JPH0846556A - Signal processing unit - Google Patents

Abstract

PURPOSE:To reduce the arithmetic operation quantity required for the signal processing unit. CONSTITUTION:A reception signal coded and multiplexed and received from a remote end channel is demultiplexed into a reception signal of each band by a demultiplexer section 12 and the result is outputted to a band split decoder 13. The band split decoder 13 decodes a reception signal of each band and outputs the result to a band split echo cancellation section 3 and a band synthesis processing 14. The band synthesis processing section 14 synthesizes reception signals of each band and outputs the result to a near end channel. A band split processing section 9 divides the transmission signal including an echo of the reception signal into plural bands and outputs it to a band split echo cancellation section 3. The band split echo cancellation section 3 cancels echo for each band and applies band division to the signal after echo cancellation and provides an output to a multiplexer section 11.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a signal processing device for voice or audio signals used in a voice conference communication system or a long distance communication system.

[0002]

2. Description of the Related Art In long-distance telephone communication using a communication satellite, a submarine cable or the like, echo generated by impedance mismatch in a 2-wire / 4-wire converter of a line deteriorates call quality. Also, in audio conference communication systems and loudspeaker phones, the output signal from the speaker is reflected indoors,
Echoes created by wrapping around the microphone interfere with a comfortable call. For example, in the remote audio conference, when the speaker of the other terminal speaks while the own terminal and the other terminal are connected via the transmission line, the voice is input from the microphone,
When it reaches the own terminal via a long transmission path, the voice (received signal) from the other terminal circulates from the speaker to the microphone at the own terminal, and when it reaches the other terminal after a long transmission path, echoes to the speaker of the other terminal. Is heard and it becomes difficult to talk.

The echo canceller is intended to cancel this echo, and in order to prevent the echo from being heard at the other terminal, the echo canceller is installed between the own terminal and the transmission path. The echo canceller inputs the transmission signal in which the voice (echo) in which the voice (received signal) from the other terminal sneaks into the microphone from the speaker of the own terminal and the voice of the speaker of the other terminal are input, and the other terminal A pseudo echo is generated based on the received signal from the signal and a signal (residual signal) from which the echo is removed by subtracting the pseudo echo from the transmission signal is sent to the partner terminal. As a result, the speaker of the other terminal cannot hear the echo.

The same applies to the case of the speaker of the own terminal. For the voice of the own terminal, the echo of the speaker of the own terminal can be removed by an echo canceller installed between the other terminal and the transmission path. Further, even when the speakers of both the own terminal and the partner terminal speak at the same time, echo cancellation is similarly performed by the echo cancellers installed between the partner terminal and the transmission path. Pseudo-echo is obtained by passing the received signal through a filter having the same characteristics as the impulse response of the echo path from the received signal output of the echo canceller to the transmitted signal input, and the echo can be canceled by subtracting this from the transmitted signal. . The echo canceller estimates the impulse response of the echo path, and performs a product-sum operation of the impulse response estimation value and the past received signal sequence to filter the received signal.
Here, assuming that the impulse response estimation value of N samples and the received signals of past N samples are stored in the echo canceller and N times of product-sum calculation is performed, it is realized by such a product-sum calculation. Since the filter has a finite impulse response of N samples,
It is called a FIR (Finite Impulse Response) filter.

If the echo path is a linear system, its characteristics are uniquely determined by the impulse response. Therefore, when the impulse response of the echo path and the impulse response estimated by the echo canceller are equal, this FIR is used. The pseudo echo output by the filter and the echo signal generated in the echo path become equal, and the echo canceller can cancel the echo by subtracting this pseudo echo from the transmission signal. However, if the estimated impulse response is different from the actual impulse response of the echo path, the echo cannot be canceled and a residual echo occurs.
For the impulse response of the echo canceller, estimate the difference between the impulse response of the actual echo path and the impulse response of the actual echo path from the residual signal obtained by subtracting the pseudo echo from the transmitted signal and the received signal so that the impulse response of the actual echo path approaches Operate.

In recent years, in order to cancel the echo generated in a very long echo path from a speaker to a microphone, such as a voice conference communication system, a band division type in which a transmission / reception signal is divided into a plurality of bands and echo cancellation is performed for each band. Echo cancellers are in the spotlight. The band division type echo canceller has the following advantages as compared with an echo canceller which performs signal processing without band division. (1) Parallel processing is easy to realize because signal processing is performed for each band. (2) Due to the nature of the impulse response estimation algorithm, the estimation speed tends to decrease for signals with large autocorrelation, such as speech signals, but band-divided signals have smaller autocorrelation than signals before division. , It is possible to increase the estimated speed.

Hereafter, the terminal side where the microphone and the speaker are present is referred to as a near-end speech path, and the transmission path side is referred to as a far-end speech path. A signal including an echo from the near-end speech path is called a transmission signal, a signal obtained by subtracting a pseudo echo from the transmission signal is called a residual signal, and a signal from the far-end speech path is called a reception signal. An echo path extends from the reception signal output end of the echo canceller to the transmission signal input end of the echo canceller via the speaker / microphone on the terminal side. Conventionally, there is a band division type echo canceller shown in FIG. FIG. 11 is shown in Japanese Unexamined Patent Publication No. 3-243020, where 1 is a received signal band division processing unit for dividing the band of the received signal from the far-end speech path, and 2 is a band of the transmission signal from the near-end speech path. A transmission signal band division processing unit for performing division, 3 is a band division echo cancellation unit for performing echo cancellation processing for each of the divided bands, and 4 is a band synthesis processing unit for combining the band-divided signals.

Next, the operation will be described. The received signal input from the far-end speech path side is output to the near-end speech path, divided into a plurality of bands by the received signal band division processing unit 1, and output to the band division echo cancellation unit 3. Further, the transmission signal input from the near-end speech path side shown in FIG. 11 is received by the transmission signal band division processing unit 2 by the reception signal band division processing unit 1.
Are divided into the same number of bands as and are output to the band division echo canceller 3. The band division echo cancellation unit 3 estimates the impulse response of the echo path for each band, and the received signal of each band input from the received signal band division processing unit 1 is converted into an FIR filter having the estimated impulse response of the echo path as a coefficient. A pseudo echo signal is generated by passing the signal, and the echo is canceled by subtracting the pseudo echo signal from the band-divided transmission signal input from the transmission signal band division processing unit 2. The signal after echo cancellation of each band, that is, the residual signal is output to the band synthesis processing unit 4. The band synthesis processing unit 4 synthesizes the residual signal input from the band division echo cancellation unit 3 into one original band and outputs it to the far-end speech path.

Next, the above-mentioned echo canceller has a drawback that sufficient echo cancellation cannot be performed in the presence of near-end speech. That is, when the impulse response is estimated when the near-end speaker transmits, it operates as if the characteristics of the echo path have changed, and the estimation accuracy decreases, resulting in insufficient echo cancellation. Therefore, it is possible to add a near-end call detection function to prevent estimation deterioration by stopping the estimation of impulse response when a near-end call is detected. FIG. 12 shows an example of the configuration of an echo canceller equipped with such a near-end speech detector. In the figure, 3 is a band division echo cancellation unit having an impulse response estimation operation / stop control function, 5 is a reception signal level calculation unit for calculating the reception signal level of each band, and 6 is for calculating the transmission signal level of each band. A transmission signal level calculation unit, 7 is a residual signal level calculation unit for calculating the residual signal level after echo cancellation in each band, 8
Is a near-end call detector that determines whether or not a near-end call signal is present, and 18 is a control signal from the near-end call detector 8, which activates or stops the impulse response estimation of the band division echo canceller 3.

Next, the operation of the band division type echo canceller shown in FIG. 12 will be described. The reception signal band division processing unit 1, the transmission signal band division processing unit 2, and the band synthesis processing unit 4 perform the same operations as in the case of FIG. The reception signal level calculation unit 5 receives the band-divided reception signal output from the reception signal band division processing unit 1, calculates the signal level of each band, and outputs the calculation result to the near-end speech detection unit 8. The transmission signal level calculation unit 6 receives the band-divided transmission signal output from the transmission signal band division processing unit 2, calculates the signal level of each band, and outputs the calculation result to the near-end speech detection unit 8. The residual signal level calculation unit 7 inputs the residual signal after echo cancellation of each band output from the band division echo cancellation unit 3, calculates the signal level of each band, and outputs the calculation result to the near-end speech detection unit 8. Output. The near-end call detection unit 8 receives the received signal level of each band,
Based on the transmission signal level and the residual signal level after echo cancellation, it is determined whether or not there is a near-end speech signal for each band, and if it is determined that there is no near-end speech signal in all bands. The control signal 18 controls the band-division echo canceller 3 to estimate the impulse response, and otherwise stops the impulse response estimation.

The operation of the band division echo canceller 3 will be described with reference to FIG. FIG. 13 is an internal configuration diagram of one band of the band division echo cancellation unit 3, 31 is a received signal storage unit, 32 is an estimated impulse response storage unit, 33 is a pseudo echo generation unit, 34 is a subtraction unit, and 35 is an impulse. The response estimation unit, 18 is a control signal from the near-end call detection unit 8. The reception signal storage unit 31 inputs the reception signal of the band to be processed from the reception signals of the respective band-divided bands output from the reception signal band division processing unit 1 in FIG. Store samples. The estimated impulse response storage unit 32 stores the estimated impulse response for N samples, for example. The pseudo echo generation unit 33 includes a past received signal sequence stored in the received signal storage unit 31,
A pseudo echo is generated by performing a product-sum operation with the impulse response estimation value stored in the estimated impulse response storage unit 32, and outputs the pseudo echo to the subtraction unit 34.

The subtraction unit 34 inputs the transmission signal of the band to be processed from the transmission signals of the respective band-divided bands output from the transmission signal band division processing unit 2, and from this, the pseudo echo generation unit 33. The pseudo echo output by the above is subtracted, and the residual signal is output to the band synthesis processing unit 4 and the impulse response estimation unit 35. In this way, the received signal storage unit 31, the estimated impulse response storage unit 3
2. The FIR echo filter is configured by the pseudo echo generation unit 33. The impulse response estimation unit 35 estimates the impulse response of the echo path by using the past received signal sequence stored in the received signal storage unit 31 and the residual signal after echo cancellation output from the subtraction unit 34, The contents of the estimated impulse response storage unit 32 are updated. However, when the control signal 18 from the near-end call detection unit 8 instructs the estimation of the impulse response, the control signal 18 is updated and the control signal 18 is stopped so as to stop the estimation of the impulse response.
Do not update if instructed by.

By operating as described above, the band division type echo canceller shown in FIG. 12 can stop the estimation of the impulse response of the echo path in the presence of the near-end speech signal. In addition to the method of calculating the signal level in all bands to detect the near-end call, a method of detecting the near-end call by monitoring the signal level of only one band is also conceivable. Monitoring the level allows more reliable detection.
As a method of obtaining the signal level, there are a method of sampling the signal at a constant time interval and performing a moving average of the square of the amplitude, a method of performing the moving average of the amplitude of the amplitude, and the like.

[0014]

As described above, the band-division type echo canceller and the band-division type high-efficiency codec have a problem that the amount of calculation increases because the band-division / combining process is required. . Further, since the band-division type echo canceller having the near-end speech detection function needs to calculate the signal level of each band, there is a problem that the amount of calculation required for calculating the signal level increases.

The present invention has been made to solve such a problem, and a first object thereof is to reduce the calculation amount required for the band division / synthesis processing without impairing the echo canceling function. is there. A second object is to reduce the amount of calculation required for near-end speech detection without degrading the near-end speech detection capability of the echo canceller in the signal processing device. A third object is to reduce the calculation amount of impulse response estimation without impairing the echo canceling performance of the echo canceller in the signal processing device.

[0016]

A signal processing apparatus according to the first invention comprises a separating means for separating, for each band, a received signal from a far-end speech path coded and multiplexed for each of a plurality of bands. Band division decoding means for decoding the reception signal separated for each band, band combination processing means for combining the decoded reception signals for each band and outputting to the near-end speech path, and reception combined by the band combination processing means Band division processing means for dividing the transmission signal from the near-end speech path including the echo of the signal into a plurality of bands, and a pseudo echo is generated by referring to the reception signal decoded by the band division decoding means, and the band division processing means The echo canceling means for subtracting the pseudo echo from the band-divided transmission signal output by each to output the residual signal for each band, the band-division coding means for coding the residual signal for each band, and the code for each band Become Multiplexing the residual signal is obtained by a multiplexing means for outputting the far-end communication path.

The signal processing apparatus according to the second aspect of the present invention comprises a separating means for separating, for each band, a received signal from the far-end speech path that is highly efficient coded and multiplexed by adaptive quantization for each of a plurality of bands. Band division decoding means for decoding the reception signal separated for each band, band combination processing means for combining the decoded reception signals for each band and outputting to the near-end speech path, and reception combined by the band combination processing means Band division processing means for dividing a transmission signal from the near-end speech path including signal echo into a plurality of bands, and impulse response of the echo path in the near-end speech path is estimated for each band and decoded by the band division decoding means. Echo canceling means for generating a pseudo echo in accordance with the estimated result of the received signal and the impulse response, subtracting the pseudo echo from the band-divided transmission signal, and outputting the residual signal of each band; Band division coding means for highly efficient coding the residual signal for each band by adaptive quantization and outputting the quantization width, residual signal level conversion means for converting the quantization width to a residual signal level, and band division processing Near-end speech detection means for detecting the presence or absence of near-end speech based on the level of the transmission signal band-divided by the means, the level of the reception signal decoded by the band-division decoding means, and the residual signal level converted by the residual signal level conversion means. And an impulse response estimation control means for stopping the estimation of the impulse response of the echo canceling means when the near-end speech detection means detects a near-end speech.

A signal processing apparatus according to a third aspect of the present invention includes a separating means for separating, for each band, a received signal from a far-end speech path which is highly efficient coded and adaptively multiplexed by adaptive quantization for each of a plurality of bands. Decode the received signal separated for each band,
A band division decoding unit that outputs a band-divided reception signal and a quantization width, a reception signal level conversion unit that converts the quantization width into a reception signal level, and a band division reception signal that the band division decoding unit outputs Band combining processing means for combining and outputting to the near-end speech path, and band dividing processing means for dividing the transmission signal from the near-end speech path including the echo of the received signal combined by the band combining processing means into a plurality of bands, The impulse response of the echo path in the near-end speech path is estimated for each band, a pseudo echo is generated according to the estimation result of the received signal and the impulse response decoded by the band division decoding means, and the pseudo echo is generated from the band-divided transmission signal. Echo canceling means for reducing and outputting the residual signal of each band, the level of the band-divided transmission signal, and the level of the residual signal output by the band-dividing echo canceling means. , And the near-end call detecting means for detecting the presence or absence of a near-end call from the received signal level output by the received signal level converting means, and the impulse response of the echo canceling means if the near-end call detecting means detects the near-end call And impulse response estimation control means for stopping.

A signal processing device according to a fourth aspect of the present invention comprises a separating means for separating, for each band, a received signal from a far-end speech path which is highly efficient coded and multiplexed by adaptive quantization for each of a plurality of bands. Decode the received signal separated for each band,
A band division decoding unit that outputs a band-divided reception signal and a quantization width, a reception signal level conversion unit that converts the quantization width into a reception signal level, and a band division reception signal that the band division decoding unit outputs Band combining processing means for combining and outputting to the near-end speech path, and band dividing processing means for dividing the transmission signal from the near-end speech path including the echo of the received signal combined by the band combining processing means into a plurality of bands, The impulse response of the echo path in the near-end speech path is estimated for each band, a pseudo echo is generated according to the estimation result of the received signal and the impulse response decoded by the band division decoding means, and the pseudo echo is generated from the band-divided transmission signal. Echo cancellation means for subtracting and outputting the residual signal of each band, and band division for highly efficient coding of the residual signal for each band by adaptive quantization and outputting the quantization width Encoding means, residual signal level converting means for converting the quantization width to a residual signal level, level of the band-divided transmission signal, residual signal level converted by the residual signal level converting means, and received signal level Near-end call detecting means for detecting the presence or absence of a near-end call from the received signal level output by the converting means, and impulse response for stopping the estimation of the impulse response of the echo canceling means if the near-end call detecting means detects the near-end call And an estimation control means.

A signal processing device according to a fifth aspect of the present invention is a reception signal band division processing means for dividing a reception signal from the far-end speech path into a plurality of bands, and a near signal including an echo of the reception signal from the far-end speech path. A transmission signal band division processing means for dividing the transmission signal from the end speech path into a plurality of bands, an impulse response of the echo path in the near end speech path is estimated for each band, and a reception signal decoded by the band division decoding means And an echo canceller that generates a pseudo echo according to the estimation result of the impulse response, subtracts the pseudo echo from the band-divided transmission signal, and outputs the residual signal of each band, and a predetermined band among the divided bands in the echo canceler. Bands are grouped into a plurality of band groups, and thinning control means is provided for controlling the estimation of the impulse response to stop for each of the plurality of band groups.

[0021]

In the first aspect of the invention, the separating means separates the received signal from the far-end speech path coded and multiplexed for each of a plurality of bands for each band, and the band division decoding means for each band. The received signal is decoded. Then, the band synthesis processing means synthesizes the decoded received signals for each band and outputs the synthesized signal to the near-end speech path, and the band division processing means from the near-end speech path including the echo of the received signal synthesized by the band synthesis processing means. The transmission signal of is divided into a plurality of bands. Then, the echo canceling unit generates a pseudo echo by referring to the reception signal decoded by the band division decoding unit, subtracts the pseudo echo from the band-divided transmission signal output by the band division processing unit, and subtracts the residual error for each band. Output a signal. Further, the band division coding means codes the residual signal for each band, and the multiplexing means multiplexes the coded residual signal for output to the far-end speech path.

In the second aspect of the invention, the separating means separates the received signal from the far-end speech path, which is highly efficient coded and multiplexed by adaptive quantization for each of the plurality of bands, into the band division decoding means. Decodes the received signal separated for each band. Then, the band synthesizing processing unit synthesizes the decoded received signals for each band and outputs the synthesized signals to the near-end speech path. Further, the band division processing unit divides the transmission signal from the near-end speech path including the echo of the reception signal combined by the band combination processing unit into a plurality of bands. Then, the echo canceling means estimates the impulse response of the echo path in the near-end speech path for each band, and generates a pseudo echo according to the estimation result of the received signal and the impulse response decoded by the band division decoding means, and divides the band. The pseudo echo is subtracted from the transmission signal and the residual signal of each band is output. Further, the band division coding means performs high efficiency coding of the residual signal for each band by adaptive quantization and outputs a quantization width, and the residual signal level converting means converts the quantization width into a residual signal level. Then, the near-end speech detection means uses the level of the transmission signal band-divided by the band-division processing means, the level of the reception signal decoded by the band-division decoding means, and the residual signal level converted by the residual signal level conversion means. The presence or absence of a call is detected, and the impulse response estimation control means stops the estimation of the impulse response of the echo canceling means when the near-end speech detection means detects a near-end speech.

In the third invention, the separating means separates, for each band, the received signal from the far-end speech path which is highly efficient coded and adaptively multiplexed by adaptive quantization for each of the plurality of bands, and the band division decoding means. Decodes the received signal separated for each band, and outputs the band-divided received signal and the quantization width. Then, the reception signal level conversion means converts the quantization width into a reception signal level. The band synthesizing processing unit synthesizes the band-divided reception signals output by the band division decoding unit and outputs the synthesized signal to the near-end speech path. Next, the band division processing unit divides the transmission signal from the near-end speech path including the echo of the reception signal combined by the band combination processing unit into a plurality of bands. Then, the echo canceling means estimates the impulse response of the echo path in the near-end speech path for each band, and generates a pseudo echo according to the estimation result of the received signal and the impulse response decoded by the band division decoding means, and divides the band. The pseudo echo is subtracted from the transmission signal and the residual signal of each band is output. Then, the presence or absence of near-end speech is detected from the level of the transmission signal band-divided by the near-end speech detection means, the level of the residual signal output by the band-division echo cancellation means, and the reception signal level output by the reception signal level conversion means. Then, the impulse response estimation control means stops the estimation of the impulse response of the echo canceling means when the near-end speech detection means detects the near-end speech.

In the fourth aspect of the invention, the separating means separates the received signal from the far-end speech channel, which is highly efficient coded and adaptively multiplexed by adaptive quantization for each of the plurality of bands, into the band division decoding means. Decodes the received signal separated for each band, and outputs the band-divided received signal and the quantization width. Then, the reception signal level conversion means converts the quantization width into a reception signal level. The band synthesis processing means synthesizes the band-divided reception signals output by the band division decoding means and outputs them to the near-end speech path, and the band-division processing means includes the near-end including the echo of the reception signals synthesized by the band synthesis processing means. The transmission signal from the communication path is divided into a plurality of bands. Then, the echo canceling means estimates the impulse response of the echo path in the near-end speech path for each band, and generates the pseudo echo according to the estimation result of the received signal and the impulse response decoded by the band division decoding means, and divides the band. The pseudo echo is subtracted from the transmission signal and the residual signal of each band is output. Then, the band division coding means performs high efficiency coding of the residual signal for each band by adaptive quantization and outputs a quantization width, and the residual signal level conversion means converts the quantization width into a residual signal level. The near-end speech detection means detects the presence or absence of near-end speech from the level of the band-divided transmission signal, the residual signal level converted by the residual signal level conversion means, and the reception signal level output by the reception signal level conversion means, The impulse response estimation control means stops the estimation of the impulse response of the echo canceling means when the near-end speech detection means detects the near-end speech.

In the fifth invention, the received signal band division processing means divides the received signal from the far end speech path into a plurality of bands, and the transmission signal band division processing means echoes the received signal from the far end speech path. The transmission signal from the near-end speech path including is divided into a plurality of bands. Then, the echo canceling means estimates the impulse response of the echo path in the near-end speech path for each band, and generates a pseudo echo according to the estimation result of the received signal and the impulse response decoded by the band division decoding means, and divides the band. The pseudo echo is subtracted from the transmission signal and the residual signal of each band is output. Further, the thinning-out control unit groups a predetermined band among the divided bands in the echo canceling unit into a plurality of band groups, and controls the impulse response estimation to be stopped for each of the plurality of band groups.

[0026]

【Example】

Example 1. Conventionally, the output of the band-divided echo canceller was combined into one original band and then transmitted, but this embodiment does not combine the output of the echo canceller but multiplexes the band-divided signals for transmission and reception. It is intended to synthesize the band on the side. FIG. 1 shows an embodiment of the present invention, in which an echo canceller for each band and a high efficiency encoder for each band are connected. In FIG. 1, 3 is a band division echo cancellation unit, 9 is a band division processing unit, 10 is a band division encoder, and 1
1 is a multiplexer, 12 is a demultiplexer, 13 is a band division decoder,
Reference numeral 14 is a band synthesis processing unit.

Next, the operation will be described. The separation unit 12 separates the reception signal from the transmission line, which is divided into bands, high-efficiency-encoded and then multiplexed, into high-efficiency-encoded reception signals for each band, and the band-division decoder 13 determines each band. The high-efficiency coded received signal is decoded and output to the band division echo cancellation unit 3 and the band synthesis processing unit 14. The band synthesis processing unit 14 synthesizes the decoded received signals of the respective bands and outputs the synthesized signals to the near-end speech path side. The band division processing unit 9 divides the transmission signal from the near-end speech path including the echo of the received signal into a plurality of bands and outputs the band to the band division echo cancellation unit 3, and the band division echo cancellation unit 3 impulse response of the echo path. And a received signal for each band from the band division decoder 13 is generated as a pseudo echo according to the impulse response estimation of the echo path, and the pseudo echo is subtracted from the band-divided transmission signal to obtain the residual signal of each band. The divided encoder 10 is output. The band division encoder 10 performs high-efficiency encoding of the residual signal for each band, and multiplexes the high-efficiency encoded residual signal of each band.
Output to 1. The multiplexing unit 11 multiplexes the high-efficiency-coded residual signals of each band and outputs the multiplexed signals to the far-end speech path side.

Therefore, in this embodiment, the residual signals after echo cancellation that have been band-divided are not combined but multiplexed while being band-divided and separated on the receiving side, so that the band required for generating a pseudo echo on the receiving side. It is not necessary to generate a divided reception signal, that is, comparing this example with the conventional example,
Conventional example The band synthesis processing unit 4 in FIG. 11 corresponds to the band synthesis processing unit 14 in FIG. 1, and the one corresponding to the received signal band division processing unit 1 in FIG. 11 is not included in FIG. Therefore, the present embodiment can reduce the amount of calculation required for the processing of the received signal band division processing unit 1 of the conventional example FIG.

Example 2. This embodiment has the configuration of the first embodiment,
A function to add a near-end call detection function, which seeks to reduce the amount of calculation for detecting the near-end call by obtaining the residual signal level based on the quantization width calculated in high-efficiency coding. Is. FIG. 2 shows the configuration of the signal processing apparatus according to this embodiment. In FIG. 2, reference numeral 10 denotes a band division encoder which performs high efficiency encoding and outputs a quantization width obtained in the encoding process. Reference numeral 15 is a residual signal level conversion unit that converts the quantization width of each band from the band division encoder 10 into a residual signal level. In addition, the band division processing unit 9, the multiplexing unit 11, the separation unit 12, the band division decoder 1
3, the band synthesis processing unit 14 is the same as that of the first embodiment shown in FIG. 1, and the band division echo cancellation unit 3, the reception signal level calculation unit 5, the transmission signal level calculation unit 6, the near-end call detection unit 8, and the control signal 18
12 is similar to that of the conventional example shown in FIG.

Next, the operation of FIG. 2 will be described. Band division processing unit 9, band division encoder 10, multiplexing unit 11, separation unit 1
2, the band division decoder 13 and the band synthesis processing unit 14 perform the same operations as in the case of FIG. 1 shown in the first embodiment. The reception signal level calculation unit 5 inputs the band-divided reception signal output from the band division decoder 13, calculates the signal level of each band, and outputs the calculation result to the near-end speech detection unit 8. The transmission signal level calculation unit 6 inputs the band-divided transmission signal output from the band division processing unit 9, calculates the signal level of each band, and outputs the calculation result to the near-end speech detection unit 8. Next, regarding the residual signal level, the residual signal level calculating unit 7 shown in FIG. 12 of the conventional example is not provided, and instead, the band division encoder 10 and the residual signal level converting unit 15 try to obtain the residual signal level. It is a thing.

The band division encoder 10 performs high efficiency encoding of the residual signal input from the band division echo canceller 3 for each band, and multiplexes the high efficiency encoded residual signal of each band.
1 and outputs the quantization width obtained in the coding process of each band to the residual signal level conversion unit 15. The residual signal level conversion unit 15 uses the quantization width of each band to
The residual signal level after echo cancellation is obtained and output to the near-end call detection unit 8. The near-end speech detection unit 8 determines whether or not there is a near-end speech signal for each band based on the reception signal level, the transmission signal level of each band, and the residual signal level after echo cancellation. When it is determined that the near-end speech signal does not exist in the band, the control signal 18 controls the band-division echo canceller 3 to estimate the impulse response, and otherwise controls to stop.

Next, a method of obtaining the residual signal level based on the quantization width obtained in the process of encoding the residual signal will be described below. FIG. 3 shows an internal configuration diagram of the band division encoder 10. This figure shows the structure of one band of the band division encoder 10, in which 100 is an adaptive quantizer, 101 is a quantizer, 102 is an encoder, and 103 is a quantization width adaptor. The quantization unit 101 inputs the residual signal of the band to be processed from the residual signals after echo cancellation of each band output from the band division echo cancellation unit 3 in FIG. The residual signal is quantized according to the quantization width output by the adaptation unit 103. The encoding unit 102 includes the quantization unit 1
The quantized residual signal is input from 01, and the quantized residual signal is encoded based on the quantization width output from the quantization width adaptation unit 103.

The high-efficiency coded residual signal, that is, the coded code is output to the quantization width adaptation unit 103 and the multiplexing unit 11 in FIG. The quantization width adaptation unit 103 inputs the coding code output from the coding unit 102, performs control such that the quantization width is widened when the coding code becomes large, and narrowed when the coding code becomes small. The quantization width is output to the quantization unit 101 and the encoding unit 102, and the residual signal level conversion unit 15 in FIG.
Output to. In this way, the adaptive quantizer 100 adaptively changes the quantization width according to the level of the residual signal, so that even a signal with a wide dynamic range can be coded without degrading the SN ratio.

Here, it should be noted that the quantization width of the adaptive quantizer 100 changes in accordance with the level of the input signal, which is narrow when the level of the input signal is low and wide when the level of the input signal is high. . Therefore, the adaptive quantizer 1
If the relationship between the quantization width and the input signal level in 00 is obtained in advance, the input signal level, that is, the residual signal level can be obtained from the quantization width. For example, it is assumed that the adaptive quantizer 100 shown in FIG. 3 has a characteristic that the quantization width changes as shown in FIG. 4 according to the input signal level. It can be seen in FIG. 4 that the input signal level is b when the quantization width is a value. Therefore, the residual signal level conversion unit 15 in FIG. 2 stores the relationship between the quantization width and the residual signal level shown in FIG. 4 in the memory in advance, and reads the quantization width as an address to read the residual signal level. Can be asked.

As described above, in this embodiment, since the residual signal level after echo cancellation is calculated without referring to the memory, the signal is sampled at constant time intervals as in the conventional example, and the amplitude is squared. The amount of calculation can be reduced as compared with the method of obtaining the signal level by moving average. It should be noted that although FIG. 3 shows an example in which the quantization width adaptation unit 103 obtains the quantization width from the high-efficiency coded signal, there is a method of adaptively obtaining the quantization width from the residual signal, but the same applies. The residual signal level can be obtained from the quantization width.

Example 3. This embodiment has the configuration of the first embodiment,
A function to detect the near-end call is added, and the amount of calculation for detecting the near-end call is reduced by obtaining the received signal level based on the quantization width calculated in the high-efficiency decoding. . FIG. 5 shows the configuration of the signal processing apparatus according to this embodiment. In the figure, 13 is a band division decoder, which outputs the quantization width obtained in the decoding process. Reference numeral 16 is a reception signal level conversion unit that converts the quantization width from the band division decoder into a reception signal level. The other band division processing unit 9, multiplexing unit 11, demultiplexing unit 12, and band synthesis processing unit 14 are the same as those in FIG. 1 of the first embodiment, and the band division echo cancellation unit 3, transmission signal level calculation unit 6, residual signal. The level calculator 7, the near-end call detector 8, and the control signal 18 are the same as those in FIG.

Next, the operation of FIG. 5 will be described. Band division processing unit 9, band division encoder 10, multiplexing unit 11, separation unit 1
2, the band division decoder 13 and the band synthesis processing unit 14 perform the same operations as in the case of FIG. 1 shown in the first embodiment. Regarding the received signal level, the received signal level calculation unit 5 shown in FIG. 11 of the conventional example is not provided, but instead, the band division decoder 13 and the received signal level conversion unit 16 try to obtain the received signal level. The band division decoder 13 decodes the high-efficiency-coded received signal (encoded code) of each band from the separation unit 12 to obtain a quantization width, and outputs the quantization width to the received signal level conversion unit 16. The reception signal level conversion unit 16 obtains the reception signal level based on the quantization width of each band and outputs it to the near-end speech detection unit 8.

The residual signal level calculation unit 7 inputs the band-divided residual signal output from the band-division echo cancellation unit 3,
The signal level of each band is calculated and the calculation result is output to the near-end speech path detection unit 8. The transmission signal level calculation unit 6 inputs the band-divided transmission signal output from the band division processing unit 9,
The signal level of each band is calculated and the calculation result is output to the near-end call detection unit 8. The near-end speech detection unit 8 determines whether or not there is a near-end speech signal for each band based on the reception signal level, the transmission signal level of each band, and the residual signal level after echo cancellation. When it is determined that the near-end speech signal does not exist in the band, the control signal 18 controls the band-division echo canceling unit 3 to estimate the impulse response, and otherwise controls it to stop.

Next, a method for obtaining the received signal level from the quantization width obtained in the process of decoding the highly efficient coded received signal will be described below. FIG. 6 shows the band division decoder 1
3 shows an internal configuration diagram of No. 3. This figure shows the configuration for one band of the band division decoder 13, where 131 is a quantization width adaptation unit and 132 is a decoding unit. Quantization width adaptation unit 131
Input the coded code of the band to be processed from the high-efficiency coded received signals (coded code) of each band from the separation unit 12 in FIG. The quantization width is calculated under the same control as that of the quantization width adaptation unit 103, and the quantization width is output to the decoding unit 132 and the reception signal level conversion unit 16 in FIG.

The quantization width in the quantization width adaptation unit 131 is wide when the level is high and narrow when the level is low, similar to the quantization width adaptation unit 103 in FIG. Therefore, the reception signal level conversion unit 16 stores the relationship between the quantization width and the reception signal level shown in FIG. 4 in the memory in advance, as in the residual signal level conversion unit 15 in FIG. The reception signal level can be obtained by reading the address.

As described above, according to the present embodiment, the quantization width in the quantization width adaptation unit 131 is obtained by converting it to the received signal level by referring to the memory, so that the signal is sampled at a constant time interval as in the conventional example, The amount of calculation can be reduced by the method of obtaining the signal level by moving average of the square of the amplitude. Note that FIG. 6 shows an example in which the quantization width adaptation unit 131 calculates the quantization width from the received coded code, but a method in which the quantization width is included in the coded code is also sent. The received signal level can be obtained from the quantization width.

Example 4. In this embodiment, the residual signal level is obtained from the quantization width obtained by encoding the residual signal shown in the second embodiment, and the received signal level is obtained from the quantization width obtained by decoding the received signal shown in the third embodiment. Then, the level of the transmission signal is calculated by a conventional method, and the amount of calculation for detecting the near-end call is reduced by detecting the near-end call from these signal levels. FIG. 7 shows the configuration of the signal processing apparatus according to this embodiment. In the figure, reference numeral 10 denotes a band division encoder which performs high efficiency encoding and outputs a quantization width obtained in the encoding process. A band division decoder 13 outputs the quantization width obtained in the decoding process. 1
Reference numeral 5 denotes a residual signal level conversion unit that converts the quantization width of each band from the band division encoder 10 into a residual signal level. 1
Reference numeral 6 is a reception signal level conversion unit that converts the quantization width from the band division decoder into a reception signal level. Other band division processing unit 9, multiplexing unit 11, separation unit 12, band synthesis processing unit 1
4 is the same as that in FIG. 1 of the first embodiment, and the band division echo canceller 3, the transmission signal level calculator 6, the near-end speech detector 8, and the control signal 18 are the same as those in FIG.

Next, the operation of FIG. 7 will be described. Band division processing unit 9, band division encoder 10, multiplexing unit 11, separation unit 1
2, the band division decoder 13 and the band synthesis processing unit 14 perform the same operations as in the case of FIG. 1 shown in the first embodiment. The band division encoder 10 includes a band division echo canceller 3 for each band.
The input residual signal is subjected to high-efficiency coding, the high-efficiency-coded residual signal of each band is output to the multiplexing unit 11, and the quantization width obtained in the coding process of each band is the residual difference. It is output to the signal level conversion unit 15. The residual signal level conversion unit 15 obtains the residual signal level after echo cancellation based on the quantization width of each band, and outputs it to the near-end speech detection unit 8.
The band division decoder 13 decodes the high-efficiency-coded received signal (encoded code) of each band from the separation unit 12 to obtain a quantization width, and outputs the quantization width to the received signal level conversion unit 16.
The reception signal level conversion unit 16 obtains the reception signal level based on the quantization width of each band and outputs it to the near-end speech detection unit 8.

The transmission signal level calculation unit 6 receives the band-divided transmission signal output from the band division processing unit 9, calculates the signal level of each band, and outputs the calculation result to the near-end call detection unit 8. The near-end speech detection unit 8 determines whether or not there is a near-end speech signal for each band based on the reception signal level, the transmission signal level of each band, and the residual signal level after echo cancellation. When it is determined that the near-end speech signal does not exist in the band, the control signal 18 controls the band-division echo canceling unit 3 to estimate the impulse response, and otherwise controls it to stop. As described above, in the present embodiment, the residual signal level is obtained based on the quantization width calculated in the high-efficiency encoding as in the second embodiment, and is obtained in the high-efficiency decoding as in the third embodiment. Since the received signal level is obtained based on the quantization width obtained, the amount of calculation for detecting the near-end call can be reduced. Example 2
Similarly, the method of adaptively obtaining the quantization width from the residual signal can similarly obtain the residual signal level from the quantization width. Further, similarly to the third embodiment, the received signal level can be obtained from the transmitted quantization width even in the method in which the encoded width includes the quantization width and is transmitted.

Example 5. This embodiment is intended to reduce the amount of calculation by thinning out the estimation of the impulse response in a part of the band, and FIG. 8 shows the configuration of the band division echo canceller of the signal processing device. In the figure, 3
Is a band-division echo canceller, which is added with a function of thinning the estimation of the impulse response of a part of the band in the operation of band-division echo cancellation. Reference numeral 17 is a thinning control unit. The reception signal band division processing unit 1, the transmission signal band division processing unit 2, and the band synthesis processing unit 4 are the same as those shown in FIG. 11 of the conventional example.

The operation will be described below. The decimation control unit 17 controls so as to thin out the estimation of the impulse response of a part of the bands at the sampling time intervals of the transmission signal, the reception signal, and the residual signal after band division, that is, at each subband sampling time interval. Therefore, the band division echo canceller 3 is divided into three groups as shown in FIG. That is,
In the figure, 36 is a first echo canceller, 37 is a second echo canceller, and 38 is a third echo canceller. Then, the reception signal band division processing unit 1 and the transmission signal band division processing unit 2 divide the band-divided signals into three groups, respectively, a first echo cancellation unit 36, a second echo cancellation unit 37, and a third echo cancellation unit 37. Output to the echo canceller 38.

Here, for example, the first echo canceller 36
Represents the signals in the I bands, the second echo canceller 37 handles the signals in the J bands, and the third echo canceller 38 handles the signals in the K bands. The total sum is N.
FIG. 9 shows an example in the case of N = 16, I = 8, J = 4, K = 4. The first echo canceling unit 36 always estimates the impulse response, and the second echo canceling unit 37 and the third echo canceling unit 38 follow the control signal from the thinning control unit 17 at subband sample time intervals. The impulse response is estimated alternately. That is, the second echo canceller 37 estimates the impulse response at a certain subband sample time interval, the third echo canceller 38 stops the impulse response estimation, and the second echo canceller 38 stops at the second subband sample time interval. The echo canceller 37 stops estimating the impulse response, and the third echo canceller 38 estimates the impulse response.

Therefore, the impulse response of the echo path is estimated alternately for the I + J and I + K bands at each subband sample time interval. The configuration and operation of the echo canceling section in each band are shown in FIG.
The description is omitted because it is the same as that shown using 3. Also,
The thinning control unit 17 in FIG. 8 controls the operation / stop of impulse response estimation by the control signal 18 in FIG. next,
As a method of grouping each band, the frequency characteristic of a signal to be handled is obtained in advance and grouping suitable for the characteristic is performed. For example, the signal processing device according to the present embodiment is
It is applied to a voice conference system capable of transmitting and receiving a voice signal in a band up to 8 kHz, and the voice signal is divided into 16 bands for processing.

FIG. 10 shows an example in which a spectrum of a general audio signal is shown and grouping of each band is shown in the echo canceller. As shown in the figure, the spectrum of the audio signal tends to have a high level at a low frequency and a low level at a high frequency. Therefore, FIG.
As shown in FIG. 7, the first echo canceller 36 performs the echo canceling for the band having a high signal level, and the second echo canceller 37 and the third echo canceller 38 perform the echo canceling for the other bands. . Then, even if the impulse response estimation speed becomes slower in the high frequency band and the echo cancellation amount becomes smaller than that in the low frequency band, the echo cancellation performance is hardly impaired because the signal strength in the high frequency band is originally small. You can

As described above, according to the present embodiment, impulse response estimation of the entire band, that is, N bands, is not performed for each subband sample time, and I + J or I
Since impulse response estimation of + K bands may be performed,
The amount of calculation required for impulse response estimation can be reduced. It should be noted that the whole band is divided into three groups, and two of them are alternately estimated for impulse response.
The present invention is not limited to this, and the impulse response may be estimated by alternating three or more groups.

[0051]

According to the first aspect of the present invention, the residual signals after echo cancellation, which have been band-divided, are not combined but multiplexed while being band-divided and separated on the receiving side, so that a pseudo echo is generated on the receiving side. It is not necessary to generate a necessary band-divided reception signal, and the amount of calculation required for this processing can be reduced. In the second invention, the residual signal level is obtained based on the quantization width calculated in the high-efficiency coding, so that the amount of calculation for detecting the near-end call can be reduced. In the third invention, since the received signal level is obtained based on the quantization width obtained in the high efficiency decoding, the amount of calculation for detecting the near-end call can be reduced. According to the fourth aspect of the invention, the residual signal level is obtained based on the quantization width calculated in high-efficiency coding, and the received signal level is obtained based on the quantization width obtained in high-efficiency decoding. It is possible to reduce the amount of calculation for detecting end calls. In the fifth invention, the impulse response estimation of the echo canceling section in a part of the band is thinned out, so that the amount of calculation can be reduced.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing a signal processing device according to a first embodiment of the present invention.

FIG. 2 is a configuration diagram showing a signal processing device according to a second embodiment of the present invention.

FIG. 3 is an internal configuration diagram showing processing for one band of the band division encoder.

FIG. 4 is a diagram showing a relationship between an input signal of a quantization width adaptation unit and a quantization width.

FIG. 5 is a configuration diagram showing a signal processing device according to a third embodiment of the present invention.

FIG. 6 is an internal configuration diagram showing processing for one band of the band division decoder.

FIG. 7 is a configuration diagram showing a signal processing device according to a fourth embodiment of the present invention.

FIG. 8 is a configuration diagram showing a band division type echo canceller in a signal processing device according to a fifth embodiment of the present invention.

FIG. 9 is an internal configuration diagram of a band division echo canceller of a band division echo canceller in a signal processing apparatus according to a fifth embodiment of the present invention.

FIG. 10 is a diagram showing a band division method in a signal processing device according to a fifth embodiment of the present invention with respect to a spectrum of a general audio signal.

FIG. 11 is a configuration diagram of a conventional band division type echo canceller.

FIG. 12 is an example of a configuration in which a near-end speech detector is provided in a conventional band division type echo canceller.

FIG. 13 is an internal block diagram of one band of a conventional band division echo canceller.

Claims (5)

[Claims] 1. A signal processing device having the following components. 1. 1. Separation means for separating, for each band, a received signal from the far-end speech path encoded and multiplexed for each of a plurality of bands. 2. Band division decoding means for decoding the received signal separated for each band. 3. Band combining processing means for combining the decoded received signals for each band and outputting to the near-end speech path; 4. Band division processing means for dividing the transmission signal from the near-end speech path including the echo of the reception signal combined by the band combination processing means into a plurality of bands. A pseudo echo is generated by referring to the received signal decoded by the band division decoding means, and the pseudo echo is subtracted from the band divided transmission signal output by the band division processing means to obtain a residual signal for each band. 5. Echo canceling means for outputting, 6. Band division coding means for coding the residual signal for each band, A multiplexing unit that multiplexes the residual signals encoded for each band and outputs the multiplexed residual signals to the far-end speech path. 2. A signal processing device having the following components. 1. 1. Separation means for separating, for each band, a received signal from the far-end speech path that is highly efficient coded and multiplexed by adaptive quantization for each of a plurality of bands. 2. Band division decoding means for decoding the received signal separated for each band. 3. Band combining processing means for combining the decoded received signals for each band and outputting to the near-end speech path; 4. Band division processing means for dividing the transmission signal from the near-end speech path including the echo of the reception signal combined by the band combination processing means into a plurality of bands. The impulse response of the echo path in the near-end speech path is estimated for each band, and a pseudo echo is generated according to the received signal decoded by the band division decoding means and the estimation result of the impulse response, and the band-divided transmission signal is generated. 5. Echo canceling means for subtracting the pseudo echo from and outputting the residual signal in each band. 6. Band division coding means for high-efficiency coding the residual signal for each band by adaptive quantization and outputting a quantization width. 7. Residual signal level converting means for converting the quantization width into a residual signal level, The presence or absence of near-end speech is detected from the level of the transmission signal band-divided by the band division processing means, the level of the reception signal decoded by the band division decoding means, and the residual signal level converted by the residual signal level conversion means. 9. Near-end call detecting means for Impulse response estimation control means for stopping the estimation of the impulse response of the echo canceling means when the near-end speech detection means detects a near-end speech. 3. A signal processing device having the following components. 1. 1. Separation means for separating, for each band, a received signal from the far-end speech path that is highly efficient coded and multiplexed by adaptive quantization for each of a plurality of bands. 2. Band division decoding means for decoding the received signal separated for each band and outputting the band-divided received signal and the quantization width. 3. Received signal level converting means for converting the quantization width into a received signal level, 4. Band combining processing means for combining the band-divided reception signals output by the band division decoding means and outputting to the near-end speech path; 5. Band division processing means for dividing the transmission signal from the near-end speech path including the echo of the reception signal combined by the band combination processing means into a plurality of bands. The impulse response of the echo path in the near-end speech path is estimated for each band, and a pseudo echo is generated according to the received signal decoded by the band division decoding means and the estimation result of the impulse response, and the band-divided transmission signal is generated. 6. Echo canceling means for subtracting the pseudo echo from and outputting the residual signal in each band. Near-end speech detection for detecting the presence or absence of near-end speech from the level of the band-divided transmission signal, the level of the residual signal output by the band-division echo cancellation means, and the reception signal level output by the reception signal level conversion means. Means8. Impulse response estimation control means for stopping the estimation of the impulse response of the echo canceling means when the near-end speech detection means detects a near-end speech. 4. A signal processing device having the following components. 1. 1. Separation means for separating, for each band, a received signal from the far-end speech path that is highly efficient coded and multiplexed by adaptive quantization for each of a plurality of bands. 2. Band division decoding means for decoding the received signal separated for each band and outputting the band-divided received signal and the quantization width. 3. Received signal level converting means for converting the quantization width into a received signal level, 4. Band combining processing means for combining the band-divided reception signals output by the band division decoding means and outputting to the near-end speech path; 5. Band division processing means for dividing the transmission signal from the near-end speech path including the echo of the reception signal combined by the band combination processing means into a plurality of bands. The impulse response of the echo path in the near-end speech path is estimated for each band, and a pseudo echo is generated according to the received signal decoded by the band division decoding means and the estimation result of the impulse response, and the band-divided transmission signal is generated. 6. Echo canceling means for subtracting the pseudo echo from and outputting the residual signal in each band. 7. Band division coding means for high-efficiency coding the residual signal for each band by adaptive quantization and outputting a quantization width. 8. Residual signal level converting means for converting the quantization width into a residual signal level, Near-end speech detection for detecting the presence or absence of near-end speech based on the level of the band-divided transmission signal, the residual signal level converted by the residual signal level conversion means, and the reception signal level output by the reception signal level conversion means. Means, 10. Impulse response estimation control means for stopping the estimation of the impulse response of the echo canceling means when the near-end speech detection means detects a near-end speech. 5. A signal processing device having the following components. 1. 1. Reception signal band division processing means for dividing the reception signal from the far-end speech path into a plurality of bands. 2. Transmission signal band division processing means for dividing a transmission signal from the near-end speech path including echo of a received signal from the far-end speech path into a plurality of bands. The impulse response of the echo path in the near-end speech path is estimated for each band, and a pseudo echo is generated according to the received signal decoded by the band division decoding means and the estimation result of the impulse response, and the band-divided transmission signal is generated. 3. Echo canceling means for subtracting the pseudo echo from and outputting the residual signal in each band. A thinning-out control unit that groups a predetermined band among the divided bands in the echo canceling unit into a plurality of band groups and controls so that the estimation of the impulse response is stopped for each of the plurality of band groups.