KR20050080357A - Noise signal filtering method for acoustic echo canceller - Google Patents

Abstract

The present invention relates to a noise signal filtering method of an acoustic echo canceller, and the conventional center clipping method performs distortion in a signal due to a discontinuity of a signal as a nonlinear operation is performed to remove impulse noise. There was a problem that could adversely affect. In view of the above problems, the present invention is to determine whether the adaptive filter's coefficients are adapted to the acoustic echo canceller and to adapt the coefficients of the adaptive filter. ; Calculating an average value of the current sample and a past predetermined sample to perform median filtering to remove impulse noise; If the adaptation of the adaptive filter is stopped, the noise elimination by median filtering is not performed. Therefore, there is no discontinuity in the signal from which the impulse noise has been removed, and a separate signal level measurement process is unnecessary. It is effective.

Description

NOISE SIGNAL FILTERING METHOD FOR ACOUSTIC ECHO CANCELLER}

The present invention relates to a noise signal filtering method of an acoustic echo canceller, and more particularly, to a noise signal filtering method of an acoustic echo canceller capable of removing an impulse noise generated through an ambient noise signal or an analog / digital conversion.

The acoustic echo canceller has a characteristic that the signal cannot be removed when the signal of the micro-input signal is different from the echo signal in the remaining echo signal after the acoustic echo is effectively removed. In general, this is the impulse noise generated by the ambient noise signal or the analog-to-digital conversion.

In the conventional center clipping method, as shown in FIG. 1 to remove the impulse noise from the residual echo signal remaining after the acoustic echo canceller, an impulse noise is removed by using a center clipping filter that replaces a signal below a predetermined signal level with zero. Remove

FIG. 2 is a diagram illustrating an example of determining a clipping level by measuring a noise level in the center clipping method of FIG. 1. As illustrated in FIG. 2, a clipping level is determined by measuring a noise level of a residual echo signal.

Thereafter, the residual echo signal is passed through the center clipping filter to perform a nonlinear operation to perform clipping as shown in FIG. 3. It can be seen that distortion has occurred in the clipped signal.

The center clipping method has a problem that it is difficult to determine the clipping level according to various signal levels. Normally, it is necessary to estimate the noise level of the input signal to determine the clipping level, and to obtain the average value of the signal.

In addition, since the center clipping method is a nonlinear operation, there is a problem that distortion occurs in the signal due to signal discontinuity. Therefore, a process of artificially adding white noise is necessary for the naturalness of the clipped signal. In this case, by adding noise, the user may feel natural but may adversely affect the clearly transmitted voice signal.

Accordingly, the present invention has been made in view of the above-described problems, and a noise signal filtering method of an acoustic echo canceller which can remove an impulse noise of a residual echo signal by using a median filter which obtains an average of a current sample and a past sample. The purpose is to provide.

In order to achieve the above object, the present invention is to determine whether the adaptive filter coefficient of the acoustic echo canceller is adapted to adapt the coefficient of the adaptive filter. Reading in; Calculating an average value of the current sample and a past predetermined sample to perform median filtering to remove impulse noise; If the coefficient adaptation of the adaptive filter is stopped, it is characterized in that the step of performing the noise cancellation by the median filtering.

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

The noise signal filtering method of the present invention removes the impulse noise of the residual echo signal by using the center clipping method using a median filter instead of the conventional center clipping method.

The following code is the code of the center clipping method proposed in the present invention. updat indicates whether the adaptive filter is adaptive to coefficients, and i is a sample index. e [i] represents the residual echo signal at the output of the acoustic echo canceller.

As you can see in the code, we perform median filtering through three average calculations.

if (updat && i> = 4)

{

e [i-1] = (e [i-2] + e [i]) / 2;

e [i-2] = (e [i-3] + e [i-1]) / 2;

e [i-3] = (e [i-4] + e [i-2]) / 2;

}

According to the present invention, the median filtering is performed while determining whether the adaptive filter of the acoustic echo canceller adapts the coefficients and adapting the coefficients of the adaptive filter. When the number of coefficients is stopped, the noise cancellation by the median filtering is not performed. That is, when the adaptation of the coefficient is stopped due to the near-end talker signal, noise cancellation by median filtering is not performed.

Impulse noise cancellation performs median filtering by calculating the average of the current sample and a few past samples. The code performs median filtering by taking the current sample and the past four samples and calculating three average values.

FIG. 4 is an exemplary diagram illustrating a process of removing impulse noise in the noise signal filtering method according to the present invention. As shown in FIG. 4, when the fourth (e [4]) and fifth (e [5]) samples are input, FIG. Explain how impulse noise is removed

If the second (e [2]) sample is an impulse noise, when the fourth (e [4]) sample is input, median filtering is performed by calculating the average value of three times ((a) of FIG. 4) and the fifth (e [ 5]) Similarly, when the sample is input, the median filtering is performed through three average calculations (FIG. 4 (b)) to remove the impulse noise. It can be seen that the magnitude of the impulse noise is reduced by two median filtering.

FIG. 5 is a waveform diagram illustrating a result of removing the impulse noise of the residual echo signal by the noise signal filtering method of the present invention. As shown in FIG. 5, the result of removing the impulse noise from the residual echo signal will be described. Compared with FIG. 3, the naturalness of the signal can be seen, and the level of the residual echo signal can also be reduced.

As described in detail above, the present invention not only removes impulse noise included in the residual echo signal by performing median filtering through three average calculations, but also reduces background noise and improves call quality by reducing the noise level. It has an effect.

In addition, since the median filtering is used, there is no signal discontinuity and a separate signal level measurement process is unnecessary, thereby removing the impulse noise with a small amount of calculation.

1 is an exemplary view showing the input and output characteristics of the center clipping method for removing the conventional impulse noise.

2 illustrates an example of determining a clipping level by measuring a noise level in the center clipping method of FIG. 1.

3 is an exemplary view illustrating a clipped signal obtained by performing a nonlinear operation using the clipping level defined in FIG. 2.

Figure 4 is an exemplary view illustrating a process of removing the impulse noise in the noise signal filtering method of the present invention.

Figure 5 is a waveform diagram showing the result of removing the impulse noise of the residual echo signal by the noise signal filtering method of the present invention.

Claims (2)

Reading a current sample and a past predetermined sample of the residual echo signal passed through the acoustic echo canceller if the adaptive filter coefficient is determined by determining whether the adaptive filter of the acoustic echo canceller is adapted; Calculating an average value of the current sample and a past predetermined sample to perform median filtering to remove impulse noise; And if the coefficient adaptation of the adaptive filter is stopped, noise canceling by median filtering is not performed. The method of claim 1, wherein the median filtering is performed in place of the center clipping filtering.