JP3507020B2 - Echo suppression method, echo suppression device, and echo suppression program storage medium - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an echo suppression method, an echo suppression device, and an echo suppression program storage medium for suppressing an echo signal in a 2-wire 4-wire conversion system, a voice communication system, and the like.

[0002]

2. Description of the Related Art FIG. 4 shows an example of a voice communication system. Figure 4
, 101 and 103 are microphones for transmission and 10
Reference numerals 2 and 104 are receiver speakers, 105 and 107 are transmitter signal amplifiers, 106 and 108 are receiver signal amplifiers, 109 is a transmission line, 110 is a transmitter, and 111 is a receiver. The transmitted voice uttered by the transmitter 110 includes a transmission microphone 101, a transmission signal amplifier 105, a transmission line 109,
The signal is transmitted to the receiver 111 via the receiver signal amplifier 108 and the receiver speaker 104.

Unlike the conventional telephone call system, this voice call system does not require a handset to be held in the hand, so that it is possible to make a call while working, or to realize a natural face-to-face call. It has advantages and is widely used in teleconferences, videophones, and loudspeakers.

However, the existence of echo has been a problem as a drawback of the above-mentioned voice communication system. That is, FIG.
In, the voice transmitted from the speaker 104 to the receiving side is received by the microphone 103, and the transmission signal amplifier 1
07, the transmission path 109, the reception signal amplifier 106, and the speaker 102, and is reproduced on the transmission side. For the speaker 110, this phenomenon is an echo phenomenon in which the voice uttered by the speaker is reproduced from the speaker 2, and is called an acoustic echo or the like. This echo phenomenon causes a bad influence such as a call trouble or discomfort in a voice call system.

Further, the sound reproduced from the speaker 102 is received by the microphone 101 and forms a closed loop of the signal. When the loop gain is larger than 1, the howling phenomenon occurs and the call becomes impossible. An echo canceller is used to overcome the problems of the voice communication system. The echo canceller is configured with either an adaptive filter unit, a nonlinear echo suppression processing unit, or a combination thereof.

Here, the nonlinear echo suppression processing refers to echo suppression processing other than adaptive filters (linear processing) such as voice switches and center clippers. For details on adaptive filters and nonlinear echo suppression processing, see "Echo Canceller Technology" (Japan Industrial Technology Center, Sho 61) supervised by Shigeo Tsujii. Also, ITU Recommendations P.201 and P20
4, G165, G167, etc., the configuration and the required performance are also presented. By these processes, it is possible to guarantee sufficient call quality in conventional communication conferences, videophones, loudspeakers, etc., which are used in a relatively quiet environment.

However, in recent years, the usage form of the voice call is expanding. For example, hands-free communication in a car in a noisy environment, desktop video conference using a bucket network with a large transmission delay, and remote lectures using a reverberant auditorium are examples. In such a usage environment, it is difficult to guarantee sufficient speech quality by the conventional adaptive filter and nonlinear echo suppression treatment.

Under high noise, adaptive filters generally have difficulty in fully estimating the impulse response of the echo path. Filter coefficient length (tap length) of adaptive filter
Is set based on the reverberation time of the echo path. Therefore, when the reverberation time is long, more tap lengths are required, which causes a decrease in the convergence speed and an increase in the hardware scale of the device. Furthermore, when the transmission delay is large, the echo becomes more audible, and the residual echo that cannot be canceled by the adaptive filter alone deteriorates the speech quality. On the other hand, the nonlinear echo suppression process has an advantage that the echo can be suppressed significantly robustly by controlling the insertion loss.

However, in the case where the transmitted voice exists at the same time as the echo, they are suppressed without distinction, so that there arises a problem that the transmitted voice is distorted or the sound is interrupted. That is, the non-linear echo suppression processing has a problem that it causes deterioration of call quality during two-way simultaneous call (double talk).

To solve the above problems, a frequency domain echo suppression method has been proposed, and the suppression method will be briefly described with reference to FIG. The echo suppression method is the method described in Japanese Patent Application Laid-Open No. 11-331046. In FIG. 5, 201 and 202 are fast Fourier transform units, 203 is a fast inverse Fourier transform unit, 20
Reference numeral 4 is a one-sided speech state detection unit, 205 is an echo path coupling amount calculation unit, 206 is an echo signal power calculation unit, 207 is an echo suppression gain determination unit, and 208 is an echo suppression processing unit. Here, the same parts as those in FIG. 4 are denoted by the same reference numerals.

First, the reception signal x (k) and the transmission signal (echo superimposed signal) y (k) are each subjected to fast Fourier transform to obtain short-time spectra X and Y of the signal. The one-sided utterance state detection unit 204 detects whether or not there is only the utterance signal by using the utterance signal x (k) and the transmission signal y (k). And when the call state is only the receiving signal,
The estimated echo path coupling amount PHe is calculated in the echo path coupling amount calculation circuit 205 from the powers PX and PY (= PE) of the reception signal and the transmission signal (echo signal in this case) converted previously.

Next, in the echo signal power calculation unit 206,
The estimated echo signal power PEe is calculated by multiplying the received signal power PXT by the estimated echo path coupling amount PHe. The determined predicted echo signal power PEe and the short-term spectra X and Y of the reception signal and the transmission signal are input to the echo suppression gain determination unit 207. The echo suppression gain determination unit 207 calculates the echo suppression gain G based on the ratio of the echo signal b (k) superimposed on the transmission signal y (k). The echo suppression unit 208 multiplies the transmission signal (echo superimposed signal) Y by the value to obtain a processed signal Se in which the echo is suppressed, and the fast inverse Fourier transform unit 203 inverse Fourier transforms the value. , The time signal se (k) in which the echo signal b (k) is suppressed and the transmission signal s (k) is emphasized is obtained.

When this method is used, only the echo signal b (k) is suppressed from the transmission signal y (k) on which the echo is superimposed, only the transmission signal s (k) is emphasized, and transmitted to the other party. You can That is, it is possible to suppress only the echo signal b (k) without interrupting the transmission signal s (k) even during double talk, although it is the nonlinear echo suppression process.

[0014]

As described above, the conventional method has the advantage of having sufficient performance when the signal-to-echo ratio is good and the propagation delay of the echo path is small. There are three major problems under various conditions.

First, when the signal-to-echo ratio is bad, the transmitted voice s (k) after processing is distorted and the speech quality deteriorates.
In the echo suppression method described in Japanese Patent Laid-Open No. 11-331046, the echo signal b is added to the suppressed transmission signal s (k).
The echo suppression gain is calculated on the assumption that (k) does not remain at all (set to 0). However, since an error occurs in the estimated value of the echo signal b (k) that is necessary when calculating the echo suppression gain G, it is difficult to accurately obtain the echo suppression gain, and the inaccurate echo suppression gain is inaccurate. Will be superimposed, and the transmitted voice s (k) will be distorted. Therefore, the first problem is how to calculate the echo suppression gain.

Next, the echo suppression method does not consider the propagation delay of the echo path. That is, when used in a usage mode with a large transmission delay, echo suppression cannot be performed. For example, considering the case where the received signal x (k) and the transmitted signal (echo superimposed signal) y (k) are converted into the frequency domain to calculate the echo suppression gain, the propagation delay of the echo path is equal to or longer than the frequency analysis frame length. In this case, at the time of multiplying the calculated echo suppression gain, the echo signal earlier than the signal to be suppressed at that time is superimposed on the transmission signal (echo superimposed signal) y (k). That is, in principle, it is impossible to suppress the echo because a time lag occurs between the processing frames.
Therefore, the second issue is to consider the propagation delay.

Finally, the above echo suppression method requires transmission / reception detection (one-sided speech state detection unit) in order to calculate the echo path coupling amount. This makes the echo suppression gain G inaccurate when the transmission / reception state is erroneously detected,
May cause deterioration of call quality. for that reason,
The third issue is how to calculate the amount of echo path coupling that does not require transmission / reception detection.

The present invention has been made in view of the above circumstances, and in a two-wire to four-wire conversion system, a loudspeaker communication system, etc., a call is made by suppressing an echo signal that causes a cause of a housing and an auditory obstacle. An object is to provide an echo suppression method, an echo suppression device, and an echo suppression program storage medium for improving quality.

[0019]

In order to achieve the above object, the present invention provides an echo component (echo signal b (echo signal b (e) from a picked-up signal (echo superimposed signal y (k)) to a received signal (x (k))). k)), which is an echo suppression method, wherein an echo signal is estimated based on the received signal and the collected sound signal (estimated echo signal be (k)), and the power (Py) of the collected sound signal is calculated. The power value (Py- (Pbe-Pd), that is, P, obtained by subtracting the power (Pbe-Pd in the embodiment) of the echo signal corrected by the masking threshold value.
se + Pd, where Pse is the estimated power of the voice signal on the sender side) is standardized by the power (Py) of the collected sound signal to calculate an echo suppression gain (G), and the calculated echo suppression gain ( G) the collected signal (y (k))
Is multiplied by.

In the echo suppression method described above,
The power ratio between the received signal and the picked-up signal is calculated for each predetermined period, and among the power ratios that are sequentially calculated, the minimum value is held as the echo path coupling amount, and the echo path coupling amount is set to the reception signal. It is characterized in that the echo signal is estimated by multiplication. In the echo suppression method described above, the feature amount of the received signal and the picked-up signal is extracted, the echo path propagation delay amount is estimated using the cross-correlation of the feature amount, and the received signal and the echo path are obtained. A second reception signal is generated from the propagation delay amount, and at least the echo signal is estimated by using the second reception signal instead of the reception signal.

Further, in the echo suppression method described above,
The reception signal and the picked-up signal are any of a waveform coded signal, an analysis and synthesis coded signal, a hybrid coded signal, and a signal converted into a frequency domain. Further, in the echo suppression method described above, when the received signal and the picked-up signal are hybrid-encoded signals, a line spectrum pair coefficient required for hybrid encoding as the feature amount,
At least one of a driving sound source component, a noise component, and a gain is used.

Further, the present invention is an echo suppressor for suppressing an echo component due to a received signal from a collected signal, and an echo signal estimating means for estimating an echo signal based on the received signal and the collected signal. An echo for calculating an echo suppression gain by normalizing a power value obtained by subtracting the power of the echo signal estimated by the echo signal estimating means from the power of the picked-up signal corrected by the masking threshold value with the power of the reception signal. Suppression gain calculation means,
And a multiplying unit that multiplies the collected sound signal by the echo suppression gain calculated by the echo suppression gain calculating unit and outputs the multiplied signal.

The present invention also provides a computer-readable recording medium having recorded therein a program for realizing the method realized by the echo suppression method using a computer.

The above summary of the invention does not enumerate all the elements necessary for the present invention, and any combination of these elements can be implemented.

By the way, the conventional echo suppression gain calculation has been performed by calculating the echo suppression gain G satisfying the following equation (1). Pse = G · Py (1) (Py = Ps + Pb) In the equation (1), Pse, Py, Ps, and Pb are
Processed transmission signal se (k), transmission signal (echo superimposed signal) y (k), desired signal s (k), echo signal b (k)
The power of each is shown. Here, if the echo suppression gain G can be accurately obtained, the echo signal b
(K) can be completely suppressed, and the transmission signal s input by the correspondent
It is possible to take out (k) completely. However, since each signal and the echo path are non-stationary in time, when the echo suppression gain G is calculated using the statistical property, an error always occurs in the estimated echo signal. When the signal-to-echo ratio is good, even if no audible distortion is detected in the processed transmission signal se (k), if the signal-to-echo ratio becomes poor, the distortion is easily detected, resulting in deterioration of speech quality. .

In contrast to the conventional problem as described above, in the present invention, first, the feature amount of the received signal x (k) and the picked-up signal y (k) is extracted, and the echo path is obtained by using the cross-correlation of the feature amount. A propagation delay amount is calculated, and a second reception signal x _r (k) is generated from the echo path propagation delay amount and the reception signal x (k),
The power ratio between the received signal x (k) and the picked-up signal y (k) is calculated for each predetermined period, and the minimum value of the power ratios sequentially calculated is held as the echo path coupling amount, and the echo path coupling amount is held. The echo signal be (k) considering the signal delay is estimated by multiplying the second received signal x _r (k) by, and the estimated echo signal be (k) is further corrected by the masking threshold to obtain the power (Pbe− Pd) is subtracted from the power (Py) of the sound pickup signal, and this power value (Py- (Py)
be-Pd)) is standardized by the power (Py) of the sound pickup signal to calculate the echo suppression gain (G).

Therefore, the echo signal be (k) to be estimated
Becomes a person who compensates for an error such as a signal delay and obtains the echo suppression gain based on this signal, so that the distortion of the echo superimposed signal y (k) in which the echo signal is suppressed can be significantly reduced. Further, by suppressing the sound pickup signal to be equal to or less than the masking threshold, it is possible to suppress the residual echo component to the extent that it cannot be perceptually detected, so that the communication quality can be improved.

[0028]

DETAILED DESCRIPTION OF THE INVENTION An embodiment of the present invention will be described below with reference to the drawings. The signals described below are defined as encoded signals or signals converted into the frequency domain.

<< First Embodiment >> FIG. 1 shows a first embodiment of the present invention.
It is a figure which shows the echo suppression apparatus by embodiment of this. In the figure, first, the voice uttered by the sender is the received signal x
It is input from the transmitting end 1 as (k). Received signal x
(K) is branched into a plurality of paths on the path sent to the receiving end 2 on the receiver side, and is input to the delay unit 303, the echo path propagation delay estimation unit 302, and the echo path 601 respectively. Here, the echo path 601 is an actual acoustic echo path or a line echo path (hybrid circuit).
Is intended. The received signal x (k) has an echo path 60.
Propagation of 1 outputs the echo signal b (k). On the other hand, the voice of the receiver side is input from the transmitting end 2 as the transmission signal s (k).

The echo signal b (k) after propagating through the echo path 601 and the transmission signal s (k) from the receiver side are input to the adder 602, and these signals are superposed by the adder 602 and echo superposed. The signal y (k) becomes the signal y (k) and is input to the echo path propagation delay estimator 302 in the subsequent echo suppressor 301.

The echo path propagation delay estimator 302 determines the echo path 6 from the correlation of the feature amount between the received signal x (k) input earlier and the echo superimposed signal y (k) input this time.
The propagation delay amount of 01 is calculated and output. Hereinafter, the calculation processing of the propagation delay amount performed by the echo path propagation delay estimation unit 302 will be described in detail. First, the echo superimposed signal y (k)
Is a signal in which the echo signal b (k) and the transmission signal s (k) are superimposed by the adder 602. Further, the echo signal b (k) is the same as the reception signal x (k) on the echo path 601. It is a signal derived by propagating. Therefore, the echo signal b (k) retains the feature amount of the reception signal x (k).

The echo path propagation delay estimator 302 estimates the propagation delay of the echo path 601 by calculating the cross-correlation for the feature quantities of the received signal x (k) and the echo superimposed signal y (k). That is, the cross-correlation function r _xy (k) of the received signal x (k) and the echo superimposed signal y (k) is given by the following equation (2) when both are complex variables.

R _xy (k) = (1 / N) Σ {x (p) y (p + k) ^* } (2) (where p = 0, 1, ..., N−1) The above (2) In the formula, y (p + k) ^* is y (p +
k) is a conjugate complex number, and Σ is a total sum when p = 0 to N−1. Echo path propagation delay estimation unit 302
Is the cross-correlation function r _xy (k) obtained by the above equation (2).
The propagation delay amount is calculated from the maximum value of, and the calculated propagation delay amount is output to the delay device 303.

The delay unit 303 creates an FIR filter based on the amount of propagation delay input from the echo path propagation delay estimation unit 302, and the received FIR signal x is received by the created FIR filter.
Pass (k). The reception signal x (k) after passing through the FIR filter becomes the second reception signal x _r (k), which is output to the subsequent echo path coupling amount estimation unit 304 and echo suppression gain calculation unit 305.

The echo path coupling amount estimation unit 304 receives the second received voice signal x _r (k) and the echo superimposed signal y.
(K) is used to generate a second signal for the echo superimposed signal y (k).
By updating the minimum value of the audio signal x _r (k) ratio, the path coupling amount of the echo path 601 (hereinafter referred to as the echo path coupling amount) is calculated and output.

The echo path coupling amount calculation processing performed by the echo path coupling amount estimation unit 304 will be described in detail below.
The echo path coupling amount is a power ratio (Py / Px) between a signal input to the echo path 601 and a signal after propagating through the echo path 601. That is, considering the signal delay, the received signal x _r which is the input signal to the echo path 601.
It is the power ratio of the echo signal b (k) which is the output signal with respect to (k).

However, the echo signal b (k) is independent
Since it is impossible to extract the
2 received signal x_r(K) and echo superimposed signal y (k)
Is used to calculate the echo path coupling amount. That is, a predetermined period
The second received signal x for the echo superimposed signal y (k) for each
_rThe ratio of (k) was calculated, and the ratio acquired this time and the ratio acquired this time were acquired.
The ratio is compared and the smaller one is used as the echo path coupling amount.
That is, the echo path coupling amount estimation unit 304 has acquired
Echo superposed signal y (k) and second received signal x _r(K)
The smallest value in the power ratio of the echo path coupling amount
Hold as. The echo path coupling amount estimation unit 304
Outputs the amount of path coupling to the echo suppression gain calculation unit 305
To do.

In the echo suppression gain calculation section 305, the second
The echo suppression gain G is calculated using the received signal x _r (k), the echo superimposed signal y (k), and the echo path coupling amount input from the echo path coupling amount estimation unit 304 at a predetermined interval. Specifically, the echo superposition signal y is obtained by multiplying the second received signal x _r (k) by the echo path coupling amount.
Estimate the echo signal be (k) superimposed on (k) (for the echo signal b (k) with no error, estimate
> Let the echo signal be (k). Further, an echo suppression gain G for suppressing the estimated echo signal be (k) below a masking threshold formed by other audible signals is calculated, and this echo suppression gain is output to the multiplier 306.

The echo suppression gain calculation processing performed by the echo suppression gain calculation section 305 will be described below. First,
The echo suppression gain calculation unit 305 calculates the echo suppression gain G that satisfies the following expression (3). Pse + Pd = G · Py (where Py = Ps + Pb) (3)

In the above equation (3), Pd is an audible signal other than the echo signal b (k), for example, the receiver side received signal x (k), the receiver side transmitted signal s (k), and the periodic noise. , Shows the masking threshold (level) formed by line noise. The masking threshold Pd can be calculated from the masking threshold due to noise. Since it is impossible to estimate the masking threshold of the transmission signal s (k) for each predetermined period, the long-term characteristic is estimated from the subjective evaluation of the desired echo suppression gain obtained experimentally in advance. . Here, there are various methods for calculating the echo suppression gain G that satisfies the above relational expression, but in the present embodiment, a case will be described in which it is obtained by a solution method based on Wiener filtering.

The above equation (3) can be expressed as the following equation (4): Pse + Pd = G · (Ps + Pb) (4) When the above equation (4) is used as the equation for the echo suppression gain G, G = (Pse + Pd ) / (Ps + Pb) = (Py−Pbe + Pd) / Py (5)

In the above equation (5), Pbe is the power of the estimated echo signal be (k) obtained by multiplying the second received signal x _r (k) by the echo path coupling amount, Pse.
Is the estimated voice signal se obtained by subtracting the power of the above-mentioned estimated echo signal be (k) from the power of the echo superimposed signal y (k).
It is the power of (k). Echo suppression gain calculation unit 305
Is an echo suppression gain G that satisfies the above equation (5),
The obtained echo suppression gain G is output to the multiplier 306.

The multiplier 306 outputs the echo superimposed signal y (k).
The echo signal b (k) of the input echo superimposed signal y (k) is suppressed by multiplying by the echo suppression gain G output from the echo suppression gain calculation unit 305, and the transmission signal input from the transmission end 2 is transmitted. A signal that is as close as possible to the signal s (k) is output to the receiving end 1.

In this embodiment, the echo path propagation delay estimator 303 has shown an example of calculating the propagation delay amount using a general method using a cross-correlation function, but the received signal x (k) It is also possible to calculate the cross-correlation using the other feature amount of the echo signal y (k). For example, a signal encoded by a hybrid encoding method represented by a GELP (Code Excited Linear Prediction) method has a line spectrum pair coefficient, a driving excitation component, a noise component, and a gain feature amount. Therefore, the echo path propagation delay estimation unit 302
Can calculate the propagation delay of the echo path 601 by calculating the cross-correlation of these feature quantities.

Further, the echo suppression gain calculation section 305
The echo suppression gain G can be obtained by applying spectral subtraction, maximum likelihood estimation, Wiener filtering, MMSE, and the like.

Second Embodiment Next, FIG. 2 shows the configuration of an echo suppressor according to a second embodiment of the present invention. Second
The echo suppressor in the above embodiment is a frequency analysis processing type echo suppressor for suppressing acoustic echo or line echo. The echo suppressor according to the second embodiment has almost the same configuration as the echo suppressor according to the above-described first embodiment, except that the echo path coupling amount calculation and the echo suppression gain calculation are performed in the frequency domain. The difference is that the frequency analysis units 401 and 402 are further provided because they are performed using signals. In addition, the frequency analysis units 401 and 40
Due to the provision of 2, the frequency synthesizer 4 is provided at the output end of the signal.
03 is provided.

[0047] According to the second embodiment, the second received signal x _r output from the delay unit 303 (k) is input to a frequency analysis unit 401, wherein the second received signal x _r (k )
Are processed in the frequency domain. Further, the echo superimposed signal y (k) is also processed in the frequency domain by the frequency analysis unit 402. Then, the second received signal x _r ′ (k) converted into the frequency domain and the echo superimposed signal y (k)
Is output to the echo path coupling amount estimation unit 304 and the echo suppression gain calculation unit 305, and the above-described processing is performed in the frequency domain. As a result, the echo suppression device according to the second embodiment can calculate the echo suppression gain with higher accuracy than in the first embodiment, and can improve the call quality.

<< Third Embodiment >> Next, FIG. 3 shows a configuration of an echo suppressor according to a third embodiment of the present invention. Third
The echo suppressor in the above embodiment is substantially the same as the echo suppressor in the above-described first and second embodiments,
The difference is that an echo canceller configured by an adaptive filter is further provided in the preceding stage of the echo suppressor 301 having the same configuration as the first embodiment or the second embodiment. For this reason, the echo suppressor 301 of the present embodiment uses the signal y ′ (k) (transmitted signal) in which the residual echo e (k) superposed to some extent is erased by the adaptive filter instead of the echo superposed signal y (k). s (k) + residual echo e (k))
Is an input signal. In this case, the echo path coupling amount is a coupling amount including the preceding-stage echo canceller (adaptive filter), but no problem occurs in the processing of the echo suppressor.

The program for realizing the function of the echo suppression device according to the first, second and third embodiments shown in FIGS. 1, 2 and 3 is recorded in a computer-readable recording medium. Various processes may be executed by causing a computer system to read and execute the program recorded in this recording medium. The “computer system” mentioned here includes an OS and hardware such as peripheral devices.

The "computer system" is WW.
If the W system is used, the homepage providing environment (or display environment) is also included. The "computer-readable recording medium" means a floppy (registered trademark) disc, a magneto-optical disc, an RO.
M, a portable medium such as a CD-ROM, and a storage device such as a hard disk built in a computer system. Further, the "computer-readable recording medium" is a volatile memory (RAM) inside a computer system which serves as a server or a client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. In addition, those that hold the program for a certain period of time are also included.

Further, the above program may be transmitted from a computer system in which the program is stored in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the "transmission medium" for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line. Further, the program may be a program for realizing some of the functions described above. further,
It may be a so-called difference file (difference program) that can realize the above-mentioned functions in combination with a program already recorded in the computer system.

Although the embodiment of the present invention has been described in detail above with reference to the drawings, the specific structure is not limited to this embodiment, and includes a design etc. within the scope not departing from the gist of the present invention. Be done.

[0053]

As described above, according to the present invention,
Estimate the echo signal based on the received signal and the picked-up signal,
The power value obtained by subtracting the power of the echo signal corrected by the masking threshold from the power of the picked-up signal is normalized by the power of the picked-up signal to calculate the echo suppression gain, and the calculated echo suppression gain is multiplied by the picked-up signal. To do. As described above, in the present invention, in the calculation of the echo suppression gain, the estimated echo signal is masked and used so as to allow the residual echo below the masking threshold value, and therefore the error of the echo suppression gain using the statistical property is used. Can be compensated for, the processing voice distortion can be reduced, and the call quality can be improved.

Further, according to the present invention, the power ratio between the received signal and the picked-up signal is calculated for each predetermined period, and the minimum value of the power ratios successively calculated is held as the echo path coupling amount. The echo signal is estimated by multiplying the received signal by. In this way, by always estimating the echo coupling amount, it becomes unnecessary to detect the transmission / reception, and it is possible to reduce the processing amount and solve the processing defect due to the erroneous detection. As a result, it is possible to eliminate the need for detecting the one-sided speech state, which is conventionally required, and prevent deterioration of the call quality due to erroneous detection.

Further, according to the present invention, the feature amounts of the received signal and the picked-up signal are extracted, the echo path propagation delay amount is estimated by using the cross-correlation between the feature values, and the received signal and the echo path propagation delay amount are calculated. The second received signal is used instead of the received signal for generating the second received signal and at least estimating the echo signal. Thus, by calculating the propagation delay amount of the echo path and using this propagation delay amount for the estimation of the echo suppression gain, it is possible to avoid the time lag between signals due to the propagation delay and to estimate the echo suppression gain. Is possible. As a result, echoes can be suppressed accurately without any time lag between processed frames.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of an echo suppressor according to a first embodiment of the present invention.

FIG. 2 is a diagram showing a configuration of an echo suppressor according to a second embodiment of the present invention.

FIG. 3 is a diagram showing a configuration of an echo suppressor according to a third embodiment of the present invention.

FIG. 4 is a diagram showing an example of a general voice call.

FIG. 5 is a diagram showing a configuration of a conventional echo suppressor.

[Explanation of symbols]

301 Echo suppressor 302 Echo Path Propagation Delay Estimator 303 delay device 304 Echo path coupling amount estimation unit 305 Echo suppression gain calculator 306 multiplier 401, 402 Frequency analysis unit 403 Frequency synthesizer 501 echo canceller 601 echo path 602 adder

─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Kazuhiko Yamamori 2-3-1, Otemachi, Chiyoda-ku, Tokyo Nippon Telegraph and Telephone Corporation (56) Reference JP-A-7-226697 (JP, A) JP JP-A-9-116472 (JP, A) JP-A-9-116615 (JP, A) JP-A-10-150343 (JP, A) JP-A-11-331046 (JP, A) JP-A-2001-95084 (JP, A) JP 2001-94480 (JP, A) JP 2002-152093 (JP, A) JP 2002-237769 (JP, A) JP 2003-284183 (JP, A) Sumiuchi Sakauchi, Masafumi Tanaka, Yutaka Kaneda, “Objective Evaluation of Desired Echo Suppression Based on Auditory Characteristics”, Proceedings of Autumn Meeting of the Acoustical Society of Japan 1999, September 29, 1999, pp. 451-452, (2-5-19) Sakauchi Sakauchi, Shoji Makino, "Analysis of Echo Suppression Required by Frequency Band in Loudspeaker Communication System", Proceedings of the 1996 Autumn Meeting of ASJ I, September 25, 1996, pp. 547-548, (2-7-17) Sumiuchi Sakauchi, Yoichi Haneda, "Examination of Echo Suppression Required by Frequency Band in Loudspeaker Communication System (Part 2)", The Japanese Society of Acoustics 1997 Spring Study Presentation Lectures Proceedings I, March 17, 1997, pp. 623-624, (2-P-25) (58) Fields investigated (Int.Cl. ⁷ , DB name) H04B 3/00 H04B 7/00 H04M 1/60 H04R 3/00 INSPEC (DIALOG) JISST file ( JOIS)

Claims (5)

(57) [Claims] 1. A reverberation suppressing method for suppressing an echo component due to a received signal from a collected signal, wherein an echo signal is estimated based on the received signal and the collected signal, and a masking threshold is calculated from the power of the collected signal. The echo suppression gain is calculated by standardizing the power value obtained by subtracting the power of the echo signal corrected by the power of the collected sound signal, and the calculated sound suppression gain is multiplied by the collected sound signal. Reverb suppression method. 2. A power ratio between the received signal and the picked-up signal is calculated for each predetermined period, and a minimum value of the power ratios calculated successively is held as an echo path coupling amount, and the echo path coupling amount is held. The echo suppression method according to claim 1, wherein the echo signal is estimated by multiplying the received signal by. 3. A feature amount of the received signal and the picked-up signal is extracted, an echo path propagation delay amount is estimated by using a cross-correlation between the feature amounts, and the echo path propagation delay amount is estimated from the received signal and the echo path propagation delay amount. The echo suppression method according to claim 1 or 2, wherein a second received signal is generated and at least the echo signal is estimated by using the second received signal instead of the received signal. . 4. An echo suppressor for suppressing an echo component due to a received signal from a picked-up signal, the echo signal estimating means for estimating an echo signal based on the received signal and the picked-up signal, and a masking threshold value. Echo suppression gain calculation means for calculating an echo suppression gain by normalizing a power value obtained by subtracting the power of the echo signal estimated by the echo signal estimation means from the power of the corrected sound collection signal to calculate the echo suppression gain. And a multiplying unit that multiplies the collected sound signal by the echo suppression gain calculated by the echo suppression gain calculation unit and outputs the multiplied signal. 5. A computer-readable recording medium in which a program for echo suppression for suppressing an echo component due to a received signal from a collected signal is recorded, wherein the program is an echo based on the received signal and the collected signal. A signal value is estimated, a power value obtained by subtracting the power of the echo signal from the power of the picked-up signal corrected by the masking threshold is normalized by the power of the received signal to calculate an echo suppression gain, and the calculated echo suppression A computer-readable recording medium recording a program for causing a computer to execute a multiplication of the collected sound signal by a gain.