated signal and the second separated signal are input, and the transmission at the time of transmission is performed. According to a determination signal indicating whether the rate is α or β, the first separation signal is transmitted when the transmission rate is α, and the first separation signal is transmitted when the transmission rate is β. According to the signal of the number of quantized bits allocated according to the rate β, it is distributed to the upper bits of the quantized signal of each band signal, and the second separated signal is divided into the respective bands according to the signal of the number of differential bits A quantized signal of each band signal, which is an output of the synthesizing circuit, and a synthesizing circuit for allocating the quantized signal to the lower bit and synthesizing the distributed upper bit and lower bit for each band signal. A decoding device comprising at least a dequantizer for dequantizing in accordance with the output of the adder, and a band synthesizing circuit for inputting and synthesizing each band signal which is the output of the inverse quantum or less.