JPH07106978A - Digital speech signal coding/decoding method for transfer of data - Google Patents

Abstract

PURPOSE:To produce the signals by compressing and coding these speech signals in a real-time processable range and decoding them at the receiving side based on the compressed and coded signal when the digital speech signals are transmitted through a prescribed transmission line, e.g. telephone circuit. CONSTITUTION:The digital speech input signal 4 supplied at a prescribed sampling speed is transmitted through a narrow band filter 10, and the signal component SF (F, T) is calculated together with the time base maximum TMAX (F) and the frequency axis maximum value FMAX (T) of the signal component against a frequency band F divided into M pieces and continuous N pieces of processing time T. Then the component SF (F, T) is normalized. The normalized speech signal NS (F, T) is compared with the audible voice threshold value TH (F) which is previously stored. If the speed signals of all time value are less than the value TH (F) within the same frequency band, these speech signals are not transmitted. The quantity of data on the speech signals to be transmitted can be extremely reduced to the quantity of data not to the transmitted. Thus the speech signals are coded. This procedure is reversed when the speech signals are reproduced.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for encoding and decoding a digital voice signal in data transmission, and particularly to I
The present invention relates to a method for compressing and encoding data of a digital audio signal in SDN in real time and a method for decoding the encoded signal.

[0002]

2. Description of the Related Art Coding of digital audio signals and digital recording / transmission of audio signals related thereto have already been put into practical use. For example, regarding recording with a digital compact cassette, Takefumi Fujimoto, "Key points of Philips DCC system: Psycho-Acoustic PA
Characteristics and details of SC code ", AIA Publishing Co., Ltd., Radio Technical Journal, December 1991, pp. 156-161. Here, high-efficiency speech signal coding (PA
SC: Precision Adaptive Subb
and Coding) is used.

In this encoding method, a voice signal is first introduced into a bandpass filter, and this signal is divided into, for example, 32 equally spaced bands. In DCC systems, a bandwidth of 750 Hz is typically used because the sampling frequency is 48 kHz. And using 512 input data,
Obtaining these data in 32 bands, taking into account the frequency dependence of human audible voice signal level and voice sensitivity,
It quantizes and encodes audio signals.

As is well known, there is a significant frequency dependence with respect to the detection of voice signals. That is, an acoustic signal (sound pressure) having a frequency near 0 Hz and a frequency of about 15 kHz or higher cannot be detected by the human ear. Especially, the detection sensitivity of the acoustic signal is high at 2 to 5 kHz, and paying attention to this point, the PASC hardly deteriorates the reception quality of the audio and the encoding of the audio signal is made efficient to record the high quality audio signal. It is possible.

On the other hand, the Electric Power Corporation (NTT) has already made 19
The advanced information and communication system (INS: Information Ne) as the ISDN plan proposed in 1986.
In the “work system”, a data transmission rate of 64 kbps is used. Such a transmission speed naturally imposes a greater restriction on the transmission amount of the digital audio signal than the DCC system, that is, a limitation on the transmission amount of data per unit time.

When a coding method such as PASC is used in such a system having a limited data transmission amount, it is possible that a small signal may become zero due to coarse quantization. . Therefore, when the sound is interrupted or a small signal continuously flows, if a pulse-like additional signal is input, the small actual voice signal disappears.

[0007]

DISCLOSURE OF THE INVENTION Problems to be Solved by the Invention
Even if there are some restrictions on the amount of data transmission such as SDN,
It is possible to efficiently compress and encode compression-encoded voice signal data so that a high-quality voice signal can be retained and nevertheless reproduced in real time, and at the same time, the compression-encoded voice signal can be decoded. To provide a way to do.

[0008]

SUMMARY OF THE INVENTION According to the present invention, the above object is to provide the following processing steps in transmission of a digital audio signal, a) input a digital audio signal of a predetermined sampling frequency to a narrow band multiplex digital filter, The signal component S (F, T) of each frequency band F is obtained by dividing into frequency bands and the sequential time T of a fixed time interval, and b) the absolute value of the signal component S (F, T) of each frequency band F is calculated. The maximum value TMAX (F) on the time axis, which is the maximum value, is obtained, and c) the signal component S within each time T.
Frequency axis maximum value FMAX which is the maximum value of (F, T)
(T) is calculated, and d) the signal component S (F, T) is divided by the smaller one of the time axis maximum value TMAX (F) and the frequency axis maximum value FMAX (T) to obtain the normal signal component NS (F, T). And e) time axis maximum value TMAX (F) in each frequency band F
Is compared with the audible voice threshold TH (F), and the converted time axis maximum value CTMAX (F) is set to 0 when the former is smaller than or equal to the latter, and when the former is larger than the latter, the time is Use the value of axis maximum value TMAX (F),
f) When the converted time axis maximum value CTMAX (F) is 0,
The subsequent transmission of the signal at all times T within the frequency band F is prohibited, and g) the normal signal component NS (F, T) is quantized by at least one code having a smaller number of bits.
The frame sync signal SYNC as the data to be converted into S (F, T) and h) to be transmitted, the converted time axis maximum value CTMAX
(F), frequency axis maximum value FMAX (T) and error code CRC are sequentially arranged to form a frame information code, and all quantized signals QS on the same time axis of each frequency band F are formed.
(F, T) is further added as unique audio signal data, which is solved by a compression encoding method of a digital audio signal.

Further, according to the present invention, there is provided the following processing step i) for receiving and decoding the compression coded data signal obtained by the above processing step in the transmission of the digital audio signal. From the frame synchronization signal SYNC in the received data block, the converted time axis maximum value CTMAX (F) and frequency axis maximum value FMAX are converted by the following data.
Decode (T), j) Error code C in data block
Checking the data for errors from the RC, taking out the error-free data signal, k) taking the unique data attached to the voice signal from the data block and decoding it to form the denormalized voice signal TNS (F, T), l) Conversion time axis maximum value C
The inverse normalized voice signal TNS (F, T) is multiplied by the smaller value of TMAX (F) and the frequency axis maximum value FMAX (T) to obtain the inverse voice signal (F, T), and the inverse m) is narrowed. A band digital inverse filter is used to obtain a digital output audio signal of a desired sampling frequency from the inverse audio signal TS (F, T).
Has been solved by a method of decoding a compressed audio signal, which performs

Further advantageous configurations according to the invention are described in the dependent claims.

[0011]

According to the present invention, the extremely narrow band multiplex digital filter proposed by the present inventor is used (patent pending). Since this filter is capable of high-speed digital calculation, it is particularly suitable for real-time transmission processing of a digital audio signal (PCM). In the case of one embodiment of the present invention, since the filter has frequency bands divided into 512 between 0 and 15 kHz, the bandwidth of each filter is 29.
Hz. In particular, in the filter used in this invention,
Since the action of each band-splitting filter affects only adjacent bands, a narrow band filter with extremely sharp cut is realized. Therefore, the audio signal can be decomposed with a very high resolution. Further, according to the present invention, a plurality of audio signal components at different times within the same frequency band are processed.
In reality, since the number of frequency bands is large, even if the quantization of the PCM audio signal is considerably roughened, the distortion occurs only in a narrow band, and therefore the quality of the reproduced audio signal is not deteriorated.

As used in the above-mentioned PASC, when the signal level curve of the audible voice threshold and the frequency dependence of the detected sound pressure are taken into account, the signal processing of the inaudible voice signal is omitted. You can This makes it possible to effectively secure the amount of information to be transmitted in the transmission of an audio signal by ISDN, which has a limited amount of transmission. In this way, the amount of signal data characterizing the audio signal can be considerably compressed without degrading the quality of the audible audio signal. In the present invention, the compressed encoded audio signal is decoded and reproduced, but the procedure is the reverse of the encoding method, and a calculation routine similar to that at the time of encoding can be used. .

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Below, based on the preferred embodiments shown in the drawings,
The contents of the present invention will be described in more detail.

As shown in FIG. 1, a digital voice input signal (PCM signal) having a predetermined sampling frequency indicated by symbol 4 is introduced into a very narrow band multiplex filter 10 used in the present invention. With this filter, a signal component in a narrow band obtained by dividing the audible frequency band into M equal parts can be taken out. This frequency division processing is executed N times, and after all, MxN signal components S (F, T); 0 <F ≤ M, 0 <T ≤ N are stored in the buffer 12. Stored signal component S
(F, T) can be represented by a matrix arrangement specified by the index F of the frequency band and the index T of the time axis as illustrated. In this embodiment, the number M of divided bands to be used and the processing time N are M = 512 N = 10. The number of bits used for signal processing is 16 bits or more.

Next, in order to normalize these frequency-divided signal components S (F, T) in the processing step 20, first, the frequency band axis and time axis of the absolute value of the signal component S (F, T) are calculated. The maximum values FMAX (T) and TMAX (F) with respect to each processing time T and each frequency band F are obtained. That is, FMAX (T) = MAX {| S (F, T) |; F = 1 to
M}, T = 1, 2 ... N, TMAX (F) = MAX {| S (F, T) |; T = 1 to
N}, F = 1, 2 ... M.

Next, with respect to the signal component S (F, T) designated by the frequency band F and the time axis T, the maximum value FMAX (T) of the signal components in the frequency band F and the signal component in the time axis T are obtained. Of the maximum value TMAX (F) of
The value obtained by dividing (F, T) is the normalized signal component NS (F, T). That is,

The signal components NS (F, T) thus normalized are calculated for the entire range of the frequency band F and the time axis T, and these are stored in the buffer 22.

Here, preparation is made for data compression in consideration of the audible frequency characteristic of the audio signal. To this end, first, in the bit allocation determining unit 55, a discrete value TH (F) of the frequency characteristic of the audible voice threshold value that is stored in advance in the predetermined location 50 of the processing device and is introduced through the transfer path 74; F = 1, 2, ... M and the maximum value TMAX (F) in each frequency band F introduced by the transfer path indicated by symbol 70 are compared over all frequency bands F. If TMAX (F) is smaller than the discrete value TH (F) of the frequency characteristic of the audible voice threshold value, it is assumed that the voice signal in this band F cannot be heard, which will be described in more detail later. , Is set to 0 as the maximum value CTMAX (F) on the new time axis in order to specify the processing that does not perform subsequent signal information transmission over the entire processing time N of the frequency band F.
Otherwise, the value TMAX (F) as it is is used. In other words, Is converted to. This is shown in the first half (step S10) of the coded bit allocation determination processing routine of FIG.

Furthermore, the latter half of the coded bit allocation determination processing routine of FIG. 2 (steps S11 to S15) shows the preparation processing before compressing the data to be transmitted. Here, as described above, the signals NS (F, T) in the frequency band F when CTMAX (F) = 0 are not all transmitted, so that there is a margin in the amount of data actually transmitted. In addition, in the embodiment according to the present invention, the number of bits is 16
The initial voice signal S (F, T) or NS that is the above
(F, T) is normally compressed to 1.6 bits, but CTM
For a large value of AX (F), the number of bits of all data in the frequency band F is further increased, that is, the signal having a higher resolution is compressed. In this embodiment, CTMAX (F) having a value other than 0 is further classified into three types of groups. That is, the values are classified into large, medium and small, and 4,2.4 and 1.6 bits are assigned to the corresponding data at the time of data compression. As an index showing this bit allocation Stipulate.

Further, according to the embodiment of the present invention, the number of 4-bit compressed data and the number of 2.4-bit compressed data is 1 :.
Allocate to 2. When such a rule is applied, the data transmission rate of the final transmission, for example, 6 in the case of this embodiment.
4 kbps, which is the amount of data that need not be transmitted due to the characteristics of the input audio signal (CTMAX (F) = 0 described above)
By the total number of bits of the signal NS (F, T) in the frequency band F at, ALOC (F) = O, ALOC (F) =
The number of data corresponding to 1 and ALOC (F) = 2 can be calculated.

In step S11, the margin bit number SBIT within one processing time of the buffer 22 determined by the transmission data output speed is SBIT = M × N Total bit number given to the buffer−N
x16 is given. Here, the last term on the right side is the time axis N (=
10) It is assumed that all the pieces of data are incompatible with each other and that they are all 1.6 bits. CTMAX (F) = 0 in SBIT
Non-transmitted data 16 × P (P is step S10
The number obtained by adding CTMAX (F) = 0 in (1) is the total number of remaining bits that can be used when all the data of CTMAX (F) ≠ 0 is processed with 1.6 bits. . In step S11, the remaining total number of bits is divided by 40 to obtain the integer quotient Q and the remainder R (where 40 is the 1: 2 allocation of 4-bit data and 2.4-bit data described above and the number of data in the band). = 10). As a result, in step S12, k ₂₄ is obtained as the number of bands F to which 2.4 bits are allocated, and k ₄₀ is obtained as the number of bands F to which 4 bits are allocated. Next, in steps S13, S14 and S15, the index A that finally indicates the bit allocation
All LOC (F) are designated.

After the coded bit allocation determination process of FIG. 2, the signal NS (F, T) normalized by the first quantization processing unit 30 (FIG. 1) is quantized to the above-mentioned coarse bit number. Be converted. This is performed according to the procedure shown in FIG. The index ALOC (F) that directs the bit allocation of each frequency band F introduced from the bit allocation determination processing unit 55 via the transfer path 76 is determined in step S20, and a coefficient is determined according to the value of the index ALOC (F). The value of PPX is designated, and the signal NS (F, T) in the buffer 22 is requantized in step S21. Here, Int (X) means the maximum integer value that does not exceed X, and >> means right shift by 1 bit and divide by 2 to discard the remainder. The final processing is performed to make the positive and negative bit distribution of the signal symmetrical with respect to the 0 signal level (see FIG. 4). The signal QS thus quantized
(F, T) is stored in the buffer 32.

In the second quantization processing unit 60 (FIG. 1), the maximum value CTMAX (F) in each frequency band and the maximum value FMAX (T) in each time axis are transferred via the transfer paths 78 and 72, respectively. Introduced, where 6-bit logarithmic quantization (2
dB step). Let these logarithmically quantized values be QTMAX (F) and QFMAX (T), respectively.

Finally, the quantized signal QS (F,
In order to output T), QTMAX (F) and QFMAX (T) to the output signal 8 sent to the output transmission line, these quantized signals are coded by the data coding processing unit 40 (FIG. 1), and at each processing time. The frame signal is made into blocks (frame information signal and compressed audio signal).

As shown in FIG. 5, first, in step S30, a synchronization signal SYNC (8 bits) is added to the beginning of the frame. Then, QTMAX (F) is sent out in step S31, and QFMAX (T) is sent out in step S32.
Further, the cyclic code CRC (7 bits) is added to complete the code of the frame information.

Next, the compressed audio signal is added to the free format section following the frame information by the data encoding process shown in FIG. In this case, the data encoding processing unit 40 informs the quantized audio signal QS (F,
T) in addition to the bit allocation index ALO via the transfer path 77.
C (F) is also introduced. This index ALOC (F) is first determined in step S40. In the case of 1.6 bits (when ALOC (0)), in step S41, the data of 1 to 5 on the time axis is converted into one ternary display value SD.
As 0, the data up to 6 to 10 on the time axis is set as another ternary display value SD1 and set to 8 in step S44.
Send it out in bits. In the case of 2.4 bits (A
At the time of LOC (1)), in step S42, as shown in the figure, the data of 1 to 5 on the time axis is converted into one quinary display value SD.
The data up to 6 to 10 on the time axis is set as another quinary display value SD0 and set to 1 in step S45.
Send in 2 bits. Alternatively, in the case of 4 bits (when ALOC (2)), 7 is added to the data at each time within the corresponding frequency band F in step S43 to obtain 4
Send it out in bits.

In the above-described signal coding processing, the calculation over a large number of bands, particularly the maximum value FM on the frequency axis is performed.
AX (T) was examined in the range of M = 512. However, since voice has audible characteristics that vary greatly with frequency,
If this maximum value is obtained within the frequency range that is further divided,
The quality of the transmitted voice can be expressed more faithfully. When using M = 512 in this example, eight divisions gave the best experimentally reproduced speech quality. Such division processing needs to be executed for all the processing of the normalization processing unit 20, the quantization processing unit 30, and the data encoding processing unit 40.

M = 512, N created in this way
= 10 and the number of frequency band divisions = 8 shows compressed data in Fig. 7a. Here, CTMAX (F) is F = 3
Since it is 0, the frequency band data DT (F)
Note that DT (3) is not sent. In addition, since the frequency band is divided into eight, the maximum value on the frequency axis is expressed as F (FB, T). In this case, the division indices FB = 1, 2, ... Further, each data DT (F) in the free format section is formed by grouping all the data on the same time axis. The data arrangement of 1.6, 2, 4 and 4 bits in this data DT (F) is shown in FIG. 7b.

Next, the digital audio signal having the frame signal arrangement shown in FIG. 7 introduced through a certain transmission circuit or detected by some digital signal reader is decoded to obtain the original audio signal. The processing method for converting to will be described.

FIG. 8 shows a digital audio signal 9 obtained by the compression coding method described above with reference to FIGS.
Are introduced into the decoding unit 41 for the frame information on the receiving device side. This receiving device may be an ISDN-compatible modem, or may be one in which the detection signal obtained from the reading head of the digital compact cassette is waveform-shaped and shaped into a digital audio signal having the frame signal arrangement shown in FIG. Good. This received signal is returned to TS (F, T) to a value close to the original voice signal, and the narrow band multiplex filter 1
The inverse filter is performed by 1 to finally obtain the digital audio signal 5. The details will be described below.

In the decoding unit 41, first, the maximum value QT on the time axis is displayed.
MAX (F) and frequency axis maximum value QFMAX (T) (Fig. 9
Step S60) of FIG. 9, and as shown in step S61 of FIG. 9, the cyclic code CRC is detected, and the received frame synchronization signal TMAX (F), FMAX (T) is used for CR.
A C code is created and used as a CRCR, and it is determined whether the transmitted CRC code matches the CRCR. If they match, the process directly proceeds to the next step. If they do not match, the calculation using the data of the frame is not substantially performed. Thus, the logarithmically quantized reception time axis maximum value QTMAX (F) and frequency axis maximum value QFMAX (T) are respectively dequantized through the transfer paths 86 and 85.
, And both values are subjected to inverse logarithm (exponential function) conversion to determine the time-axis maximum value CTMAX (F) and the frequency-axis maximum value FMAX (T).

Next, in the bit allocation decision processing unit 56, the frequency band F of the data section which does not require the transmission of the voice data.
(The band when TMAX (F) = 0) and the total number P of such frequency bands are obtained in step S50 of FIG.
Then, from this total number P, the remaining number of bits used for transmission in the bit allocation determination processing unit 56 is calculated based on exactly the same calculation method as that of the encoding described in FIG. 2, that is, steps S51, S52, It is determined by S53, S54 and S55. In this case, as described in the compression encoding, the same bit allocation, that is, the data of the frequency band of the normal signal strength is set to 1.6 bits, and the data of the frequency band F of the stronger signal strength and the data of the stronger frequency band F of the signal strength are set. To 1:
Allocate to 2 bits at a ratio of 2 and 2.4 bits. Determine the index ALOC (F) that indicates the bit allocation for their corresponding groups and the number of frequency bands for each bit.

Next, in the data decoding processing section 43, as shown in FIG.
1, the compressed coded audio signal quantized from the data DT (F) on the time axis of the data free format section based on the bit allocation instruction index ALOC (F) sent from the bit allocation determination processing unit 56. Calculate QS (FT). This process corresponds to the inverse conversion of the process of FIG. In this case, HDATA (0, J) = 3 ^4-J HDATA (1, J) = 5 ^4-J .

Further, as shown in FIG. 12, from the quantized compression-encoded voice signal component QS (F, T) in the frame, the normalized voice is processed by a process reverse to that of step S12 of FIG. The signal TNS (F, T) is obtained. This is stored in the buffer 23.

Next, in the normalization processing unit 21, the time-axis maximum value CTMAX (F) and the frequency-axis maximum value FMAX obtained by the dequantization processing unit 61 and introduced through the transfer paths 89 and 90.
Using (T), the normalized voice signal TNS (F,
T) is converted into an initial audio signal TS (F, T) and stored in the buffer 13. The conversion in this case is the reverse of the normalization of FIG. That is, Is. The normalized voice signal TNS (F,
T) and the initial audio signal TS (F, T) are the normalized audio signal NS (F, T) and the initial audio signal S shown in FIG.
It is substantially different from (F, T). This is because the quantization processing 30 shown in FIG. 1 converted the strength of the audio signal component with a considerably coarse bit resolution. Since the roughness of this signal level is determined in consideration of human audibility, it can still be heard as high quality voice when actually converted into voice.

Finally, the block of the matrix-shaped voice signal TS (F, T) in the buffer 13 is passed through the narrow band multi-band inverse filter 11 according to the present invention to obtain a digital voice signal (PCM) indicated by symbol 5. Can be taken out as. This voice signal can be stored in a storage medium in a predetermined storage device, or can be heard as voice with the help of a predetermined voice conversion device (playback device).

When the stereo left and right audio signals are respectively sent to the 2B channel by ISDN, the skew of the 2B channel is not compensated. Therefore, the skew is shifted in advance by shifting the sending time point by about a half cycle of the frame. Absorb.

The decoding method of the compressed digital audio signal according to the present invention has been described mainly for the case of ISDN. However, this invention is not limited to ISDN,
It can also be used for playback on digital compact cassettes and magnetic tapes. In these cases, since the data amount per unit time has a margin more than in the case of ISDN, the bit coding is further increased, and a signal coding / compression method capable of maintaining high sound quality by fine steps and a decoding method therefor are also provided. It is possible.

Furthermore, the present invention is not limited to the values of the parameters M = 512, N = 10, FB = 8 and the type of ALOC (F) = 3 used in the above embodiment, but 2.4. The ratio of bits to 4 bits is not limited to 2: 1. These parameters can be
It can be appropriately selected and changed. Two of the three bits are the CTMAX (F) of each frequency band as described above.
CT is not decided in the order of size (that is, relative)
It can also be evaluated by the absolute value of the value of MAX (F). As is well known, the characteristic curve TH (FB) of the audible voice threshold at the divided frequency FB shows a characteristic that is considerably different depending on the frequency range. In that case, the absolute value can be determined by appropriately weighting the characteristic curve TH (FB) of the audible voice threshold at the division frequency FB. Such an operation should be decided experimentally.

[0040]

As described above, according to the present invention, a method of compressing and encoding digital audio data and decoding after receiving it via a transmission circuit is as follows.
A voice signal can be maintained in sufficiently high quality with a small amount of data transmission. That is, (1) by dividing a signal into a number of bands with a narrow band digital filter, even if a signal in each band is quantized very roughly, quantization distortion occurs only in a narrow band. There is little deterioration in the above sound quality. (2) Since the signal before the coarse quantization is normalized by the smaller one of the maximum values on the time axis and the frequency axis, it is preferable to normalize by only one of the maximum values.
The signal level after normalization rises overall, the effect of coarse quantization is reduced, and the signal level after normalization does not become extremely small even with rapid changes in the time axis and frequency axis. , The voice is no longer lost even with coarse quantization. (3) Since the coded bit allocation instruction value can be reproduced by calculation from the maximum value on the time axis on the receiving side, it is not necessary to transmit. This measure makes it possible to greatly increase the amount of transmitted data.

[Brief description of drawings]

FIG. 1 is a processing diagram schematically showing each step of data compression encoding processing according to the present invention and a signal attached thereto.

FIG. 2 is a flowchart of a subroutine for determining coded bit allocation.

FIG. 3 is a flowchart of a subroutine for performing quantization.

FIG. 4 is a diagram showing a bit arrangement of a signal whose resolution is roughened by quantization.

FIG. 5 is a flowchart of a subroutine for processing data frames and performing data encoding to provide data information.

FIG. 6 is a subroutine for encoding compressed audio data.

FIG. 7 is a layout (a) of data to be transmitted and a layout (b) of a compressed audio signal in one data block.
Indicates.

FIG. 8 is a process diagram schematically showing each step of decoding compression-coded audio data according to the present invention and a signal attached thereto.

FIG. 9 is a schematic flowchart of a routine for receiving a code error check and a frequency axis maximum value during reception.

FIG. 10 is a flowchart of a bit allocation determination routine.

FIG. 11 is a flowchart of a data decoding routine.

FIG. 12 is a flowchart of a routine for calculating a normalized audio signal component.

[Explanation of symbols]

 4 Input Speech Signal (Sampled PCM Signal) 5 Output Speech Signal 8 Compressed Output Speech Coded Signal 9 Compressed Reception Speech Signal 10 Narrow Band Multiplex Filter 11 Narrow Band Demultiplexing Filter 20 Normalization Processing Unit 21 Inverse Normalization processing unit 30 Quantization processing unit I 40 Data encoding processing unit 41 Maximum value decoding unit 43 Data decoding unit 50 Storage unit for numerical values of audible voice threshold curve 55 bit allocation determination unit 56 bit allocation determination unit 60 Quantization Processor II 61 Inverse Quantization Processor

Claims (10)

[Claims] 1. The following processing steps in the transmission of a digital audio signal: a) A digital audio signal of a predetermined sampling frequency is input to a narrow band multiplex digital filter, separated into multiple frequency bands, and a sequential time of a fixed time interval ( Signal component (S (F, T)) of each frequency band (F) in T), and b) signal component (S (F, T)) in each frequency band (F).
Time-axis maximum value (TMAX maximum value of the absolute value of
(F)) is obtained, and c) The frequency axis maximum value (FMAX (T)), which is the maximum value of the signal component (S (F, T)) within each time (T), is obtained, and d) The signal component (S (F, T) is the maximum value on the time axis (TMA
X (F)) and frequency axis maximum value (FMAX (T)) are divided to obtain the normal signal component (NS (F, T)), and e) Time domain maximum value in each frequency band (F) (TMAX
(F)) with an audible voice threshold (TH (F)),
As the converted time axis maximum value (CTMAX (F)), a value of 0 is set when the former is smaller than or equal to the latter, and a time axis maximum value (TMAX is set when the former is larger than the latter.
(F)), f) When the conversion time axis maximum value (CTMAX (F)) is 0, the subsequent transmission of the signal at all times (T) within the frequency band (F) is performed. G) quantizing the normal signal component (NS (F, T)) with at least one code having a smaller number of bits (QS)
(F, T), h) Frame sync signal (SYN) as data to be sent
C), converted time axis maximum value (CTMAX (F)), frequency axis maximum value (FMAX (T)) and error code (CRC)
Are sequentially arranged to form a frame information code, and all the quantized signals (QS (F, F,
T)) is further added as unique audio signal data (DT (F)). 2. A converted time axis maximum value (CTMAX (F))
Is 0, the number of bits in the frequency band (F) with a large time axis maximum value (CTMAX (F)) is taken into consideration, considering the amount of data not used for data transmission and the amount of data transmission specified by the transmission rate. A code having a large value is distributed and converted into a quantized signal (QS (F, T)).
The compression encoding method according to. 3. The frequency axis maximum value (FMAX (T)) and the operations such as normalization and quantization over the frequency band attached to it are divided into a plurality of sub frequency bands (FB). The compression encoding method according to claim 1 or 2. 4. A converted time axis maximum value (CTMAX (F))
4. The compression coding method according to claim 1, wherein the frequency axis maximum value (FMAX (T)) is transmitted as a logarithmically quantized value. 5. The number of frequency bands to be used is 512, the number of time axis data is 10, and the number of band divisions is 8. The converted time axis maximum value (CTMAX (F)) and frequency axis maximum value (FMA).
The logarithmic coding of X (T) is a 6-bit 2 dB step, and the bit number of the code of the compressed data is 1.6,
5. The compression encoding method according to claim 4, wherein there are three types of 2.4 and 4 bits. 6. The following processing steps for receiving and decoding a data signal compressed and encoded according to claim 1 in the transmission of a digital audio signal, i) a frame synchronization signal in a received data block ( S
YNC), the converted time axis maximum value (CTMAX (F)), frequency axis maximum value (FMAX)
(T)) decoding, j) checking the data error from the error code (CRC) in the data block, taking out an error-free data signal, k) taking out the unique data attached to the voice signal from the data block, Denormalized speech signal (TNS) by decoding
(F, T)), and 1) the smaller one of the converted time axis maximum value (CTMAX (F)) and the frequency axis maximum value (FMAX (T)) is used as the inverse normalized audio signal (TNS (F)). , T)) to obtain the inverse audio signal (TS (F, T)), and m) the inverse audio signal (T
A method for decoding a compression-encoded digital audio signal, which comprises: obtaining a digital output audio signal of a desired sampling frequency from S (F, T). 7. The inverse normalized speech signal (TNS (F,
T)) is formed by the converted time axis large value (CTMAX (F))
Is 0, the number of bits in the frequency band (F) with a large time axis maximum value (CTMAX (F)) is taken into consideration, considering the amount of data not used for data transmission and the amount of data transmission specified by the transmission rate. The decoding method according to claim 6, wherein the decoding is performed by allocating a large code. 8. The frequency axis maximum value (FMAX (T)) and the operations such as normalization and quantization over the frequency band attached to it are divided into a plurality of sub frequency bands (FB). The decoding method according to claim 6 or 7. 9. A converted time axis maximum value (CTMAX (F))
And the frequency axis maximum value (FMAX (T)) are the received converted time axis maximum value (CTMAX (F)) and the frequency axis maximum value (FMAX).
The decoding method according to any one of claims 6 to 8, wherein the decoding method is performed by performing inverse logarithmic quantization of MAX (T)). 10. The number of frequency bands to be used is 512, the number of time axis data is 10, the number of band divisions is 8, and the number of bits of the code of the compressed data is 1.6, 2.4 and 4 bits. The decoding method according to claim 1, wherein there are three types.