JPH08279811A - Voice data converter - Google Patents

Abstract

PURPOSE: To attain inter-conversion of code data without modification by interconnecting a VBR-CODEC and a GBR-CODEC. CONSTITUTION: A 2nd voice signal processing unit 106 of an abort cell compensation processing unit 100 receives a fixed bit rate voice (GBR)-CODEC coding data S7 being an output of a cell separator 300 outputs a decoded voice S5. A voiced state abort cell compensation processing unit 105 outputs abort cell compensation data as a signal S4. A white noise generator 104 outputs the white noise as a signal S3. Furthermore, a non-voice state abort cell compensation processing unit 103 outputs cell abort compensation data estimate non-voice data as a signal S2. Furthermore, a 2nd selector 102 outputs the signal S4 in the voice state and outputs the signal S2 in the voice state as a signal S1, and a voice coder 101 outputs coded voice data as a signal S6. Then a voice detector 600 conducts the same processing as the voice detection conducted by a variable bit rate voice (VBR)-CODEC and outputs voice/non-voice information as a signal S21.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice coding / decoding device used in a telephone line switching network using a high-speed digital leased line, and particularly at a constant bit rate regardless of whether the input voice is voiced or not. The present invention relates to an interconnection between a CBR-CODEC for encoding and a VBR-CODEC for assembling and transmitting only voice encoded data of a voiced part of an input voice into a cell having a certain number of bits.

[0002]

2. Description of the Related Art Conventionally, in a network in which an ATM network and an STM network are mutually operated, VBR-CODEC and ST on the ATM network are used.
When interconnecting CBR-CODECs on the M network, it was necessary to temporarily decode all encoded data into PCM voice at the connection point.

In order to explain the prior art, VBR-CO
The DEC encoder is shown in FIG.

As shown in FIG. 2, in the VBR-CODEC encoder, the input PCM audio signal is the audio encoder 70.
2 and the voice detector 701. Voice detector 701
Performs the process of detecting the voiced section of the input voice. The speech coded data encoded by the speech encoder 702 and the voiced section information detected by the speech detector 701 are input to the cell assembler 703. The cell assembler 703 adds encoded data and a header having a certain number of bits such as a transmission time stamp and a cell number, and then makes one cell, and
It is transmitted to the transmission line. This transmission cell is transmitted only when the voice detector determines that there is sound.

The silence compression function means that a cell is transmitted only when there is a sound, and is not transmitted when there is no sound. The VBR-CODEC in the ATM transmission network refers to a voice CODEC in which cells are transmitted only when there is sound and cells are transmitted intermittently.

Next, the VBR-CODEC decoder side will be described with reference to FIG.

In the VBR-CODEC decoder, the speech coded data and the header are separated by the cell separator 803. Cell discard is detected by the cell number or the like. The separated encoded data is input to the audio decoder 802 and decoded into an audio signal.

Only the coded data of the cells which can be received when there is a voice and without discarding the cells are decoded.
When no cell is transmitted from the encoder side when there is no sound, or when the cell is discarded on the transmission line even when there is sound, the voice cannot be decoded as usual, so white noise is decoded. The discarded cell compensation processor 801 executes the discarded cell compensation processing that suppresses the deterioration of voice quality.

On the other hand, on the CBR-CODEC encoder side,
As shown in FIG. 4, a CBR-CODEC speech encoder 9
All the voice coded data coded by 01 are sent out regardless of whether there is sound or no sound, and are transmitted to the CBR-CODEC decoder side.

On the CBR-CODEC decoder side,
As shown in FIG. 5, assuming that all coded data is transmitted from the encoder regardless of whether voice is present or not.
The received coded data is all decoded in the same manner by the CBR-CODEC audio decoder 1001,
Output as M audio signal.

Next, conventional VBR-CODEC and CBR-C
A technique for interconnecting ODECs is shown and described in FIG. As described above, the VBR-CODEC is different from the CBR-CODEC in that the encoder side performs the silence compression process and the decoder side performs the discarded cell compensation process.

Therefore, in order to compensate a silent section or a discarded cell portion for CBR-CODEC and to realize a silent compression function in an ATM transmission network for VBR-CODEC, as shown in FIG. In addition, it was necessary to once decode all the voice coded data into PCM voice.

CBR-CODEC and V in the prior art
The details of the problem when connecting BR-CODEC are described below.

CBR-CODEC and VBR-CODEC
Is directly connected to the PCM voice without decoding at the time of connection, that is, by the CBR-CODEC that does not have means such as discard cell compensation processing by converting the data format of the encoded data of the VBR-CODEC encoder. When decoding, there is a problem that uncomfortable voice is decoded when there is no sound or when cells are discarded due to congestion.

Therefore, in order to convert the voice coded data coded by the VBR-CODEC into the CBR-CODEC coded data, the VBR-CODEC is temporarily changed.
1101 completes the P-coded speech coded data in which the cell discard compensation processing is applied to the cell discard section and the silent section.
Decoded into a CM voice signal, and again CBR-CODEC1
It was encoded at 102. On the contrary, CBR-CODEC
In order to convert the coded data of VBR-CODEC into the coded data of VBR-CODEC, the coded data coded by CBR-CODEC1102 is decoded by CBR-CODEC, and silence is compressed again by VBR-CODEC1101. It was encoded above.

[0016]

According to the prior art,
If VBR-CODEC and CBR-CODEC are interconnected, all signals are always CBR-C at the connection point.
Voice code of ODEC and VBR-CODEC
It is inevitable that the process of decoding and encoding is once performed by C, and there is a problem that the voice quality is deteriorated.

The above-mentioned problem is caused when a mutually interconnected network is constructed by a system in which two systems of an ATM transmission network using VBR-CODEC and an STM transmission network using CBR-CODEC are mixed, that is, VBR- C
There is a problem that the voice quality corresponding to one voice CODEC link always deteriorates at the connection point when the ODEC and CBR-CODEC encoded data are mutually connected.

[0018]

SUMMARY OF THE INVENTION A voice coded data conversion apparatus according to the present invention is a variable bit rate voice CODEC (Variable) obtained by subjecting a telephone band voice signal to a silence compression process.
e Bit Rate Audio CODEC: VBR-C
ODEC) and a fixed bit rate audio CODEC (Cons) that does not apply silence compression to the telephone band audio signal.
tant Bit Rate Voice CODEC: CB
R-CODEC) and the VBR
-CODEC and the CBR-CODEC are audio data converters that enable mutual connection, and the VBR-CO
Asynchronous transfer mode configured by DEC (Asyn
Chronous Transmission Mod
e: An ATM network for transmitting and receiving digital data in the following, referred to as ATM, and a synchronous transfer mode (Synchronouou) configured by the CBR-CODEC.
s Transmission Mode: STM
It is provided between the ATM network and the STM network that transmits and receives digital data, and receives a cell from the ATM network that is composed of voice coded data of a certain number of bits and a header of a certain number of bits. The V
A cell separator having means for separating the BR-CODEC voice coded data, outputting the coded data of the first format, and detecting that the cell has been discarded on the ATM network; A cell discard detection signal received from the cell separator, the cell discard detection signal being received by the cell separator when the cell discard is detected by the device. A first selector for selecting the first-format speech-encoded data output from the cell separator and the second-format speech-encoded data output from the discard cell compensation processor; and the STM. From network CBR-COD
A first voice decoder for receiving the EC encoded data and decoding the voice, and a voiced / unvoiced information of the voice by monitoring the power of the decoded voice output from the first voice decoder. The output voice detector and the CBR-CODEC encoded data received from the STM network are divided into blocks of a certain number of bits, and a header of a certain number of bits is added to the block to form a transmission cell of a certain number of bits, A voice data converter configured by a cell assembler that sends out cells only when there is a voice according to voiced / unvoiced information that is the output of the voice detector, wherein the header with a certain number of bits of the cell is the voice detection. Irrespective of the presence / absence information output of the device, all cells are counted one by one, and the presence / absence information output of the voice detector is used as the presence / absence cell And a second number counted by 1 only for cells transmitted by the packet assembler, the cell separator separates the header portion in cells received from the ATM network, and the second number in the header. Cell number separating means for extracting a cell number of 1 and the second cell number, discarding of cells from the first cell number and the second cell number separated by the cell number separating means, and further silence. Cell information determining means for determining cell discard or voiced cell discard and outputting cell discard information; coded data separating means for separating voice coded data in cells received from the ATM network; The cell information determination means has a speed converting means for converting the encoded data separated by the encoded data separating means into the speech encoded data of the first format. The stage compares the first cell number and the second cell number in the order in which the cells are received, and when the first cell number is discontinuous, detects that the cell is a silent cell discard and outputs the cell. When the second cell number is discontinuous, it has means for detecting that the cell has been discarded due to congestion or the like on the ATM network and outputting the cell, and the voice coded data of the first format is the CBR. A bit string composed only of voice coded data output by a CODEC encoder is transmitted to the VBR-CODEC at a constant speed regardless of whether there is sound or no sound, and network congestion; The discarded cell compensation processor includes a second voice decoding means for decoding the voice coded data of the first format received from the cell separator, and a second voice decoding means for decoding the voice coded data of the first format. A voiced discarded cell compensating means for performing a discarded cell compensation process for a cell discarded when there is a voice based on the encoded voice and the cell discarded information output from the cell separator, and means for generating a white noise signal And a silent cell discard cell compensation processing means for calculating the average amplitude information of the silent section of the decoded speech received from the second speech decoding means and adjusting the average amplitude of the white noise signal output from the white noise signal generating means. According to the cell information received from the cell separator, the discarded cell compensation data which is the output of the voiced cell discard compensating means is selected when the voice is present, and is the output of the silent cell discard compensating means when the voice is silent. The cell assembling means, comprising: second selecting means for selecting the discarded cell compensation data, and voice coding means for encoding the discarded cell compensation data which is the output of the second selecting means. The cell number generating means for generating a number by a counter, and the first cell number according to claim 2 received from the cell number generating means only when there is a sound according to the sound / silence information output from the voice detector. Second cell number and CBR-received from the STM network
The means for multiplexing the CODEC encoded data into a predetermined cell format for cellizing and the means for generating said cell number have two counters, the first counter having a voiced silence received by the voice detector. The counter is a counter that counts the number of cells only when there is a sound cell according to the information, the second counter has a counter that counts the number of cells regardless of whether there is sound or no sound, and the voice encoder has a plurality of excitation sources. The synthesized speech is decoded from the excitation codebook composed of vectors using the linear prediction coefficient, the excitation source vector that gives the synthesized speech closest to the input speech is searched, the optimum value is selected, and the index (code LD-CELP (Low-Delay) for outputting and transmitting the number
Code Excited LinearPredic
in the voiceless cell discarding cell compensation processing means, in order to prevent the ending of the voiced cell section of the voiced cell of the decoded voice received from the second voice decoding means, The present invention is characterized by including a silent cell discard cell compensation processing means for calculating the average amplitude of a fixed section of only the background noise at the tail end and adjusting the average amplitude of the white noise signal output from the white noise signal generation means.

[0019]

Next, the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of the present invention.

In FIG. 1, the voice data converter is C
Terminals P1 and P2 for inputting / outputting BR-CODEC encoded data, terminals P3 and P4 for inputting / outputting VBR-CODEC encoded data cells, discard cell compensation processor 100, first selector 200, and cell separator 300 and cell assembler 4
00, a first speech decoder 500 and a speech detector 600.

The cell separator 300 receives the VBR-CODEC coded data cell S12 input from the P3 terminal, separates only the coded data from S12, and outputs the coded data signal S10. 302
And the signal S10 is input, and the signal S is input irregularly.
If 10 is a constant speed and there is no signal S10,
That is, while the cell S12 is not received, the speed converter 301 that outputs the coded data signal S7 that can be decoded by the CBR-CODEC in which the coded data is "0", and the signal S12 are input, and the signal S12 is input. 2 cell numbers are taken out from the cell number separator 304, which outputs the two cell numbers as signals S11 and S13, and S11 and S13, and outputs as a signal S8 whether it is silent or voiced, Signal S9 indicating whether or not cell discard has been performed
And a cell information judging device 303 for outputting as.

The cell assembler 400 receives the signal S21 as an input and outputs as S15 a first cell number that changes only when there is a voiced cell, and outputs a second cell number that changes regardless of whether the cell is a voiced or silent cell. Cell number generator 4 for outputting as S18
01 and signals S15, S18, S19, and S21 are input, and the VBR-CODEC encoded voice data cell is set to the signal S.
And a multiplexer 402 for outputting as 17.

The discard cell compensation processor 100 outputs the CBR-CODEC encoded data S output from the cell separator 300.
A second voice decoder 106 which receives 7 as an input and outputs a decoded voice as S5, a voiced discard cell compensation processor 105 which receives as input a signal S5 and discard cell compensation data as a signal S4, and white noise And a signal S2, S for a silent time loss cell compensation processor 103 for inputting the signals S3, S5 and outputting cell loss compensation data assuming a silence as a signal S2.
4 and the signal S8 as input and the second selector 102 which outputs the signal S4 when there is sound and outputs the signal S2 as the signal S1 when there is no sound, and the voice which inputs the signal S1 and outputs the encoded voice data as the signal S6. The encoder 101 is provided.

Next, the operation will be described.

The signal S19 input from the STM network side through the P2 terminal is input to the cell assembler 400 and the first speech decoder 500.

The first speech decoder 500 decodes the inputted signal S19 into speech just like a normal CBR-CODEC decoder and outputs it as a signal S20.
The voice detector 600 receives the signal S20 and sends it to the VBR-CO.
The same process as the voice detection performed by DEC is performed. That is, the power of the signal 21 is constantly monitored, and when the power is equal to or higher than a certain value, it is set as a voiced section, and when the power is less than a certain value, it is set as a non-voiced section including only background noise, and voiced / unvoiced information is signaled. It is output as S21.

The cell assembler 400 includes a CBR-CODE
The C encoded data (signal S19) and the voice detection result (signal S21) are input, and the signal S19 is input to the multiplexer 402.
The signal S21 includes a cell number generator 401 and a multiplexer 402.
Respectively input to.

CBR-CODE on STM network
LD-CELP audio coding is performed in the C encoder.

In the LC-CELP system speech coding, the PCM speech signal of 5 samples is set as one vector, and the speech coding process is performed in the unit of one vector. For an input 1-vector speech signal, a synthetic speech is decoded from an excitation codebook (codebook) composed of a plurality of excitation power source vectors by using linear prediction, and the synthetic speech closest to the input speech is decoded. Is searched from the codebook, the optimum value is selected, and only the index (code number) is transmitted as the output.

The cell number generator 401 has a first counter that changes regardless of a silent section and a sound section cell, and a second counter that changes only for cells in a sound section, and has a first counter value. To the multiplexer 402 as the first cell number and the signal S15 as the second cell number and as the signal S18 as the second cell number.

The multiplexer 402 multiplexes the first and second two cell numbers (signals S15 and S18) and the CBR-CODEC coded data (signal S19). At that time, CB
A VBR-CODEC coded data cell composed of R-CODEC coded data and a header containing the first and second cell numbers is assembled. Further, based on the result of the voice detection (signal S21), only the cells in the voiced section are output as the signal S17.

The signal S17 is transmitted from the P4 terminal to the ATM network.

The signal (signal S12) input from the ATM network side through the P3 terminal is input to the cell separator 300.

The signal S12 input to the cell separator is input to the cell number separator 304 and the encoded data separator 302.

In the cell number separator 304, the VBR-CO
The DEC side separates the first and second cell numbers multiplexed in the cell from the cell, and outputs the first cell number as the signal S11 and the second cell number as the signal S13.

The cell information judging device 303 detects the cell discard based on the first cell number (signal S11) and the second cell number (signal S13). The operation of the cell information determiner 303 for detecting cell discard will be described with reference to FIG.

As shown in FIG. 7, the cell information judging device 303
(See FIG. 1) compares the first and second cell numbers with the cells before and after the reception, and if the second cell numbers are not consecutive, determines that the cells in the silent section are discarded. The judgment result is output to the signal S8. In the silent cell discarding portion shown in FIG. 7, it is detected that the second cell number or the number "2" is discontinuously changed to "5", and the cell information judging unit 303 detects the silent cell Judge that it has been discarded.

When both the first and second cell numbers are discontinuous by the same number, it is determined that the cells in the sound section are discarded and the determination result is output to the signal S8. In the place where the voiced cell is discarded as shown in FIG. 7, both the first cell number and the second cell number are discontinuous by “3”, and therefore the cell information judging unit 303 determines that the voiced cell is discarded. It is determined that

In the encoded data separator 302, the signal S1
The encoded data is separated from 2, and output as a signal S10. At this time, even if the cell is discarded and the cell is not input through the signal S12, some data including the synchronization signal of the CODEC frame, for example, data of all 0 except the synchronization signal of the CODEC frame is output.

The speed converter 301 speed-converts the cell data signals S12 and S10, which are normally input intermittently at high speed, to a fixed speed that can be normally decoded by CBR-CODEC, and outputs as a signal S7.

The discarded cell compensation processor 100 has a CBR-
Coded data that can be decoded by CODEC (signal S7) and a signal S8 in which a voiced section and a silent section are determined are input.

The signal S7 is input to the second speech decoder 106, the speech is decoded in the same manner as in a normal CBR-CODEC decoder, and output as a signal S5.

The decoded voice is input to the silent cell discard cell compensation processor 103.

Based on the decoded speech up to the section immediately before the input signal S5, the speech-cell-discarded cell compensation processor 105 determines that the current cell is a cell in the speech section and that the cell is discarded. Then, a process for compensating the voice data for the discarded cell is performed. For example, in the process called pattern matching waveform interpolation method, a waveform similar to the waveform of the cell section immediately before the discarded cell section is selected from the front, that is, the past waveform, and one cell immediately after the selected waveform is selected. The waveform is used to compensate for the discarded cell section. The voice data compensated by the voiced voice discard compensation process 105 is output as a signal S4.

The white noise generator always generates white noise, which is sent to the silent loss compensation processor 103 through the signal S3.

In the silent cell discard cell compensation processor, the RMS value of the power of the last several tens of ms of the decoded speech input by the signal S5 is calculated, and the white noise input by the signal S3 has the same RMS value power. The gain of the white noise is adjusted, and the signal S2 is output.

The second selector 102 outputs the output (signal S4) of the sound cell discard cell compensation processor in the sound section by the signal S8, and outputs the silence cell discard cell compensation processor in the sound section. (Signal S2) is selected and output as signal S1.

In the speech encoder 101, the second selector 1
The output of No. 02 (signal S1) is encoded in exactly the same way as the CBR-CODEC encoder and output as signal S6.

On the basis of the signal S9, the first selector 200 outputs the discarded cell compensation processor 100 (signal S6) in the case of a voiced discard cell section or a silent cell discard cell section.
CBR-CO separated from the cell separator
The DEC encoded data (signal S7) is selected and output as the signal S14.

The signal S14 is output from the P1 terminal as CBR-CODEC encoded data.

[0051]

As described above, according to the voice data conversion device of the present invention, the VB on the ATM network is used.
R-CODEC and CBR-CO on STM network
It is not necessary to perform PCM decoding even in the DEC interconnection, and mutual conversion can be performed with the encoded data as it is. This makes it possible to prevent deterioration of voice quality due to interconnection.
In addition, the device of the present invention, as shown in FIG.
It is realized with the same components as ODEC, and the device of the present invention can be realized by simply changing the connection of the components of VBR-CODEC. This has the effect of reducing the hardware scale.

The apparatus of the present invention has an effect that it is possible to construct a network in which an ATM network and an STM network are mixed and mutually connected with higher voice quality.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of an audio data converter according to the present invention.

FIG. 2 VBR-CO with silence compression function for ATM transmission network
It is a block diagram which shows a DEC encoding part.

FIG. 3 VBR-CO with silence compression function for ATM transmission network
It is a block diagram which shows a DEC decoding part.

FIG. 4 shows a CBR-CODEC encoder for STM transmission network.

FIG. 5 shows a CBR-CODEC decoding unit for the STM transmission network.

FIG. 6 illustrates the prior art of VBR-CODEC and CBR-CODEC interconnects.

FIG. 7 shows detection of cell discard by two cell numbers.

[Explanation of symbols]

 10 voice data converter 100 discarded cell compensation processor 101 speech encoder 102 second selector 103 silent cell discarded cell compensation processor 104 white noise generator 105 voiced discarded cell compensation processor 106 second speech decoding Coder 200 first selector 300 cell separator 301 rate converter 302 coded data separator 303 cell information determiner 304 cell number separator 400 cell assembler 401 cell number generator 402 multiplexer 500 first voice Decoder 600 Speech detector 701 Speech detector 702 Speech encoder 703 Cell assembler 801 Discarded cell compensation processor 802 Speech decoder 803 Cell separator 901 CBR-CODEC speech encoder 1001 CBR-CODEC speech decoding Vessel 1101 VBR-CODEC 1102 CBR-CODEC

Claims (8)

[Claims] 1. A variable bit rate voice CODEC (Vaei) by subjecting a telephone band voice signal to silence compression processing.
av Bit Rate Audio CODEC: VB below
R-CODEC) and a fixed bit rate voice CODEC (C
and an instant bit rate audio CODEC: hereinafter referred to as CBR-CODEC), which enables mutual connection of the VBR-CODEC and the CBR-CODEC. 2. An asynchronous transfer mode (Asynchronous T) configured by the VBR-CODEC.
An ATM network that transmits and receives digital data in a transmission mode (hereinafter referred to as ATM), and a CB that does not perform silent compression on the telephone band voice signal.
Synchronous transfer mode (S
ynchronous Transmission M
ode: hereinafter referred to as STM), is provided between the STM network for transmitting and receiving digital data,
Receives a cell composed of voice coded data of a fixed number of bits and a header of a fixed number of bits from the M network, separates the voice coded data of the VBR-CODEC from the cell, and encodes the first format. A cell separator that outputs the data and has means for detecting that the cell has been discarded on the ATM network; and if cell discard is detected by the cell separator, a second cell is added to the discarded cell portion. A discarded cell compensation processor having means for generating voice coded data of the form: and a first type voice coded data output from the cell separator by a cell discard detection signal received from the cell separator, A first selector for selecting a second type of voice coded data output from the discard cell compensation processor, and a CBR-C from the STM network. It receives the DEC coded data,
A first voice decoder for decoding voice, and a voice detector for outputting voiced / unvoiced information of voice by monitoring the power of the decoded voice output from the first voice decoder, CBR-C received from the STM network
The ODEC coded data is divided into blocks of a certain number of bits, and a header of a certain number of bits is added to the blocks to make a transmission cell of a certain number of bits. According to the voiced / soundless information output from the voice detector, 2. A voice data converter according to claim 1, further comprising a cell assembler which sends out cells only when sound is produced. 3. The header having a certain number of bits of the cell has a first number in which all cells are counted by 1 regardless of the presence / absence of voiced information output from the voice detector, and the header of the voice detector. 3. The voice data conversion apparatus according to claim 1, further comprising: a second number counted by 1 only for cells transmitted by the cell assembler as voiced cells according to the voiced / unvoiced information output. 4. A cell for separating the header portion of claim 3 in the cell received from the ATM network, and extracting the first cell number and the second cell number in the header by the cell separator. The number separating means and the first cell number and the second cell number separated by the cell number separating means determine cell discard and further discard of silent cells or discard of voiced cells. The cell information determining means for outputting discard information, the encoded data separating means for separating voice encoded data in the cell received from the ATM network, and the encoded data separated by the encoded data separating means are A cell information judging means for receiving the cell with the first cell number and the second cell number. When the first cell number is discontinuous, it is detected and output as a cell discard of a silent cell, and when the second cell number is discontinuous, A is output.
A voice code output by the encoder of the CBR-CODEC, having means for detecting that a cell has been discarded due to congestion or the like on the TM network and outputting the cell; The voice data conversion according to claim 1, 2, or 3, wherein a bit string composed only of the encoded data is transmitted to the VBR-CODEC at a constant speed regardless of whether there is sound or no sound, and network congestion. vessel. 5. The discarded cell compensation processor includes second speech decoding means for decoding the speech coded data of the first format received from the cell separator, and the second speech decoding. Generating a white noise signal, and a voiced discarded cell compensating means for performing a discarded cell compensation process for a cell discarded when there is a voice based on the voice decoded by the means and the cell discarded information output from the cell separator. Means for calculating the average amplitude information of the silence section of the decoded voice received from the second voice decoding means, and adjusting the average amplitude of the white noise signal output from the white noise generating means. Means and the cell information received from the cell separator according to claim 3, select the discarded cell compensation data which is the output of the voiced cell discard compensating means when there is voice, and the silent cell discard compensating means when there is no voice. of 2. The apparatus according to claim 1, further comprising: second selecting means for selecting the discard cell compensation data which is an output, and voice encoding means for encoding the discard cell compensation data which is an output of the second selecting means. 2. The voice data conversion device according to 2 or 4. 6. The cell assembler receives from the cell number generating means only when there is sound, according to cell number generating means for generating a number by a counter and the sound / silence information output from the voice detector. Means for multiplexing the first cell number and the second cell number of item 2 and the CVR-CODEC encoded data received from the STM network into a predetermined cell format, and generating the cell number. The means has two counters,
The first counter is a counter that counts the number of cells only when there is a voiced cell based on the voiced / unvoiced information received by the voice detector, and the second counter has a counter that counts the number of cells regardless of whether there is voiced or not. 4. The voice coded data conversion device according to claim 1, 2 or 3. 7. The speech coder decodes a synthesized speech from a source code extension composed of a plurality of excitation source vectors using a linear prediction coefficient, and the synthesized speech closest to the input speech is decoded. The LD-CELP method (Low-Delay Code) in which an excitation source vector that gives a value is selected, an optimum value is selected, and an index (code number) is output and transmitted.
Excited Linear Prediction
System) is used, the speech coded data conversion device according to claim 1 or 2. 8. In the silent cell discard cell compensation processing means, the voiced cell section of the decoded voice received from the second voice decoding means is located at the end of the voiced cell section in order to prevent the ending of the voiced cell section. The silent cell discard cell compensation processing means for calculating an average amplitude of a constant section of only background noise and adjusting the average amplitude of the white noise signal output from the white noise signal generating means. Alternatively, the voice data conversion device according to item 5.