KR101083710B1 - Acoustic Echo cancellation apparatus and method using convergence of adaptive filter - Google Patents

Abstract

The present invention relates to an apparatus and method for acoustic echo cancellation using convergence of an adaptive filter to improve the performance of an echo canceller in a simultaneous call section and to reduce power consumption in a section other than the simultaneous call. Simultaneous call detector for determining whether the signal and the near-end caller's voice signal are coming in at the same time to detect whether the call is in the same call section or a single call section; using different filter coefficients according to whether or not the call Adaptive filter for controlling whether and storage; Convergence checker for checking the convergence of the adaptive filter; characterized in that it comprises a.

Description

Acoustic echo cancellation apparatus and method using convergence of adaptive filter

The present invention relates to acoustic echo cancellation, and more particularly, to an apparatus and method for acoustic echo cancellation using convergence of an adaptive filter to improve the performance of the echo canceller in a simultaneous call section and to reduce power consumption in a non-simultaneous call section. will be.

The development of wireless communication environment has increased the use of hands-free telephones.

It is used in various places because of its ease of use, but when used in a narrow space such as a car, acoustic echo occurs and call quality is greatly reduced. For this reason, the echo cancellation device is built in. Due to the light weight and miniaturization of the device, it is impossible to install a large capacity battery. For this reason, it is becoming more important to reduce power consumption.

Acoustic echo refers to a phenomenon in which sound from a speaker is reflected by an object or a wall, and then enters into a microphone and is heard by the other party.

The echo path may change due to time delay and other environmental factors, but the estimated echo signal still has similarity with the far-end signal. Using these characteristics, acoustic echo can be estimated and eliminated.

In other words, the estimated echo signal is generated through a finite impulse response (FIR) filter using a far-end signal similar to the echo signal as a reference signal.

In consideration of computational constraints for real-time systems, an adaptive filter algorithm based on last mean square (LMS) is mainly used for echo path estimation.

In addition, an adaptive filter algorithm based on normalized least mean square (NLMS), which is an algorithm that normalizes the input signal to normalize the input signal size, is used, and shows better performance than the conventional LMS.

Simultaneous call with voice signal and acoustic echo signal of near-end speaker at input signal of microphone is very important factor in acoustic echo canceller.

The adaptive filter estimates the echo path on a sample basis using the far-end speaker signal and the microphone signal. However, due to the voice of the near-end talker, the adaptive filter diverges without estimating the echo path. Therefore, the adaptive filter does not operate during the duration of the simultaneous call, and the echo is removed using the estimated echo path.

 Among the algorithms for detecting simultaneous calls, there are Geigel algorithm that checks the size of the microphone input signal, algorithm that uses the cross-correlation of the microphone input signal and the residual echo signal, and the method of using both methods simultaneously.

Figure 1 shows the basic concept of the acoustic echo generation process and the acoustic echo canceller.

가 스피커를 통해서 출력되고, 반향 경로를 통과하여 음향 반향 신호

가 된다. 이렇게 생성된 반향신호는 근단 신호

와 함께 마이크의 입력 신호로 입력된다.Far end signal

Is output through the speaker, and the acoustic echo signal passes through the echo path.

Becomes The generated echo signal is a near-end signal.

It is input with the microphone's input signal.

Acoustic echo reduces the quality of the call considerably because the far end speaker hears his voice again. Therefore, these acoustic echoes must be eliminated during a call.

Finding the exact starting point of a call is very important in echo cancellation because the call detector misses a call, even if it is a short time, the filter is distorted during that time and is fixed until the call ends.

In general, complex simultaneous call detection algorithms will provide an accurate starting point, but using complex algorithms for real-time systems and in terms of power consumption is not appropriate.

In addition, there is a section in which the voice signal is loud and does not appear suddenly like the impulse signal. In other words, even if a complex algorithm is used, the missed section of the voice signal of the near-end speaker is missed, so that echo is not easily removed from the concurrent call section.

Accordingly, the adaptive filter continuously performs echo estimation and adaptive calculations for echo cancellation. However, if the adaptive filter is sufficiently converged, such an operation is unnecessary and thus power is wasted.

The present invention is to solve the problem of the acoustic echo cancellation of the prior art, the sound using the convergence of the adaptive filter to improve the performance of the echo canceller in the simultaneous call section and to reduce the power consumption in the non-simultaneous call section It is an object of the present invention to provide an echo canceling device and method.

The present invention uses a simple simultaneous call detection algorithm, and considers the simultaneous call time missed by the simultaneous call detector and uses the filter coefficients before distortion to improve the convergence of the adaptive filter to improve the echo cancellation performance in the concurrent call section. It is an object of the present invention to provide an acoustic echo canceling device and method used.

An object of the present invention is to provide an apparatus and method for canceling acoustic echo using convergence of an adaptive filter to reduce an unnecessary operation and power consumption of an entire system by controlling whether an adaptive filter is updated based on a measured ERLE.

The objects of the present invention are not limited to the above-mentioned objects, and other objects that are not mentioned will be clearly understood by those skilled in the art from the following description.

In order to achieve the above object, an acoustic echo cancellation apparatus using convergence of an adaptive filter according to the present invention is an apparatus for removing acoustic echo in a simultaneous call section, and a voice signal of a far-end speaker and a near-end speaker are simultaneously received. Simultaneous call detector for detecting whether the call is a call section or a single call section; using different filter coefficients according to whether or not the simultaneous call, adaptive filter for controlling whether the filter is updated and stored according to the convergence; And a convergence checker for checking the convergence of the adaptive filter.

Here, the convergence checker,

는 마이크의 입력신호,

는

에서 적응 필터로부터의 추정반향 신호를 감산한 오차신호,

는 수렴 검사기의 문턱값,

와

는 망각인자들인 것을 특징으로 한다.Measure Echo Return Loss Enhancement (ERLE) and determine whether the adaptive filter has converged.

Is the input signal from the microphone,

Is

An error signal obtained by subtracting the estimated echo signal from the adaptive filter at,

Is the threshold of the convergence checker,

Wow

Is characterized by oblivion factors.

를 이용하여 반향 신호를 추정하고,동시통화가 아닐 경우에는 적응 필터 계수

를 이용하여 반향을 추정을 하는 것을 특징으로 한다.And the adaptive filter is previously stored filter coefficients in case of simultaneous call detection.

Estimation of the echo signal using, If not in simultaneous call, adaptive filter coefficient

It is characterized by estimating the echo using.

를 적응 필터 계수

로 갱신하고, 아닐 경우에

을 갱신하는 것을 특징으로 한다.And the adaptive filter, if the result of the convergence checker periodically converges the previous filter coefficients

Adaptive filter coefficients

If not, update

It characterized in that for updating.

를 이용하여 반향 신호를 추정하는 단계;동시통화 구간이 아닐 경우에는 적응 필터 계수

를 이용하여 반향을 추정하고, ERLE를 측정하는 단계;상기 측정한 ERLE를 문턱값과 비교하여 문턱값 보다 낮은 경우에는 적응 필터 계수를 갱신하고, 높은 경우에는 갱신을 하지 않고 저장 주기

마다 이전 필터계수

를 저장하는 단계;를 포함하는 것을 특징으로 한다.According to another aspect of the present invention, there is provided a method for canceling acoustic echo using convergence of an adaptive filter. Determining the previous filter coefficients that are previously stored in the simultaneous call interval.

Estimating a reverberation signal using the adaptive signal coefficients;

Estimating reverberation and measuring an ERLE; comparing the measured ERLE with a threshold value to update the adaptive filter coefficient when the threshold value is lower than the threshold value, and if not, performing a storage period without updating.

Previous filter coefficient

Storing; characterized in that it comprises a.

And measuring the ERLE is,

를 이용하여 ERLE(Echo Return Loss Enhancement)를 측정하고, 여기서,

는 마이크의 입력신호,

는

에서 적응 필터로부터의 추정반향 신호를 감산한 오차신호,

는 수렴 검사기의 문턱값,

와

는 망각인자들인 것을 특징으로 한다.

Measure Echo Return Loss Enhancement (ERLE) using

Is the input signal from the microphone,

Is

An error signal obtained by subtracting the estimated echo signal from the adaptive filter at,

Is the threshold of the convergence checker,

Wow

Is characterized by oblivion factors.

The acoustic echo cancellation device and method using the convergence of the adaptive filter according to the present invention has the following effects.

First, the echo cancellation performance is improved by using a simple simultaneous call detection algorithm.

Second, it is possible to improve the performance of the echo canceller in the simultaneous call section and reduce the power consumption in the non-simultaneous call section.

Third, it is possible to reduce unnecessary computation of the entire system by adjusting whether the adaptive filter is updated based on the measured ERLE.

1 is a block diagram showing a general acoustic echo generation and removal concept
2 is a detailed block diagram of an acoustic echo cancellation device according to the present invention.
3 is a flowchart illustrating an audio signal processing process of an acoustic echo canceller according to the present invention.

Hereinafter, a preferred embodiment of an apparatus and method for acoustic echo cancellation using convergence of an adaptive filter according to the present invention will be described in detail.

Features and advantages of the acoustic echo cancellation device and method using the convergence of the adaptive filter according to the present invention will become apparent from the detailed description of each embodiment below.

2 is a detailed configuration diagram of an acoustic echo canceller according to the present invention, and FIG. 3 is a flowchart illustrating an audio signal processing process of the acoustic echo canceller according to the present invention.

The acoustic echo canceller according to the present invention is composed of a voice detector 10, an NLMS adaptive filter 20, a simultaneous call detector 30, and a convergence checker 40 as shown in FIG.

First, the voice detector 10 that receives the far-end signal as an input and measures the power of the far-end signal and judges whether the voice signal of the far-end talker and the near-end talker's voice are simultaneously received when the voice detector 10 determines that the voice section is greater than the threshold value. NLMS adaptive filter which judges whether the call is a call section or a single call section by using different filter coefficients according to whether or not the call is concurrent, and controls whether to update or store the filter according to convergence. 20, and a convergence checker 40 for checking the convergence of the adaptive filter.

The NLMS adaptive filter 20 first obtains the estimated echo using Equation (1).

와

는 각각 적응 필터 계수와 원단 신호로 이루어진 N by 1 행렬이다.here,

Wow

Are N by 1 matrices of adaptive filter coefficients and far-end signals, respectively.

The residual echo signal from which the estimated echo signal is removed from the microphone signal is calculated using Equation 2.

을 갱신하는데 사용된다.The calculated residual echo signal is expressed by Equation 3

It is used to update.

와

는 유클리디안 노옴 (Euclidean norm)이고 수렴 상수 (적응 단계의 크기)이다.

는 수렴 속도와 수렴 이후의 오차의 크기를 결정할 수 있는 중요한 값으로, 적응 필터의 안정성을 위해서

사이의 값을 가진다.here

Wow

Is the Euclidean norm and the convergence constant (the magnitude of the adaptation step).

Is an important value to determine the convergence speed and the amount of error after convergence.

Has a value between

The voice detector 10 receives a far-end signal as an input by applying Equation 4, measures the power of the far-end signal, and determines the voice section when the voice signal is larger than the threshold.

,

와

는 각각 음성검출기의 문턱값, 독립적인 망각인자

들이다.here,

,

Wow

Are the thresholds of the voice detector and the independent forgetting factor, respectively.

admit.

In addition, the simultaneous call detector 30 applies Equation 5 to detect a simultaneous call section in which the echo signal and the near-end signal are simultaneously present in the microphone signal.

와

는 각각 상호 상관법의 문턱값이고 독립적인 망각인자

들이다.here,

Wow

Are the thresholds of cross-correlation respectively and the independent forgetting factor

admit.

The convergence checker 40 serves to determine the convergence state of the NLMS adaptive filter 20.

The present invention is a method of considering the simultaneous call time missed by the simultaneous call detector 30 while using a simple simultaneous call detection algorithm.

를 이용하여 반향신호를 추정한다. 필터 계수가 수렴했는가를 판단하는 것은 수렴 검사기(40)가 수행하며 수학식 6과 같이 ERLE(echo return loss enhancement)를 측정하여 판단할 수 있다.Save the converged filter coefficients, and when the concurrent call starts, the previous filter coefficients

Estimate the echo signal using. The convergence checker 40 determines whether the filter coefficients converged, and can be determined by measuring echo return loss enhancement (ERLE) as shown in Equation (6).

와

는 각각 마이크 신호와 잔여반향 신호로 이루어진 N by 1 행렬이다.here,

Wow

Are N by 1 matrices each consisting of a microphone signal and a residual echo signal.

In general, since the defined ERLE has many calculations and difficult to obtain stable values as shown in Equation 6, the ERLE is measured and the convergence is determined by applying the forgetting factor as shown in Equation 7.

와

는 각각 수렴 검사기(40)의 문턱값이고 망각인자들이다.here,

Wow

Are thresholds and forgetfulness factors of the convergence checker 40, respectively.

경우에 수학식 3의 적응 연산을 더 이상 수행할 필요가 없어지는 것이다.Since continuous filter coefficient storage is a burden in terms of storage space, it is necessary to store it periodically every storage period M. The adaptive filter includes updating the filter every sample. However, when the adaptive filter converges, the continuous update is meaningless. In equation (7)

In this case, it is no longer necessary to perform the adaptive operation of Equation 3.

Therefore, by updating the adaptive filter based on the measured ERLE, unnecessary computation and power consumption of the entire system can be reduced.

를 측정하여(S301), 음성 구간인지를 판단한다.(S302)3 is a flowchart illustrating an operation of an acoustic echo canceller according to an exemplary embodiment of the present invention. First, a far-end signal is received from the voice detector 10 to receive power of a far-end signal.

By measuring (S301), it is determined whether the voice section (S302).

를 측정하여(S303), 반향 신호와 근단 신호가 동시에 존재하는 동시 통화 구간인지를 판단한다.(S304)

In operation S303, it is determined whether the echo signal and the near-end signal exist simultaneously in the same call period.

를 이용하여 반향 신호를 추정한다.(S305)Previously stored filter coefficients in the concurrent call segment

Estimate the echo signal using (S305).

를 이용하여 반향을 추정하고(S306), ERLE를 측정한다.(S307)And adaptive filter coefficients when not in the concurrent call interval.

To estimate the echo using (S306) and to measure the ERLE (S307).

마다(S309)

를 저장한다.(S311)Subsequently, the measured ERLE is compared with the threshold value (S308) and the adaptive filter is updated when the threshold value is lower than the threshold value (S310).

Every time (S309)

Save it. (S311)

The apparatus and method for acoustic echo cancellation using the convergence of the adaptive filter according to the present invention uses a simple simultaneous call detection algorithm and uses the filter coefficient before distortion in consideration of the simultaneous call time missed by the simultaneous call detector. This is to improve the echo cancellation performance in the section.

It will be understood that the present invention is implemented in a modified form without departing from the essential features of the present invention as described above.

Therefore, the described embodiments should be considered in descriptive sense only and not for purposes of limitation, and the scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the equivalent scope are included in the present invention. It should be interpreted.

10. Voice Detector 20. NLMS Adaptive Filter
30. Simultaneous call detector 40. Convergence checker

Claims (6)

In the device for removing the acoustic echo in the simultaneous call interval,
A simultaneous call detector which determines whether the voice signal of the far-end caller and the voice signal of the near-end caller are simultaneously input and detects whether it is a call section or a single call section;
An adaptive filter that uses different filter coefficients according to whether or not a simultaneous call is used, and controls whether or not the filter is updated and stored according to convergence;
And a convergence checker for checking the convergence of the adaptive filter.
The adaptive filter periodically filters previous filter coefficients when the convergence checker results in convergence.

Adaptive filter coefficients

If not, update

Apparatus for acoustic echo cancellation using convergence of the adaptive filter, characterized in that for updating the. The method of claim 1, wherein the convergence checker,

Measure Echo Return Loss Enhancement (ERLE) and determine whether the adaptive filter converges,
here,

Is the input signal from the microphone,

Is

An error signal obtained by subtracting the estimated echo signal from the adaptive filter at,

Is the threshold of the convergence checker,

Wow

Acoustic echo cancellation apparatus using the convergence of the adaptive filter, characterized in that the forgetting factors. The method of claim 1, wherein the adaptive filter,
Previous filter coefficients previously stored in case of simultaneous call detection

To estimate the echo signal,
Adaptive filter coefficients for non-simultaneous calls

Echo cancellation apparatus using the convergence of the adaptive filter, characterized in that for estimating the echo using. delete Receiving a far-end signal from the voice detector and determining whether the voice signal is a voice section;
Determining whether the echo signal and the near-end signal exist at the same time;
Previously stored filter coefficients in the concurrent call segment

Estimating an echo signal using;
Adaptive filter coefficients for non-interval calls

Estimating the echo using and measuring the ERLE;
The measured ERLE is compared with the threshold value and the adaptive filter coefficient is updated when the threshold value is lower than the threshold value.

Previous filter coefficient

Storing the echo reflection method using the convergence of the adaptive filter comprising a. The method of claim 5, wherein measuring the ERLE is,

Measure Echo Return Loss Enhancement using
here,

Is the input signal from the microphone,

Is

An error signal obtained by subtracting the estimated echo signal from the adaptive filter at,

Is the threshold of the convergence checker,

Wow

Acoustic echo cancellation method using the convergence of the adaptive filter, characterized in that the forgetting factors.

Images