KR100400214B1 - Echo removal unit in telephonic communication system and method for updating coefficient of adaptive filter thereby - Google Patents

Abstract

PURPOSE: An echo removal unit in a telephonic communication system and a method for updating a coefficient of an adaptive filter thereby are provided to improve the quality of telephonic communication by adding the echo removal unit to the telephonic communication system. CONSTITUTION: An echo removal unit in a telephonic communication system includes an echo estimator(400), a first noise remover(410), a first echo remover(406), a second echo remover(408), and a second noise remover(404). The echo estimator(400) is used for estimating an echo element of an original speaker by analyzing a voice signal of the original speaker from a microphone. The first noise remover(410) is used for removing noise from the voice signal of the microphone. The first echo remover(406) is used for outputting a filter coefficient update signal to estimate an echo signal of the original speaker. The second echo remover(408) is used for removing the echo element by subtracting an output signal of the echo estimator from an input signal of the microphone. The second noise remover(404) is used for removing a noise element from an output signal of the second echo remover.

Description

Acoustic echo canceller in telephony communication system and a method for coefficient update of adaptive filter.

The present invention relates to an acoustic echo canceller of a telephone communication means having a speakerphone, and more particularly, to an acoustic echo canceller of a telephone communication means that exhibits excellent performance even in a noisy environment and a coefficient of an adaptive filter. It is about how to update.

As an additional device for providing a user's call convenience using a telephone communication means, in particular, a portable telephone communication means, a hand-free mobile phone or a speakerphone used in a conference system has the advantage of providing the user with simplicity of call. Acoustic echo occurs because the far-end speaker's voice signal (the other party's voice) is re-input into the microphone near the far-end speaker. This acoustic echo degrades call quality and, in severe cases, makes the call itself impossible. Therefore, in order to use the speakerphone more efficiently, an acoustic echo canceller for removing acoustic echo must be installed. By the way, when the surrounding environment of the communication place is noisy, the call quality due to the noise is deteriorated, and the noise may cause an abnormality in the operation of the acoustic echo canceller already installed. Therefore, in order to use the speakerphone normally even in a noisy communication environment, it is necessary to remove the noise from the front of the acoustic echo cancellation process. That is, the voice signal of the near-end talker in communication means such as teleconference, hand-free telephone, etc. is transformed by the background noise signal and the acoustic echo signal of the far-end talker from the speaker, It is added to the voice signal and input to the microphone. Therefore, in order to realize a good quality call service, the process of removing the acoustic echo component of the far-end speaker and the noise of the background noise signal must be essential.

The apparatus shown in FIG. 1 includes a double talking detecting means 10, a line echo removing means 12, and an acoustic echo canceling means 14. In the figure, x (n) represents a far-end speaker signal, s (n) represents a near-end speaker signal, and d (n) represents a signal input to a microphone.

The double talking detection means 10 detects the far-end speaker signal (speaker end signal) and the near-end talker signal (microphone end signal) and compares the energy of both ends. As a result of the comparison, the coefficient of the line echo removing means 12 is updated when the near-end speaker family is significantly larger, and the coefficient of the acoustic echo removing means 14 is updated when the far-end speaker is extremely large. At this time, when the energy difference is insignificant, the coefficients of both removing means 12 and 14 are not updated. Here, the line echo is not very fluctuating due to the characteristics of the transmission line, while the acoustic echo is large and varied due to various ambient noises such as human movement, speaker, and microphone movement. There is difficulty.

Therefore, in order to provide a good quality communication service, an acoustic echo canceling means in which an acoustic echo canceling process and a noise canceling process are performed has emerged, and the apparatus shown in FIGS. 2 and 3 is an example.

2 and 3 are views showing a detailed configuration of a conventional acoustic echo removing means, Figure 2 is an example configuration, Figure 3 is another configuration. 2 and 3, reference numerals 142 and 142 'denote acoustic echo estimators, reference numerals 146 and 146' denote noise cancellers, and reference numerals 144 and 144 'are input from a microphone stage. Acoustic echo cancellers 144 and 144 ', respectively, which subtract the acoustic echo signals estimated by the acoustic echo estimators 142 and 142' from the signal d (n) to remove the acoustic echo components of the far-end speaker, respectively. . Also, the acoustic echo estimators 142 and 142 'input the output of the acoustic echo cancellers 144 and 144' to estimate acoustic echo.

In the case illustrated in FIG. 2, the acoustic echo estimator 142 is affected by the noise and the near-end talker because the noise canceller 146 is located at the rear end of the acoustic echo canceller 144. That is, the acoustic echo estimation process is performed in a state where noise is not removed. It is difficult to stop the process of updating the adaptive filter coefficients of the acoustic echo estimator 142 in the presence of the double talk signal. This results in transforming the adaptive filter coefficients of the acoustic echo estimator 142 to a value independent of the echopath characteristics upon double talk. To solve this problem, the apparatus shown in FIG. 3 has appeared. In the apparatus illustrated in FIG. 3, a noise removing process is performed before an acoustic echo estimation process is performed. In this case, it is possible to stop the process of updating the adaptive filter coefficients of the acoustic echo estimator 142 'in the double talk situation. However, there is a disadvantage in that it is not possible to solve the performance degradation problem of the overall echo canceller 142 'due to time-varying disturbing noise generated during the noise removal process. That is, the prior art shown in Figs. 2 and 3 has a simple implementation, but does not show satisfactory results in terms of performance. On the other hand, the process of obtaining the adaptive filter coefficients of the acoustic echo estimators 142 and 142 'uses the normalized least mean square (NLMS) method due to the simplicity of implementation. However, since this method assumes a constant step size, the convergence speed is slow, and in particular, in a place where the acoustic echo environment is rapidly changed (for example, in a car interior), the performance is sharply degraded.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and an object of the present invention is to provide an acoustic echo canceller of a telephone communication means that can be used even in a noisy environment, and to update the adaptive filter coefficient of the acoustic echo canceller accordingly. Another way is to provide a way to do this.

1 is a view showing a basic module of a speakerphone in a general communication means.

2 is a view showing an example of a conventional acoustic echo removing means.

3 is a view showing another example of the conventional acoustic echo canceling means.

4 is a view showing the configuration of the acoustic echo canceling means according to the present invention.

FIG. 5 is a diagram showing the detailed configuration of the step size updating means shown in FIG.

FIG. 6 is a diagram for describing a method of updating coefficients of an adaptive filter in the apparatus illustrated in FIG. 4.

FIG. 7 is a waveform diagram illustrating waveforms of the transmission signal e (n) in which the acoustic echo canceller having the structure of FIGS. 2 to 4 is realized in a plurality of environments.

According to the present invention for achieving the above object, the acoustic echo canceling device of the communication means having a speaker and a microphone, the voice signal of the far-end speaker radiated through the speaker is input through the micro path in the air An acoustic echo estimator for estimating an acoustic echo component of a far-end speaker, a first noise canceller for removing noise from a signal input through the microphone, and the acoustic echo from a signal output from the first noise canceller A first acoustic echo canceller for outputting a signal obtained by subtracting the acoustic echo component estimated by the estimator, that is, a signal for updating a coefficient of a filter for estimating an acoustic echo signal of a far-end speaker, which is input next; A second acoustic echo canceller configured to remove the acoustic echo component by subtracting the signal output from the acoustic echo estimator from the received signal; And a second noise canceller for removing noise components included in the signal output from the second acoustic echo canceller.

In the present invention, it is characterized in that it further comprises a coefficient updating means connected to the rear end of the first acoustic echo canceller, for updating the adaptive filter coefficient of the acoustic echo estimator.

In the present invention, the coefficient updating means includes: a power meter for calculating the power of the transmission signal e (k) and the signal u (k) output from the speaker; the transmission signal e (k) and the speaker A correlation measurer for calculating a correlation with an output signal u (k), a step size estimator for calculating a step size used in an NLMS algorithm using the calculated power value and a correlation degree value; And a coefficient updater for updating the adaptive filter coefficient of the acoustic echo estimator using the estimated step size, the transmission signal, and the microphone input signal.

)을 감산하여 전송신호(e(k))를 만드는 과정;To estimate the acoustic echo component of the current time input from the microphone by adaptively filtering by using a predetermined filter coefficient by collecting the voice signal of the far-end speaker output to the speaker stage according to the present invention for achieving the above another object The method of updating the coefficients of the adaptive filter for (a) the acoustic echo component in the signal input from the microphone (

Subtracting) to create a transmission signal e (k);

와,(b)

Wow,

의해 전송신호(e(k))와 스피커에서 출력되는 신호(u(k))의 각각의 파워(

)(

) 를 산출하는 과정;expression

Power of the transmission signal e (k) and the signal u (k) output from the speaker

) (

Calculating a);

를 이용하여 전송신호(e(k))와 스피커에서 출력되는 신호(u(k))와의 상관도(

)를 계산하는 과정;(c)

The correlation between the transmission signal e (k) and the signal u (k) output from the speaker

) The process of calculating;

에 적용하여 NLMS 알고리즘에서 사용되는 스텝 사이즈를 계산하는 과정;(d) the correlation with the calculated power value

Calculating a step size used in the NLMS algorithm by applying to;

에 의해 적응필터계수를 갱신하는 과정을 포함하는 것이 바람직하다.(e) using the obtained step size, transmission signal and microphone input signal

It is preferable to include the step of updating the adaptive filter coefficient by.

Hereinafter, with reference to the accompanying drawings will be described the present invention in more detail.

4 is a view showing the configuration of the acoustic echo canceling means according to the present invention.

The apparatus shown in FIG. 4 estimates the acoustic echo component of the far-end speaker by estimating the characteristic that the voice signal of the far-end speaker radiated through the speaker stage 412 is input to the microphone end 414 through various paths in the air. Acoustic echo estimator 400, the acoustic echo component estimated by the acoustic echo estimator 400 in the signal output from the first noise canceller 402, the first noise canceller 402 to remove noise from the signal input through the microphone A first acoustic echo canceller 406 for outputting a signal obtained by subtracting a signal, that is, a signal for updating a coefficient of a filter for estimating an acoustic echo signal of a far-end speaker, which is input next, an acoustic echo estimator from a signal input through a microphone The second acoustic echo canceller 408 for removing the acoustic echo component by subtracting the signal output from the 400 and the second noise for removing noise components included in the signal output from the second acoustic echo canceller 408. A remover 404.

Here, the microphone stage 414 is composed of a microphone amplifier and an analog digital converter (ADC), and the speaker stage 412 changes the frequency characteristics of the amplifier stage and the output speech to adjust the magnitude of the output analog and the digital analog converter (DAC). The main consists of a post filter. The acoustic echo estimator 400 includes an adaptive filter, a coefficient storage unit for storing coefficients of the adaptive filter, and an input signal storage unit. Output the estimated acoustic echo component of the current time.

Next, the operation of the acoustic echo canceller shown in FIG. 4 will be described.

In the acoustic echo cancellation apparatus according to the present invention, a signal input through the microphone 414 undergoes a noise removal process before the acoustic echo component removal process so that even if the communication means equipped therewith is in a severe noise environment, a good call quality can be exerted. After the acoustic echo cancellation, the noise cancellation process is performed. That is, when the noise component included in the signal input through the microphone 414 is updated by the first noise canceller 402 to update the adaptive filter of the acoustic echo component estimator 400, the microphone input signal corresponding to the error may be removed. The noise component is reduced, allowing more accurate coefficients of the adaptive filter. In addition, unlike the first noise canceller 410, the second noise canceller 404 for transmission removes noise without changing the tone.

In the apparatus shown in FIG. 4, the acoustic echo canceling means further including a step size updating means 416 inputs a signal output from the first acoustic echo canceller 406 to adjust the coefficient of the adaptive filter of the acoustic echo estimator 400. FIG. It is a device for updating.

FIG. 5 is a diagram showing the detailed configuration of the step size updating means shown in FIG.

The apparatus shown in FIG. 5 outputs from a power meter 416a, a transmission signal e (k) and a speaker which calculates the power of the transmission signal e (k) and the signal u (k) output from the speaker. A correlation measurer 416b that calculates a correlation with the signal u (k) to be used, a step size estimator 416c that calculates a step size used in the NLMS algorithm using the calculated power value and the value of the correlation degree; And a coefficient updater 416d for updating the adaptive filter coefficient of the acoustic echo estimator 400 using the estimated step size, the transmission signal, and the microphone input signal.

FIG. 6 is a diagram for describing a method of updating coefficients of an adaptive filter in the apparatus illustrated in FIG. 4.

)은 수학식 1에 의해 이루어진다.The audio signal of the far-end speaker outputted to the speaker stage is collected and digitally filtered using a predetermined update coefficient to estimate the acoustic echo component of the current time input from the microphone (step 500).

) Is made by Equation 1.

[Equation 1]

으로 입력벡터이고,here,

Input vector

으로 적응필터 계수벡터이며,

Is an adaptive filter coefficient vector

N is the number of taps of the adaptive filter.

)을 감산하여 전송신호(e(k))를 만든다(602 단계).As shown in Equation 2, the acoustic echo component estimated in step 600 in the signal input from the microphone (

) Is subtracted to produce a transmission signal e (k) (step 602).

[Equation 2]

If the correct acoustic echo component is estimated in the 600-step acoustic echo estimation process, the transmission signal corresponds to the voice signal of the near-end talker. That is, when the acoustic echo component is removed from the signal output from the microphone, only the voice signal of the near end speaker is left. This transmission signal is transmitted to the telephone transmitting end to make a telephone call.

On the other hand, the power of the transmission signal e (k) and the signal u (k) output from the speaker is calculated by the following equation (3).

[Equation 3]

Here, lambda is 0.9 for calculating the weighted average.

The correlation between the transmission signal e (k) and the signal u (k) output from the speaker is calculated using Equation 4 below (step 606).

[Equation 4]

The calculated power value and the degree of correlation are calculated using the following equation (5). The step size used in the NLMS algorithm is calculated (step 608). It is proportional to the similarity (correlation value) of the input signal and is inversely proportional to the estimated error magnitude and the input signal magnitude. This means that if the input signal is composed of an acoustic echo signal, the correlation between the transmission signal and the input signal increases, In this case, the convergence speed is improved, and if the size of the transmission signal is larger than the size of the input signal, it means that the acoustic reflection component ratio of the input signal is reduced, thereby reducing the divergence probability due to the estimation error. In addition, during simultaneous calls, the power of the input signal increases and the degree of correlation decreases, so the step size will be smaller. In this way, acoustic echo cancellation can be performed without the assistance of simultaneous call detection stages. In other words, the method proposed by the present invention is similar to that of the conventional call detector used in the conventional method, but the possibility of malfunction is reduced due to the low possibility of malfunction. There is an advantage to be reduced.

[Equation 5]

The adaptive filter coefficient is updated by the following Equation 5 using the step size obtained through Equation 5, the transmission signal and the microphone input signal (step 610).

[Equation 6]

That is, the adaptive filter coefficients are updated by the NLMS algorithm, stored in the filter coefficient storage unit of the acoustic echo estimator, and used to estimate the next acoustic echo component.

The present invention described above can be used in the speakerphone of a general telephone, a hand-free car attached to an automobile, and a conference call system.

Table 1 shows the results of measuring echo return loss enhancement (ERLE) in multiple surroundings of the acoustic echo canceller having the structure of FIGS. 2 to 4, and FIG. 7 shows the acoustic echo canceller having the structure of FIGS. Is a waveform diagram showing waveforms of a transmission signal e (n) realized in a number of environments.

TABLE 1

(Unit dB)

2 structure 3 structure 4 structure Car environment 1 29.24 12.61 31.29 Car environment 2 6.37 0.22 10.62 Car environment 3 -12.84 1.35 9.36 Office environment 1 33.44 14.01 33.58 Office environment 2 5.78 1.14 6.960 Office environment 3 -6.05 -1.74 -0.08

Here, the structure of FIG. 2 performs a noise removing process after performing an echo removing process, and the structure of FIG. 3 performs a conventional echo removing process after performing a noise removing process, and FIG. The configuration of the present invention is shown as a case of performing a noise removal process later.

In addition, vehicle environment 1 and office environment 1 (hereinafter, environment 1) are cases where only far-end speaker signals exist, and environment 2 is where both near-end and far-end speaker signals exist and the filter adaptation process is continued. It shows the ERLE value when both near-end and far-end speaker signals are present and the adaptation process of the filter is stopped. Here, the EREL value is obtained by the following Equation 7, which represents the energy ratio between the echo component and the signal from which the echo component has been removed as a db value. The larger the value, the better the performance of the half-shaped eliminator.

[Equation 7]

As described above, according to the present invention, the acoustic echo canceling device and the coefficient update method of the adaptive filter in the telephone communication system can realize high quality products by providing improved call quality even in a noisy environment.

Claims (4)

In the acoustic echo canceller of the communication means having a speaker and a microphone, An acoustic echo estimator for estimating the acoustic echo component of the far-end speaker by estimating the characteristic that the voice signal of the far-end speaker radiated through the speaker is input to the microphone through various paths in the air; A first noise canceller for removing noise from a signal input through the microphone; A signal obtained by subtracting an acoustic echo component estimated by the acoustic echo estimator from the signal output from the first noise canceller, that is, a signal for updating a coefficient of a filter for estimating an acoustic echo signal of a far-end speaker next input A first acoustic echo canceller; A second acoustic echo canceller configured to remove the acoustic echo component by subtracting the signal output from the acoustic echo estimator from the signal input through the microphone; And And a second noise canceller which removes a noise component included in the signal output from the second acoustic echo canceller. 2. The acoustic echo canceling apparatus according to claim 1, further comprising coefficient updating means connected to a rear end of the first acoustic echo canceller for updating the adaptive filter coefficient of the acoustic echo estimator. 3. The apparatus of claim 2, wherein the coefficient updating means comprises: a power meter for calculating the power of the transmission signal e (k) and the signal u (k) output from the speaker; A correlation measurer for calculating a correlation between the transmission signal e (k) and the signal u (k) output from the speaker; A step size estimator for calculating a step size used in an NLMS algorithm using the calculated power value and a correlation degree value; And And an coefficient estimator for updating the adaptive filter coefficient of the acoustic echo estimator using the estimated step size, the transmission signal, and the microphone input signal. In the method of updating the coefficient of the adaptive filter for collecting the sound signal of the far-end speaker output to the speaker stage and adaptively filtering using a predetermined filter coefficient to estimate the acoustic echo component of the current time input from the microphone , (a) the acoustic echo component in the signal input from the microphone

Subtracting) to create a transmission signal e (k); (b)

Wow, expression

Power of the transmission signal e (k) and the signal u (k) output from the speaker

) (

Calculating a); (c)

The correlation between the transmission signal e (k) and the signal u (k) output from the speaker

) The process of calculating; (d) the correlation with the calculated power value

Calculating a step size used in the NLMS algorithm by applying to; (e) using the obtained step size, transmission signal and microphone input signal

The adaptive filter coefficient updating method comprising the step of updating the adaptive filter coefficient by.

Images