JP2002333900A - Sound-encoding/decoding method and sound-transmitting/ receiving device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a sound-encoding/decoding method and a sound-transmitting/receiving device, with which the feeling of incompatibility can be reduced concerning background noises, without increasing the overall amount of transmitting data. SOLUTION: The background noises inputted in a soundless block, which is generated immediately after the start of speaking, are sent as encoded data by a transmitting side device, a noise-synthesizing coefficient is calculated from the background noises provided by decoding the encoded data by a receiving side device, and background noises for prescribed time inputted immediately after the start of each of soundless blocks generated during speaking are sent as encoded data by the transmitting side device. In the receiving side device, the noise-synthesizing coefficient is updated, on the basis of the background noises provided by decoding the encoded data, and during a soundless term following the output term of the decoded background noises, pseudo-noises, generated on the basis of the updated noise synthesizing coefficient, are outputted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech encoding / decoding method and a speech transmitting / receiving apparatus, and more particularly to restricting data transmission from a transmitting side in a silent section and outputting pseudo background noise generated on a receiving side. The present invention relates to a voice encoding / decoding method and a voice transmitting / receiving apparatus configured to be performed.

[0002]

2. Description of the Related Art In a communication system, in order to reduce communication costs and effectively use communication resources, the amount of transmission data is reduced by compressing data and suppressing transmission of unnecessary data. 2. Description of the Related Art In an audio digital transmission system that encodes and transmits audio data, a silence compression technique is employed to reduce the amount of transmitted data. Silence compression is
The transmitting side analyzes the state of the input signal to determine whether it is a sound state or a silence state. If the sound state is present, the coded data of the input signal (voice signal) is transmitted to the communication line. The transmission of encoded data is suppressed, and pseudo noise generated on the receiving side is output during the silent period.

If the synthesis of pseudo noise in a silent section is omitted or the white noise is simply output for simplicity, if the silent section becomes longer, the listener may feel uneasy about disconnection of the line. According to the technology, it is possible to eliminate anxiety of a listener in a silent section and to reduce transmission data in a silent section which is said to occupy about 60% during voice communication. In addition, regarding the noise synthesis technology in the silent section,
For example, it is described in JP-A-05-122165 or JP-A-10-097292. Also, I
G. TU standardized in TU-T 729 Annex
B describes silent compression and noise synthesis techniques in the CS-ACELP system.

[0004]

However, in the conventional silent compression technique, in the section determined to be silent, transmission of the encoded data is stopped on the transmitting side, or a noise having a smaller data amount than the encoded audio data is transmitted. The information is transmitted intermittently, and the receiving side outputs the predicted pseudo noise at the same time as the end of the sound period. Therefore, the listener hears the actual background noise superimposed on the voice of the sender in the sound period, and hears the pseudo noise synthesized on the receiving side in the silent period. However, it is difficult to synthesize pseudo-noise accurately with very short-term prediction. For this reason, according to the conventional silence compression technology, there is a subtle difference between the pseudo noise output in the silence section and the actual background noise heard in the speech section, and the listener feels discomfort, Poor voice quality. Further, when it is determined from the state of the input signal that there is sound or no sound, there is no guarantee that a silent section is not necessarily a section in which only noise is input. For this reason, on the receiving side, if a control method is used in which a decoded signal of the received coded data is output in a sound interval and a pseudo noise output is immediately switched in a silent interval, when a switch is made from a sound interval to a silent interval, There is a problem that there is a sense of incongruity at the end of the voice and the signal quality is degraded.

On the other hand, there is a method in which noise information is transmitted at a constant cycle even in a section determined to be silent, and pseudo noise is synthesized while the coefficient information necessary for noise synthesis is updated on the receiving side as needed. This intermittent noise information transmission method has a problem that, instead of improving the quality of the pseudo noise, the overall transmission data amount increases, which contributes to data delay on the transmission path.

An object of the present invention is to provide a speech coding / decoding method and a speech transmission / reception apparatus which can reduce the discomfort of background noise without increasing the overall transmission data amount. An object of the present invention is to provide a voice encoding / decoding method and a voice transmitting / receiving apparatus capable of avoiding deterioration of voice quality at the time of switching from a voiced section to a voiceless section without increasing the overall transmission data amount. It is in.

[0007]

In order to solve the above-mentioned problems, according to the speech encoding / decoding method of the present invention, encoded data of input speech is transmitted from a transmitting apparatus at a predetermined period, and the receiving apparatus transmits the encoded data. In a voice transmission system that decodes and outputs the encoded data, the transmitting device transmits background noise input in a silent section immediately after the start of a call as encoded data, and the receiving device converts the encoded data. A noise synthesis coefficient is calculated from the background noise obtained by decoding, and the transmission-side device transmits background noise for a predetermined time inputted immediately after the start of each silent period occurring during a call as encoded data. Updating the noise synthesis coefficient based on the background noise obtained by decoding the encoded data at the receiving device, and during the silent period following the output period of the decoded background noise, And outputs a pseudo noise which is synthesized based on the sound synthesis coefficient.

[0008] According to the configuration of the present invention, when switching from a sound period to a silence period, background noise for a predetermined time inputted immediately after the start of the silence period is transmitted as coded data, and the reception-side device transmits the coded data. Since the decoded background noise is output, even if the weak signal of the sender is included in the input signal determined to be a silent section, the weak signal is transmitted to the receiver. This makes it possible to eliminate a sense of incongruity when switching from a sound section to a silent section. In addition, the receiving apparatus calculates a noise synthesis coefficient from background noise obtained by decoding the encoded data received in the first silent section immediately after the call, and calculates the code received in each silent section generated thereafter. Since the noise synthesis coefficient is updated based on the background noise obtained by decoding the encoded data, it is possible to synthesize the pseudo noise in a form following the temporal change of the background noise on the transmission side, Natural background noise can be given to the listener in the silent section.

More specifically, according to the present invention, the transmitting apparatus transmits identification information of a sound section and a silent section along with each encoded data of the input speech and the background noise, and the receiving apparatus performs According to the identification information, it identifies whether the received coded data is for input speech or for background noise, and executes processing for calculating or updating the noise synthesis coefficient when receiving coded data for background noise. It is characterized by doing. In a preferred embodiment of the present invention, the receiving device determines the final coded data for background noise based on the number of coded data for background noise received in each silent section during a call, and decodes the coded data. The obtained background noise is corrected and output so as to be gradually approximated to the pseudo noise, and is output.

[0010] A voice transmitting / receiving apparatus according to the present invention includes a transmitting unit for transmitting encoded voice to a communication line periodically and a receiving unit for decoding encoded data received from the communication line and outputting the decoded voice. The transmitting unit determines from the input audio signal a voiced section or a silent section, and converts the voice signal input in the voiced section and the background noise input in the voiceless section into encoded data. Means for converting the encoded data for a predetermined period of time generated immediately after the start of each silent section and the encoded data generated in the sound section along with the voice / silence determination information to the communication line. Means for separating the coded data received from the communication line from the voiced / silent determination information, and a first, a second voice / silence determination information from the voiced / silent determination information. Generate 2 control signals Means for calculating or updating a noise combining coefficient based on feature information extracted from the received encoded data during a specific period indicated by the first control signal; and Noise synthesis means for synthesizing pseudo noise at a predetermined period and outputting the pseudo noise intermittently according to the second control signal;
During reception of the encoded data, the speech signal or background noise generated by the decoding means is output, and when there is no encoded data, pseudo noise generated by the noise synthesis means is output. And

[0011]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows a first embodiment of a speech encoding / decoding system according to the present invention. In the figure,
101A denotes a transmitting unit of the voice transmitting / receiving device A, 102B denotes a receiving unit of the voice transmitting / receiving device B, and 100 denotes a communication line (telephone network) connecting the devices A and B. The voice transmitting / receiving device A is 10
2B, and a voice transmitting / receiving apparatus B
A is provided with a transmission unit similar to A, but here, for the sake of simplicity, a speech encoding / decoding technique according to the present invention will be described focusing on transmission / reception of encoded data between the transmission unit 101A and the reception unit 102B. I do.

An audio signal S1 input at a sampling cycle from an input terminal IN is input to an audio encoding unit 2 and a sound / non-speech determining unit 5 sequentially. The audio encoding unit 2 encodes the audio signal accumulated within a unit time (several tens of sampling periods), which is an encoding processing cycle, by a predetermined encoding method adopted in the communication system, and encodes encoded data ( Packet) S2. The sound / non-speech determining unit 3 receives the input voice S1 or the encoded data S generated by the voice coding unit 2.
2 to determine whether the input signal S1 is a sound state or a soundless state by extracting a sound power or spectrum from a part of the sound signal 2 and comparing it with a preset threshold value. Output. For example, the sound / non-speech determination signal S3 is output as a pulse signal representing a sound state by bit “1” and a silence state by bit “0”, and is input to the period determination unit 5 and the data transmission unit 6. Period determination unit 5
Is external when the receiver picks up the handset, for example,
The transmission period determination operation of the encoded data S2 is started with the call start signal S0 input from the exchange as a trigger.

FIG. 2A shows an input voice to the input terminal IN, FIG. 2B shows a speech / non-speech determination signal S3, and FIG. Indicates the sending state. In the present embodiment, for example, as can be seen from the transmission state of the encoded data shown in (C), in addition to the voiced sections (periods T1 and T3), a silent section (period T0) immediately after the start of a call (t0) A predetermined period ΔT set immediately after the start of each silent section that occurs thereafter is defined as a transmission period of the encoded data S2.

When receiving the call start signal S0, the period determination unit 5 sets the data transmission control signal S as an initial operation mode.
5 is turned on, and waits for the presence / absence determination signal S3 to change from "0" to "1". Sound / silence determination signal S3
After the state once becomes "1", the mode shifts to the normal operation mode. In the normal operation mode, the period determination unit 5 detects a period in which the determination signal S3 is in the “1” state and a predetermined period ΔT after the determination signal S3 changes from the “1” state to the “0” state,
During these periods, the data transmission control signal S5 is turned on. During other periods, the data transmission control signal S5 is kept off.

The data transmission control signal S5 is transmitted to the transmission gate 4
The encoded data S2 output from the audio encoding unit 2 passes through the transmission gate when the data transmission control signal S5 is in the ON state, and is transmitted from the data transmission unit 6 to the communication line 7. You. The coded data transmitted in the silent section (noise analysis period) T0 in the initial operation mode is obtained by encoding background noise on the transmitter side, and is based on the synthesis coefficient of pseudo noise generated in the silent section on the receiving side. Data. In the normal operation mode, the coded data transmitted in the silent section (noise update period) ΔT is used for updating the pseudo noise synthesis coefficient on the receiving side. When switching from a voiced section to a silent section, the transmitting side encodes and transmits a predetermined amount of background noise, and the receiving side modifies the pseudo-noise synthesis coefficient so that the receiving side can talk to the transmitting side. It is possible to generate realistic pseudo noise reflecting the temporal change of the background noise.

The transmission period ΔT of the background noise is obtained, for example, by counting the pulse of the bit “0” inputted in the encoding processing cycle as the voiced / silence determination signal S3, and comparing this with a predetermined threshold N. Can be set. Period determination unit 5
In the normal operation mode, the data transmission control signal S5 is turned on when the determination signal S3 is in the bit “1” state, and the count value remains at the threshold value N even after the determination signal S3 changes to “0”. Until the count value reaches the threshold value N, the data transmission control signal S5 is kept on. When the count value exceeds the threshold value N, or when the determination signal S3 returns to the "1" state before the count value reaches the threshold value N, the data transmission control signal S5 is turned on. The transmission control signal S5 can be configured by a logic circuit that switches to the off state and resets the counter.

The period T0 in the initial operation mode can be determined by setting the threshold value N to a maximum value whose count value cannot exceed. In the present invention, the noise coding in the initial operation mode may be limited to a predetermined period required for noise analysis on the receiving side. That is, a finite threshold value L (where L> N) is set in the initial operation mode, the data transmission control signal S5 is held in the ON state until the count value exceeds the threshold value N, and the count value is set to the threshold value N. When the data transmission control signal S5 is exceeded or when the determination signal S3 changes to the "1" state, the data transmission control signal S5 is switched to the off state,
The counter may be reset. in this case,
For example, as shown in FIG. 3, a predetermined period ΔT0 immediately after the start of a call (t0) is a noise analysis period. Data transmission unit 6
Transmits a voice packet to the communication line 7 in a form in which a voice / non-speech determination bit indicated by the voice / non-speech determination signal S3 is added to each encoded data S2.

On the receiver side, the encoded data sequence (speech packet sequence) transmitted to the communication line 7 is received by the data receiving unit 10, and the encoded data S10 and the sound / non-speech determination bit S are received.
Separated into 11. The encoded data S10 is input to the audio decoding unit 12, and the decoded audio signal S12 is
The signal is supplied to the noise output unit 19. Part of the information generated at the time of decoding the encoded data S10 in the audio decoding unit 12 is input to the characteristic information extracting unit 13, where the characteristic information necessary for the synthesis of the pseudo noise is extracted and stored. Stored in the unit 14.

On the other hand, the sound / non-speech determination bit S 11 separated by the data receiving unit 10 is supplied to the period determining unit 15. The period determination unit 15 is, for example, a call start signal (off-hook signal) generated when the receiver picks up the receiver.
The period determination operation is started with S0 'as a trigger. During the reception of the noise coded data (period T0, ΔT, in FIG. 2), the period determination unit 15 generates the first control signal S15 for instructing the synthesis coefficient operation, and generates the noise coded data and the voice code. During the reception period of the coded data (periods T0, T
1, ΔT, T2, ΔT,...) Generates a second control signal S150 for suppressing the output of the synthesized pseudo noise.

In the initial operation mode, the period determining section 15 determines the first silent section (noise analysis period) T0 from the sound / non-speech determination bit S11, and synthesizes the first control signal S15 during the noise analysis period. The count calculation / update unit 16 is instructed to calculate the combination coefficient. Composite count calculation / update section 16
Captures the decoded audio signal (in this case, background noise) from the characteristic information storage unit 22 during the reception period of the control signal S15 generated in the initial state, and calculates a noise synthesis coefficient. The synthesis coefficient is obtained, for example, by calculating the average value of the background noise in the noise analysis period T0 stored in the feature information storage unit 22. When the noise analysis period is the predetermined period ΔT0 as shown in FIG. 3, the number of bits “0” indicated by the sound / non-sound determination bit S11 is counted, and the time when the count value reaches the threshold L is reached. Then, the output of the control signal S15 is stopped.

After the output of the first control signal S15 is stopped immediately after the start of the call, the period determining unit 15 shifts to the normal operation mode. In the normal operation mode, a speech section and a speech section are identified from a change in the state of the speech / non-speech determination bit S11, and a control signal S15 is generated in a noise update period ΔT immediately after the start of each speech section (head). It instructs the composite count calculation / update section 16 to update the composite coefficient. Noise update period ΔT
Can be determined by starting the counting operation of the bit “0” when the voiced / silent determination bit S11 changes from the bit “1” to the bit “0”, and comparing the count value with the threshold value N.

After once calculating the synthetic coefficient of the pseudo noise, the synthetic coefficient calculating / updating section 16 shifts to the normal operation mode, and from the characteristic information accumulating section 22 during the reception period of the control signal S15 generated thereafter. Capture background noise and update existing synthesis coefficients. The update of the synthesis coefficient is performed, for example, by combining the currently held noise and the new noise with a predetermined weight,
For example, it means an averaging process. In order to avoid an extreme change in the synthesis coefficient, the existing coefficient value may be regarded as important, and a weight greater than the new noise may be given for averaging.

The period determination unit 14 receives the encoded data reception period of the background noise (periods T0 and ΔT in FIG. 2) and the encoded data reception period of the voice (FIG. 2) determined from the voiced / silent determination bit S11. In the periods T1, T2,
...), A pseudo noise output suppression signal (second control signal)
S150 is turned on. The noise synthesizing unit 18 sets the noise synthesizing coefficient of the synthesizing coefficient calculating / updating unit 16 in the noise synthesizing filter in synchronization with the clock output from the timer at a predetermined sampling period, and sets the white noise generating unit (excitation source) 1
The pseudo noise S18 is synthesized from the white noise generated in step 7 and supplied to the audio / noise output unit 19. While the data reception unit 10 is receiving the encoded data, the suppression signal S150 is turned on, the output of the pseudo noise S18 from the noise synthesis unit 18 is suppressed, and the synthesized pseudo noise data is cleared.
Therefore, from the audio / noise output unit 19, the decoded audio or background noise is supplied to the terminal O during the encoded data reception period.
During the silent period in which the coded data is output to the UT and no encoded data is received, the pseudo noise synthesized by the noise synthesis unit 18 is applied to the terminal O.
It will be output to the UT.

FIG. 4 shows a second embodiment of the speech encoding / decoding system according to the present invention. This embodiment employs a weighted moving average calculator 22 instead of the audio / noise output unit 19 in the receiving unit 102B shown in FIG.
2 is characterized in that the speech signal and the noise decoded in step 2 are temporarily stored in the coded speech storage section 21 and then supplied to the weighted moving average calculation section 22.

During reception of the encoded data of the background noise (periods T0, ΔT in FIG. 2) and reception of the encoded data of the audio (periods T1, T2,... In FIG. Alternatively, the decoding noise S12 is output and stored in the decoded voice storage unit 21. The audio signal and the background noise accumulated in the decoded audio accumulation unit 21 are output to the terminal 20 via the audio / noise weighted moving average calculation unit 22. In addition, the pseudo noise output from the noise synthesis unit 18 in the silent section is calculated by the speech / noise weighted moving average calculation unit 22.
Is output to the terminal 20 via the.

In this embodiment, in order to eliminate the discontinuity between the decoding noise output from the speech decoding unit 12 and the synthesized noise S18 output from the noise synthesizing unit 18, the end of each noise update section is eliminated. At the timing when the decoded data is stored in the decoded voice storage unit 21, the noise synthesis unit 18 starts outputting the pseudo noise S 18, and the voice / noise weighted moving average calculation unit 22 outputs the pseudo noise S 18 from the decoded voice storage unit 21. Weighted moving average processing is performed on the fetched decoded noise data and the pseudo noise S18 output from the noise synthesis unit 18, and the result is output as background noise. In this case, the period determination unit 15 compares the count value of the sound / non-sound determination bit S11 of the bit “0” with the threshold “N−1” in each noise update section ΔT, and determines that the count value is equal to the threshold “N−1”. Is returned to the off state at the point of time when the decoding noise and the pseudo noise S1 to the speech / noise weighted moving average calculation unit 22 are returned.
8 simultaneous inputs are possible.

FIG. 5 is a conceptual diagram of the weighted moving average processing. (A) shows the output 1 from the decoded audio storage unit 21;
The waveform of the decoding noise S21 for the packet, (B) is a weighting factor given to the decoding noise S21, (C) is the waveform of the pseudo noise (synthesized noise) S18 output from the noise synthesis unit 18, (D) ) Indicates the weighting factor given to the pseudo noise S18, and (E) indicates the waveform of the corrected pseudo noise output from the voice / noise weighted moving average calculation unit 22.

As shown in FIGS. 3B and 3D, during the output period of the last coded data in each noise update interval ΔT, the weighting factor given to the decoding noise S21 is gradually reduced. By performing a weighted average of the decoding noise and the pseudo noise while gradually increasing the weight coefficient given to S18, the background noise output from the voice / noise weighted moving average calculation unit 22 is gradually changed from the decoding noise S21 to the pseudo noise S18. Can be transferred to. Therefore, according to the present embodiment, it is possible to output a background noise that does not cause a sense of incongruity between a sound section and a silent section on the receiving side.

[0029] The ITU-T defines G.100, which is a standardized system for silence compression technology of CS-ACELP. 729
As a result of comparing the Annex B and the silence compression technique of the present invention by a simple simulation, according to the present invention, the total amount of transmission data in a silence section is 1/4 or less, and the number of times of data transmission is reduced.
It has been found that the problem can be reduced to 1/20 or less and the problem of data delay due to the intermittent transmission method can be solved.

[0030]

As is apparent from the above description, according to the present invention, it is possible to realize a speech encoding / decoding system which does not cause a listener to feel uncomfortable even when the data transmission amount is reduced. In addition, by transmitting encoded data of background noise at the beginning of each silent period and updating the noise synthesis coefficient on the receiving side, it is possible to synthesize pseudo noise that can flexibly cope with ever-changing background noise. Become.

[Brief description of the drawings]

FIG. 1 shows a first embodiment of a speech encoding / decoding system according to the present invention.
FIG.

FIG. 2 is a diagram showing a relationship between a voiced / silent section and a transmission period of encoded data according to the present invention.

FIG. 3 is a diagram showing another example of encoded data transmission in the present invention.

FIG. 4 shows a second example of the speech encoding / decoding system according to the present invention.
FIG.

FIG. 5 is a view for explaining a weighted moving average process executed by the voice / noise weighted moving average calculation unit 22 of FIG. 4;

[Explanation of symbols]

S0: call start signal, S1: input voice signal, 2: voice coding unit, S2: coded data, 3: voice / non-voice determination unit,
S3: Voice / silence determination bit, 4: Transmission gate, 5: Period determination unit, 6: Data transmission unit, 7: Communication line, 10: Data reception unit, S10: Coded data, S11: Voice / silence determination Bit, 12: audio decoding unit, S12: decoded signal, 13: feature information extracting unit, 14: feature information accumulating unit, 1
5: period determination unit, 16: composite coefficient calculation / update unit, 17:
White noise generation unit, 18: noise synthesis unit, S18: pseudo-synthesis noise, 19: audio / noise output unit, 20: audio output terminal,
21: decoded voice storage unit, 22: voice / noise weighted moving average calculation unit, 101A: transmission unit, 101B: reception unit.

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Toku Tsukada 216 Totsuka-cho, Totsuka-ku, Yokohama-shi, Kanagawa Prefecture Inside the Hitachi, Ltd.Communications Division (72) Inventor Mitsuhiro Noda 216 Totsuka-cho, Totsuka-ku, Yokohama-shi, Kanagawa F-term in Hitachi, Ltd., Communications Division, 5D045 CC05 DA11

Claims (6)

[Claims] An audio transmission system in which encoded data of input speech is transmitted from a transmitting device at a predetermined period, and the receiving device decodes and outputs the encoded data. The background noise input in the generated silent section is transmitted as coded data, and the receiving apparatus calculates a noise synthesis coefficient from the background noise obtained by decoding the coded data. The background noise for a predetermined time inputted immediately after the start of each silent section occurring during the transmission is transmitted as coded data, and the reception side device decodes the coded data on the basis of the background noise obtained by decoding the coded data. Update the synthesis coefficient,
A speech encoding / decoding method, comprising outputting pseudo noise synthesized based on the updated noise synthesis coefficient during a silent period following the output period of the decoded background noise. 2. A transmitting apparatus transmits identification information of a voiced section and a silent section along with each encoded data of input speech and background noise, and a receiving apparatus transmits a received code according to the identification information. Identifying whether the digitized data is for input speech or background noise,
The speech encoding / decoding method according to claim 1, wherein a process for calculating or updating the noise combining coefficient is performed when the encoded data for background noise is received. 3. A receiving-side apparatus determines final coded data for background noise based on the number of coded data for background noise received in each silent section during a call, and decodes the coded data. 3. The speech encoding / decoding method according to claim 2, wherein the background noise is corrected and output so as to gradually approximate the pseudo noise. 4. A transmitting device transmits background noise for a predetermined time inputted in a silent section generated immediately after the start of the call as encoded data, and a receiving device decodes the encoded data. A noise synthesis coefficient is calculated from the obtained background noise, and pseudo noise synthesized based on the noise synthesis coefficient is output in a silent section following an output period of the decoded background noise. A speech encoding / decoding method according to any one of claims 1 to 3. 5. A voice transmitting / receiving apparatus comprising: a transmitting section for transmitting input voice as encoded data to a communication line periodically; and a receiving section for decoding encoded data received from the communication line and outputting the decoded data as voice. Means for determining whether the input speech signal is a speech section or a non-speech section, and converting the speech signal input in the speech section and the background noise input in the non-speech section into encoded data. Means for transmitting encoded data for a predetermined period of time generated immediately after the start of each silent section and encoded data generated in a sound section to a communication line together with voice / silence determination information. Means for separating the encoded data received from the communication line from the voiced / silent determination information, and first and second voice / silence determination information from the voiced / silent determination information. For generating control signals Means for calculating or updating a noise combining coefficient based on characteristic information extracted from the received encoded data during a specific period indicated by the first control signal; and a predetermined period based on the noise combining coefficient. And a noise synthesizing means for intermittently outputting the pseudo noise in accordance with the second control signal. The speech generated by the decoding means during reception of the encoded data. A speech transmission / reception apparatus which outputs a signal or background noise and outputs pseudo noise generated by the noise synthesis means when there is no encoded data. 6. A voice transmitting / receiving apparatus comprising: a transmitting section for encoding data of an input voice and periodically transmitting the encoded data to a communication line; and a receiving section for decoding the encoded data received from the communication line and outputting it as voice. Encoding means for converting the input signal including speech and background noise into encoded data at a predetermined cycle,
Means for determining whether the input signal is a sound section or a sound section, and suppressing transmission of the encoded data except for a predetermined period and a sound section set immediately after the start of each sound section. And a means for sending the encoded data output from the gate means to the communication line together with the determination information of the voiced section or the silent section. Decoding means for decoding the received encoded data to generate an audio signal or background noise; feature extraction means for the background noise connected to the decoding means; and a sound source received from the communication line. A means for generating a control signal for calculating a noise synthesis coefficient and a pseudo noise output suppression signal based on the determination information of the section or the silent section; so Means for calculating or updating a noise synthesis coefficient from the extracted feature information; and synthesizing pseudo noise at a predetermined cycle based on the noise synthesis coefficient, and outputting the pseudo noise intermittently according to the output suppression signal. And a means for outputting, to an output terminal, an audio signal or background noise generated by the decoding means and pseudo noise supplied from the noise synthesis means. Voice transmitting and receiving device.