JPH0335300A - Voice coding and decoding transmission system - Google Patents

Abstract

PURPOSE:To allot an optimum information quantity to the frame under processing by respectively synthesizing and forming the local decoding signals of respective sub-frames and one frame from the coding devices of the different numbers of quantization levels extracted for every frame and transmitting the coding devices corresponding to the allotted patterns of the best characteristics. CONSTITUTION:The local decoding signals are formed from the coding devices 211 to 2nm of the different numbers of the quantization level extracted for every frame and the local decoding signals of the respective sub-frames and one frame are respectively synthesized and formed in accordance with the allotted patterns in such a manner that a specified transfer bit rate is attained. The coding devices corresponding to the allotted patterns of the best characteristics are multiplexed 24 and are transmitted. The optimum information quantity is allotted to the frame under processing by determining the allotted patterns of the information quantity in the frame from the currently processed frame in such a manner.

Description

[Detailed Description of the Invention] (Table of Contents) 1. Existing Industrial Fields of Application Prior Art...Fig. 14 Problems to be Solved by the Invention Means and Action for Solving the Problems First Embodiment... 2. Example 3. Example 4. Effect of the invention, Figure 4.5.6, Figure 7, Figure 8.9.10, Figure 12.13. Overview] A method that compresses information in audio signals and transmits information with high efficiency.
In particular, regarding audio coding transmission systems that change bit allocation patterns in the time domain or frequency domain, processing can be performed by determining the information allocation pattern within a frame from the frame being processed without processing all allocation patterns. The purpose is to provide a high-efficiency coding and transmission device for audio signals that can allocate an optimal amount of information to each frame. generate a signal, combine local decoded signals of each subframe in the frame according to an allocation pattern that makes the transmission bit rate constant, generate a local decoded signal of one frame, and select an allocation pattern with the best characteristics. The configuration is such that corresponding code words are multiplexed and transmitted.

[Industrial Field of Application] The present invention relates to a method for compressing information in an audio signal and transmitting the information with high efficiency, and particularly to an audio encoding/decoding transmission method that changes an allocation pattern in the time domain or frequency domain.

Audio coding and transmission methods that compress the information content of input audio signals are ADPCM that uses adaptive prediction and adaptive quantization, and time domain that adaptively allocates the number of bits in the time domain according to temporal changes in the input audio signal. A method that compresses the amount of information by combining adaptive bit allocation, and also divides the input audio signal into multiple bands and adapts the allocation pattern of the number of bits allocated to each band for each frame according to temporal changes in the input audio signal. There are methods for compressing information using frequency domain adaptive bit allocation, which allocates the number of bits automatically.

With the spread of digital lines in recent years, it is important to reduce the capacity of storage memory in order to make effective use of the lines, and when storing voice information in voice storage and response systems. There is a need for a method that compresses the amount of information in a signal and encodes it with higher efficiency.

[Conventional technology]

It compresses the amount of information by combining conventional ADPCM using adaptive prediction and adaptive quantization with time-domain adaptive bit allocation, which adaptively allocates information (bits) in the time domain according to temporal changes in the input signal. An example of a high-efficiency coding transmission method for audio signals is described in the literature 1) "'16kb/s
Adaptive Pi [・Study of Assignment Predictive Coding Algorithm], 1985 National Conference of the Telecommunications Division of the Institute of Electronics and Communication Engineers, p. 52-4
p, 2-283-2-284j. (Figure 15(a)) In this method, the input audio signal is divided into frames, and the frames are further divided into subframes, and the input audio signal is divided into subframes according to the bit allocation pattern so that the amount of information in the frame is constant. The number of allocated bits for each subframe is changed by using adaptive prediction and adaptive quantization, so the information used to select the bit allocation pattern is based on the residual of the previous frame. It uses signal power, predicts what pattern of residual signal will be generated next, and determines the allocation pattern based on the prediction.

Another example is a coding transmission method that divides the input signal into multiple frequency bands and adaptively changes the allocation pattern of the number of bits allocated to each band for each frame in response to temporal changes in the audio signal. , Reference Shinji Hayashi et al. Backward type high quality adaptive prediction adaptive bit allocation coding method 2'' Journal of the Institute of Electronics and Communication Engineers 86/10 Vol. J69-A, N
o, 10pp, 1234-1.242J.

(FIG. 15(b)) In this method as well, the allocation pattern uses the residual signal power of the previous frame, similar to the determination of the allocation pattern in the time domain.

[Problem to be solved by the invention]

Therefore, in the conventional time domain and frequency band allocation pattern determination methods, the allocation pattern is determined from the residual signal of the frame immediately before the currently processed frame. If the characteristics of the difference signal and the residual signal of the next frame are significantly different, the allocation pattern predicted from the previous frame will not be optimal for the focus allocation of the current frame, and will not be optimal when reproducing the codeword. This will cause deterioration of characteristics.

Therefore, it may be possible to encode all possible allocation patterns. In this case, the optimal allocation pattern can be selected by selecting the one with the best characteristics from among all the patterns. However, encoding all patterns requires expansion of the device configuration.
This causes major problems such as a significant increase in the amount of processing, and cannot be said to be effective when considering overall efficiency.

The present invention makes it possible to allocate the optimal amount of information to the frame being processed without processing all the allocation patterns and by finding the information amount allocation pattern within the frame from the frame currently being processed. The purpose of the present invention is to provide a highly efficient encoding and transmitting device for audio signals.

[Means to solve the problem]

FIG. 1 shows a principle diagram of a speech coding transmission method using an allocation pattern in the time domain. 111 in the figure is M pit t-i
Adaptive quantizer for the number of childization levels, ■13 to 116 are N-M
Adaptive inverse quantizer for the number of bit inverse quantization levels, 117 is N
An adaptive predictor operates based on the result of dequantization of the number of bit levels, 112 is a selection/multiplexing unit, and 118 is a locally decoded signal quality evaluation unit.

Further, FIG. 2 shows a principle diagram of a speech coding transmission system using an allocation pattern in the frequency domain. 1 in the diagram
2 is a band division filter, 211 to 2 nm is an encoder with a different number of quantization levels, 23 is a locally decoded signal quality evaluation unit, 2
4 is a selection/multiplexing section.

In order to achieve the above object, the present invention uses an allocation pattern in the time domain as a first means.
An embedded encoder includes an adaptive dequantizer 113 for the number of bit dequantization levels and an adaptive predictor 117 that operates based on the dequantization result for the number of N bit levels, and a code word from the adaptive quantizer 111. Adaptive inverse quantizers 114 to 116 are provided for inversely quantizing N+l-M hits to be decoded and Hell's number, and respective locally decoded signals are extracted by the outputs of the adaptive inverse quantizers and the outputs of the adaptive predictor; Each locally decoded signal is divided into subframes, local decoded signals for each subframe are synthesized according to an allocation pattern adapted to the transmission bit rate, and the quality is evaluated by comparing each synthesized locally decoded signal with the input audio signal. , the allocation pattern with the best characteristics is selected, and the codewords from the adaptive quantizer are deleted for each subframe in accordance with the selection, and the codewords are multiplexed and transmitted.

Furthermore, as a device that uses an allocation pattern in the frequency domain, a plurality of encoders 211 to 2 nm are provided for each band, each having a different number of quantization levels for the number of bits existing in the allocation pattern, and extracts local decoded signals from each of the devices, generates one frame worth of local decoded signals for all combinations of the allocation patterns, compares each local decoded signal with the input audio signal to evaluate its quality, and determines the characteristics. According to the selected allocation pattern, the corresponding code words are multiplexed and transmitted.

[Effect]

The first means of the present invention is a method for determining an allocation pattern in the time domain. The input audio signal is processed frame by frame. The input audio signal for one frame is quantized by an Mbit adaptive quantizer 111, inputted to a selective multiplexer 112, and input to each N, N+1 . ...M-1,
In an N-bit dequantizer, necessary upper bits are dequantized by the number of bits of each adaptive dequantization. Mbi
The L adaptive quantizer 111 constitutes an embedded encoder by an N-bit adaptive inverse quantizer 113 and an adaptive predictor 117 that operates based on the N-bit inverse quantization result. An embedded encoder is an encoder composed of an S-bit quantizer, a T (<3)-bit inverse quantizer, and an adaptive predictor. - Encoding method Tearly that makes it possible to carry other information on the T bit, document rCCITT, G722
It is also disclosed in J. Figure 3 shows an example of the quantization characteristics of the embedded encoding (↑).
The 4 bits of 000,0001 are expressed as the top 3 pinpoints of 000, oooo, ooot.

The four bits of ooio and ooti are expressed by the upper two bits of 00.

Further, the output of each of the adaptive inverse quantizers is added to the output of the adaptive predictor 117 which operates based on the N-bit inverse quantization results to generate respective local composite signals, which are input to the local decoded signal evaluation unit l18. At the same time, input signals for corresponding l frames are also input.

The local decoded signal evaluation unit 117 divides each one-frame locally decoded signal into subframes, and determines the amount of information per sample for each subframe in the range of M-N bits according to a plurality of information amount allocation patterns. The locally decoded signals changed in are combined to synthesize one frame's worth of locally decoded signals. At this time, the information amount allocation pattern is such that the total amount of information for one frame does not exceed the number of transmission bits allowed for one frame.

Among the locally decoded signals for l frames synthesized according to a plurality of information allocation patterns, the pattern with the best characteristics is selected by comparing it with the input signal, and the result is output to the selective multiplexing section 112. do. Since the codeword has a hierarchical structure, the selective multiplexing unit 112 deletes the lower bits of the Mbtt codeword for each subframe according to the information allocation pattern, and generates a codeword that corresponds to the pattern. Code words for one frame are multiplexed. Further, the information amount allocation pattern selected by the local composite signal evaluation section 118 is multiplexed and transmitted as auxiliary information.

The second means is a method for determining an allocation pattern in the frequency domain. The input audio signal is processed frame by frame. 1
A frame input audio signal is divided into n bands by a band division filter. Since the band-divided signals are independent, the quality deterioration due to encoding in each 41P band is independent, and therefore the deterioration due to encoding may be evaluated independently for each band. The signals of each band are encoded in parallel by a plurality of encoders with different encoding bit rates corresponding to the encoding bit rates assigned to each band existing in the information allocation pattern. A plurality of encoders with different encoding bit rates each output a code word and a locally decoded signal. The evaluation unit No. 23 evaluates local decoded signals of multiple encoders with different encoding bit rates for each band, combines the evaluation results while referring to the information allocation pattern, and performs a comprehensive evaluation across all bands. The information allocation pattern with the best characteristics is selected and output to the selection/multiplexing section 24. The selection/multiplexer 24 selects and multiplexes code words corresponding to the information allocation pattern for each band, and further multiplexes the information allocation pattern and sends the multiplexed code word to the transmission path. The information allocation pattern is a pattern in which the total amount of information for one frame is less than or equal to the amount of information allocated to one frame. On the decoding side, (1) a decoder is selected for each band according to the information allocation pattern for each frame, and after performing decoding processing, band synthesis is performed and a reproduced signal is output.

In the present invention, evaluation results for each band are combined with reference to information allocation patterns to comprehensively evaluate all bands, so multiple encoding patterns allocate the same amount of information to a certain band. In this case, in that band, the encoding process corresponding to the information 1IJfi will be performed - times, thereby preventing an increase in the amount of processing, and - Since the allocation pattern is calculated from the frame to which information is allocated, Optimal allocation can be made.

〔Example 〕

(First Embodiment) FIG. 4 shows a first embodiment of the time domain allocation pattern determination method according to the present invention.

The transmission bit rate of the transmission path in this embodiment is 4 bits/sample. Further, the number of subframes is assumed to be two. In this case, the possible allocation pattern for boat fishing is that the number of bits allocated to the first half and the second half is (6°2) (5, 3) (4, 4) (3
, 5) (2, 6). Therefore, the adaptive quantizer 401 has 6 bits/sample. Further, the adaptive inverse quantizers 407 to 411 are set to 2 to 6 bits/sample. The adaptive rules of the adaptive predictor 402, adaptive quantizer 401, and inverse quantizers 407 to 411 are described in the above-mentioned document RCC ITT,
Use the tail specified by G722J, 6bit
This method uses the dequantization result of a 2-bit code word with the lower 4 bits deleted, and the adaptive predictor also uses a 2-bit/2-bit code word.
Acts on sample results. Further, each of the 2 to 5 bit/sample inverse quantizers 407 to 410 has lower bit deletion units 403 to 406 for deleting lower bits corresponding to the previous stage thereof.

The length of the l frame is set in consideration of the non-stationarity of the audio signal and the amount of auxiliary information.

In the local decoded signal evaluation section, local decoded signals are cut out for each subframe according to the bit allocation pattern according to the present method, and the locally decoded signals are combined, thereby synthesizing one frame of locally decoded signals. The signal-to-noise ratio (S/N) between the locally decoded signal and the input signal corresponding to each pattern is calculated, and the S
/N Select the maximum pattern.

Based on the result, the selection/multiplexing section deletes or leaves the lower bits unchanged for each subframe to obtain a desired code word. This code word is multiplexed for one frame, and a signal indicating the information amount allocation pattern is also multiplexed and sent to the transmission path.

Hereinafter, the control operation will be explained in more detail with reference to FIG. 5, which shows a flowchart of the embodiment.

Step l) One frame of the input audio signal is input. (
Sl) Step 2) Adaptive predictor 40 based on the previous input audio signal
The error signal is extracted using the predicted signal that is the output of step 2. (32
) Step 3) Convert the error signal to a 6-bit adaptive quantizer 4.
01 quantizes one sample with 6 bits. (S3) Step 4] Each lower bit deletion unit 403 deletes the lower 1 to 4 bits from the quantized 6-bit/sample signal, and the corresponding adaptive quantizers 407 to 411 dequantize the signal. Each local decoded signal a to e is extracted from the sum with the prediction signal obtained by the bit adaptive predictor 402 and input to the quality evaluation section 412 . (S4) Step 5
) adaptive predictor 402. The parameters of the adaptive quantizer 401 and the inverse quantizers 407 to 411 are updated using the locally decoded signal from the 2-bit/sample adaptive inverse quantizer 402.

Step 6) The above operation is continued until processing of the sample deposit for one frame is completed. (S6) Step 7) Each of the locally decoded signals a''-e is divided into subframes. In this embodiment, it is divided into two.

(S7) Step 8) Locally decoded signals for l frames corresponding to the information amount allocation pattern are synthesized from the locally decoded signals for each subframe. In other words, the first half subframe of signal a and signal e
When combining with the second half of the subframe, the number of bits becomes (2, 6) for each subframe, and conversely, when combining
) is obtained, and the average number of bits of l frame in this case ~/
As for the number of samples, a 4-bit signal can be obtained.

Similarly, the bit numbers (3, 5) and (5, 3) are obtained for each subframe from the signals S and d, and (4, 4) from the signal C. (S8) Step 9) The average bit/sample number of all of these five patterns of locally decoded signals is 4 bits.

The five signals obtained in this way are respectively input audio signals f
The S/N ratio is calculated for each pattern, the pattern with the best characteristics is selected, and a signal g indicating the selected pattern is output to the lower bit deletion/multiplexing section 413. (S9) Step 10) The lower bit deletion/multiplexing unit 413 deletes the lower several bits of the input code word for each l sample for each subftherm according to the signal g. For example, if the selected pattern is (6°2), the first half of the subframe is left as is, and the lower four bits of the second half of the subframe are deleted. (SIO) Step 11) Multiplex the codewords whose lower bits have been controlled by the above operation. (Sll) Step 12) The multiplexed code word and the output signal g of the quality evaluation unit 412 are multiplexed and transmitted. (S12) The above is the control operation on the code side.

Next, FIG. 6 shows an embodiment of a decoding example corresponding to the code side. On the decoding side, inverse quantizers of 2 to 6 bit inverse quantization levels are provided in parallel corresponding to the code side. The transmitted multiplexed signal is separated by a separating section 601 into a signal indicating the allocation pattern selected on the code side and each code word. A signal indicating the allocation pattern is input to the switching unit 602 to control which inverse quantizer is input for each subframe, and a switching code word is input to any inverse quantizer for each subframe. It is something.

The signal thus obtained is combined with the predicted signal of the predictor 608 and output as a reproduced audio signal.

(Second Embodiment) As in the present invention, embedded coding is used to create locally decoded signals in each allocation pattern, and bit allocation is performed by quality evaluation using adaptive bits in both the frequency domain and time domain. It is also applicable to coding that performs allocation, and a second embodiment thereof is shown in FIG. In this case, the subframe is divided into each frequency band and further divided into the time domain.
The information amount allocation pattern is a two-dimensional pattern in the time domain and frequency domain. For example, the first half of the high-frequency subframe,
−The second half, the first half and the second half of the low-frequency side subframe,
The sum of the information amounts of all subframes is the information amount of one frame. In this embodiment, one frame is divided into four. The allocation patterns that are normally considered in this embodiment are shown below.

time time time time time time low range (54”
54) ゝ45'' 45ノ (44) However, in this embodiment, a pattern is adopted in which a larger number of bits is allocated to the low range than the high range.

An input audio signal is divided into a high band and a low band through a band division filter 701, and each corresponding time domain low band encoding unit 702 has a quantizer with a number of 2 to 6 bit dequantization levels. , although the inside of the high-frequency encoding section 703 is not shown,
It has substantially the same configuration as the low-frequency encoding section 702, except that it has dequantizers with dequantization levels of 2 to 4 levels. The quantization and inverse quantization here are the same as those described in the previous embodiment. The quality evaluation section 704 combines the locally decoded signals from each inverse quantizer for each subframe in the time domain into a locally decoded signal for each band according to the above allocation pattern. Furthermore, band synthesis is performed to generate a locally decoded signal synthesized from the four-divided signals of one frame, which is compared with the input audio signal, calculates the S/N, selects the pattern with the best characteristics, and uses the selected signal to perform the above-mentioned implementation. As in the example, this is processed by the lower bit deletion/multiplexing unit.

As in the first embodiment, the decoding section dequantizes and corrects the band according to the signals indicating the allocation pattern that are simultaneously transmitted to obtain a reproduced signal.

When allocation is performed in the time domain and frequency domain in this way, it becomes possible to increase the number of allocation patterns, and more detailed information amount allocation becomes possible.

In addition, it is possible to easily perform encoding with better characteristics than conventional simple encoding, and by using several necessary inverse quantizers instead of encoding all allocation patterns as in the conventional method. This is an excellent method that can be used in many situations.

(Third Embodiment) Next, a third embodiment of a system for adaptive bit allocation in the frequency domain is shown in FIG. 801 is a first band division filter;
802 and 803 are second band division filters that further divide the output of the first band division filter. This is a third band division filter. In the third embodiment, the band is divided into four. FIG. 9 shows an example of a focus assignment pattern that can be considered in that case. In this embodiment, control is performed according to the pattern. The transmission bit rate of the transmission path here is approximately 3 bits/sample.

The band division filters 801 to 803 include QMF (Qua
drature m1rror filter). ADPCM is used for the encoder, and the allocation pattern is a pattern that takes into account the characteristics of the audio signal and allocates a large amount of information to a low frequency band. (Figure 9 above) 9th
As shown in the figure, ADPCM encoders having the number of quantization bits existing in the allocation pattern for each band are arranged in parallel. Therefore, the number of quantization bits of ADPCM for each band is 6
．． 5.4 bit encoder 804~806.5+ 4+
These are 3-bit encoders 807 to 809.3, 2-bit encoders 810, 811.3, and 2-bit encoders 812, 813.

The present embodiment will be described in detail below with reference to FIG. 10 showing a flowchart of the third embodiment.

Step 1) One frame of audio signal is input.

(T 1 ) Step 2) 4 through band division filters 801 to 803
Divide into bands. (T2) Step 3) All ADPCM encoders 804 to 813
In each case, encoding processing for -frames is performed. (T3
) Step 4) Each of the encoders 804 to 813 has an internal decoding function, and each outputs one frame's worth of locally decoded signals to the quality evaluation section 814 via the decoding function. (T4) Step 5) For each band, each ADP
The locally decoded CM signal and the corresponding input audio signal of each band are compared, each error signal is extracted, and the sum of squares (l
frame). (T5) Step 6) Perceptually weighting is performed on the sum of squares corresponding to each ADPCM. (
T6) Step 7) When each locally decoded signal is synthesized according to the allocation pattern shown in FIG. 9, calculate the sum of the sum of squares obtained in T6 above for each band with auditory weighting for each allocation pattern. do.

(T7) Step 8) The one with the smallest sum determined in T7 is selected as the allocation pattern with the best characteristics, and a signal indicating the allocation pattern is output to the selection/multiplexing section 815. (T8
) Step 9) Each ADP input to the selection/multiplexing section 815
Among the codewords from the CMs, the codewords from the ADP CM of each band are selected according to the signal from the quality evaluation section 814 and multiplexed with the signal indicating the allocation pattern. (T9) Step 10) Output code word data for l frames to the transmission path. (TIO) Step 11) Copy the internal parameters of the ADT'CM selected for each band to other ADPCMs within each band.

This is because the amount of information allocated to each band differs for each frame, so in order to synchronize with the decoder, at the beginning of the frame for each band, the coder internally uses the previously selected quantization bit number. The parameters are copied to the ADPCMs existing in parallel in the corresponding band. (Tll) The above is the control on the code side.

Next, FIG. 11 shows the configuration of the decoding side corresponding to the code side. 1101 is a separation unit, 1102 to 1105 are switching units for each band, 1106 to 1115 are ADP CM decoders each having a different number of dequantization levels, and 1116 is a band synthesis filter.

Separation section 1101 separates the signal indicating the allocation pattern selected on the code side and each code word. Each code word is input to switching units 1102 to 1105 for assigned bands. In each switching unit, the input code word is switched and input to the ADPCM decoder corresponding to each band according to the signal indicating the allocation pattern selected on the code side separated by the separation unit 1101. The output is filtered by a band synthesis filter 11 for each band.
16, which is synthesized and output as a reproduced audio signal.

In the conventional case where all allocation patterns are encoded and decoded and the optimal one is selected, in the case of this embodiment, there are 8 allocation patterns, so the ADPCM However, according to the present invention, since there are two bands in which the data is tripled and two bands in which the data is doubled, the processing is 2.5 times as much on average. but,
Compared to a method in which all patterns are encoded, it is possible to transmit high-quality signals with extremely simple processing.

(Fourth Embodiment) In the present invention, a large number of This can be achieved with simpler processing without using ADPCM. Example 4 is shown in FIG. In the figure, the same members as in FIG. 8 are given the same reference numerals.

Further, the ADPCM 4 to 6 binot encoders 804 to 806 have an embedded encoder, the ADPCM encoder 90 has an embedded encoder, 807 to 809 have 91, and 810.811 has 9.
2, 812, 813 is 93 that performs the function,
It is. ADPCM encoding with embedded encoder! The internal configuration of S90 is shown in FIG. t3. ADPCM encoders 91 to 93 having other embedded encoders perform quantization.

The configurations are the same except for the number of bits for inverse quantization. FIG. 13 has substantially the same configuration as the first embodiment shown in FIG. The difference is that the output of each lower bit deletion section is output to an adaptive inverse quantization section and a selection/multiplexing section. With this configuration, similarly to the third embodiment, a locally decoded signal with the number of bits according to the allocation pattern in each band can be obtained. The processing in the quality evaluation section is exactly the same as in the third embodiment.

Using embedded ADPCM makes it possible to share the same adaptive update period for adaptive prediction and adaptive quantization in ADPCM, so even when multiple ADPCMs are operated in parallel, the processing of multiple dequantizations increases. Just do it. In this case, even if the amount of information allocated to each band differs from frame to frame, the processes of adaptive prediction and adaptive quantization are common, so there is no need to copy the internal parameters at the beginning of the frame, and this is further improved than in the third embodiment. This makes processing easier.

Although embodiments in the time domain, frequency domain, and both have been described above, various methods of evaluation by each quality evaluation section can be considered. For example, the cepnutrum distance, the peak value of the absolute value of the error signal, the sum of the absolute values of the error signal, the sum of squares of the error signal, a combination thereof, etc. can be considered, and the present invention is not limited to the embodiments.

(Effects of the Invention) According to the present invention, AD using adaptive prediction and adaptive quantization
It becomes possible to combine PCM with time-domain adaptive bit allocation processing with good characteristics, and a highly efficient speech coding and transmission device that efficiently utilizes temporal fluctuations in speech signals can be realized. In addition, in an audio encoding method that divides an audio signal into multiple bands and changes the amount of information allocated to each band according to temporal fluctuations in the characteristics of the audio signal for each frame,
Since the optimum one can be selected with a small increase in the amount of processing, it becomes possible to efficiently utilize the non-stationarity of the audio signal in the frequency domain, and a highly efficient encoding and transmission device for the audio signal can be configured.

Therefore, according to the present invention, it is possible to provide a speech encoding/decoding transmission system that has higher quality and requires much simpler processing than processing all patterns.

[Brief explanation of drawings]

Figure 1 is a diagram of the principle of a speech coding transmission method using an allocation pattern in the time domain, Figure 2 is a diagram of the principle of a speech coding transmission method using an allocation pattern in the frequency domain, and Figure 3 is a diagram of the principle of an audio coding transmission method using an allocation pattern in the frequency domain. 4 shows an example of the audio coding transmission method using an allocation pattern in the time domain, FIG. 5 shows a flowchart of the audio coding transmission method shown in FIG. 4, and FIG. FIG. 4 is an example of an audio encoding method that performs decoding corresponding to the encoding; FIG. 7 is an example of an audio encoding transmission method that uses allocation patterns in the frequency domain and time domain; The figure shows an example of an audio coding transmission method using an allocation pattern in the frequency domain, FIG. 9 shows an allocation pattern used in the audio coding transmission method of FIG. 8, and FIG. Figure 11 is an example of a voice encoding method that performs decoding corresponding to the encoding in Figure 8. Figure 12 is an example of a voice encoding transmission method that performs decoding corresponding to the encoding in Figure 8. An example in which an embedded encoder is used as an encoder, FIG. 13 shows the internal configuration of the ADPCM encoder using embedded encoding in FIG. 12, and FIG. 14 shows the conventional configuration. FIG. In the figure 11 112.24 113 to 116 17 118.23 Adaptive quantizer selection/multiplexing unit Adaptive inverse quantizer Adaptive predictor Quality evaluation unit Quantization Code Word (Quantization Code) 2bit BbIt 4b Quantization Code Diagram 4th Flowchart, -G Diagram Assignment Pattern.

Claims (11)

[Claims] (1) A method that compresses the amount of information of an input audio signal, in which one frame is divided into several subframes, and each subframe is given a quantum amount per sample to achieve a predetermined transmission bit rate. In an audio coding transmission method that allocates and encodes the number of quantization bits, a locally decoded signal is generated from codewords with different numbers of quantization levels extracted for each frame, and the above-mentioned decoded signal is generated according to an allocation pattern that achieves a predetermined transmission bit rate. Speech encoding characterized in that locally decoded signals of each subframe within a frame are combined to generate a locally decoded signal of one frame, and code words corresponding to an allocation pattern with the best characteristics are multiplexed and transmitted. method. (2) A method for compressing the amount of information of an input audio signal, especially ADPCM (AdpCM) that uses adaptive prediction and adaptive quantization.
veDifferentialPulseCodeMo
In an audio encoding method that combines time-domain adaptive bit allocation that adaptively allocates bits in the time domain according to temporal changes in the input audio signal, adaptive quantization with M-bit quantization levels is used. Converter (111) and N
Adaptive inverse quantizer (<M) number of bit inverse quantization levels (1
13) and an adaptive predictor (117) that operates based on the result of inverse quantization of the number of N bit levels;
Adaptive inverse quantizer with +1 to M bit inverse quantization levels (1
14 to 116) are provided, each locally decoded signal is extracted by the output of the adaptive inverse quantizer and the output of the adaptive predictor, each of the locally decoded signals is divided into subframes, and the locally decoded signal is adapted to the transmission bit rate. synthesize local decoded signals for each subframe according to the selected allocation pattern, compare each local decoded signal with the input audio signal to evaluate its quality, select the allocation pattern with the best characteristics, and perform adaptive quantization according to the selection. A voice coding transmission method that deletes the lower bits of the code word from the device for each subftherm, multiplexes it, and transmits it. (3) The audio coding transmission system according to claim (2), wherein when multiplexing code words, a signal indicating a selected allocation pattern is also multiplexed and transmitted. (4) A method that compresses the amount of information of an input audio signal, in particular ADPCM encoding using adaptive prediction and adaptive quantization, and adaptively allocating the number of bits in the time domain according to temporal changes in the input audio signal. In an audio decoding method that uses inverse quantization to decode a signal encoded by audio encoding that combines time-domain adaptive bit allocation, codewords are generated according to a signal indicating an allocation pattern transmitted from the encoder side. A switching unit (602) is provided to switch the input inverse quantizer for each subframe, and each subframe is input to the inverse quantizer with the number of inverse quantization levels corresponding to the number of levels at the time of quantization and reproduced. An audio decoding method characterized by obtaining audio signals. (5) A voice encoding/decoding transmission device comprising an encoding device based on the voice encoding transmission method according to claim (3) and a decoding device based on the voice decoding method according to claim (4). (6) A method for compressing the amount of information of an input audio signal, in particular, dividing the input audio signal into multiple frequency bands for each frame,
In an adaptive bit allocation coding method that changes the allocation pattern of the number of bits to be allocated according to temporal changes in each band, the number of bits existing in the allocation pattern has a different number of quantization levels for each band. Multiple encoders (2
11 to 2 nm), extracts each local decoded signal from the encoder, generates one frame worth of local decoded signals for all combinations of the allocation patterns, and combines each local decoded signal with the input audio signal. 1. A speech coding and transmission system characterized in that the quality of the signals is compared, the quality thereof is evaluated, the best combination of characteristics is selected, and the corresponding codewords are multiplexed and transmitted according to the selected allocation pattern. (7) The audio coding transmission system according to claim (6), wherein when multiplexing code words, a signal indicating a selected allocation pattern is also multiplexed and transmitted. (8) A method for compressing the amount of information of an input audio signal, in particular, dividing the input audio signal into multiple frequency bands for each frame,
In an audio decoding method that decodes signals that are encoded and transmitted using an adaptive bit allocation encoding method that changes the allocation pattern of the number of bits allocated according to temporal changes in each band. A code word and a signal indicating a bit allocation pattern are separated, and decoded by a dequantizer having a corresponding number of dequantization levels for each frequency domain of the code word according to the signal indicating the bit allocation pattern, and decoding is performed for each frequency domain. An audio decoding method characterized in that a reproduced audio signal is obtained by synthesizing encoded signals. (9) A voice encoding/decoding transmission device comprising an encoding device based on the voice encoding transmission method according to claim (7) and a decoding device based on the voice decoding method according to claim (8). (10) Multiple encoders corresponding to frequency bands are based on an adaptive quantizer with M-bit quantization levels, an inverse quantizer with N-bit inverse quantization levels, and an inverse quantization result with N-bit levels. Claim (7), characterized in that the embedded encoder is composed of an adaptive predictor that operates according to Audio coding transmission method. (11) A method that compresses the amount of information in an input audio signal. It divides the input audio signal into multiple frequency bands, changes the allocation pattern of the number of bits allocated according to the temporal changes in each band, and also uses adaptive prediction. In an audio coding transmission system that combines ADPCM encoding using adaptive quantization and time domain adaptive bit allocation that adaptively allocates the number of bits in the time domain according to temporal changes in the input audio signal, each frequency For each band, an adaptive quantizer with M bit quantization levels, an adaptive dequantizer with N (<M) bit dequantization levels, and an adaptive predictor that operates based on the dequantization results with N bit levels. and an adaptive inverse quantizer with a number of N+1 to M bit inverse quantization levels for decoding the code word from the adaptive quantizer, the output of the adaptive inverse quantizer and the adaptive prediction are provided. extracts each locally decoded signal from the output from the transmitter, divides each locally decoded signal into subframes, synthesizes the locally decoded signal for each band according to an allocation pattern adapted to the transmission bit rate, and further divides the locally decoded signal into subframes according to the allocation pattern adapted to the transmission bit rate. synthesize local decoded signals for one frame according to A voice coding transmission method that deletes the lower bits of the code word for each subftherm, multiplexes it, and transmits it.