JPH06338826A - Sound echo eliminator - Google Patents

Abstract

PURPOSE:To eliminate unstability for the fluctuation of echo route characteristic and to improve followup ability when a corrective operation is performed by switching algorithm to adaptive algorithm with high fluctuation followup ability after quickly eliminating an echo by using fast algorithm in the initial stage of an identification operation. CONSTITUTION:A signal from an input terminal 1 is delayed by a primary delay circuit 11 when the identification operation is started, and the correlation coefficient of a delay signal and a reception signal is calculated by a primary self correlation coefficient calculation circuit 12. The reception signal is subtracted from a signal in which a primary delay signal is multiplied by a primary autocorrelation coefficient by a non-correlation circuit 13, and the coefficient of a variable coefficient filter 9 is corrected sequentially via a coefficient correction circuit 7 so as not to generate an echo component at a transmission output terminal 4. When the fact that the power of an error signal outputted to the terminal 4 is set at zero in a fixed interval is detected by such sequential correction processing, non-correlation processing is interrupted by a convergence decision circuit 14, and the algorithm is switched to the adaptive algorithm employing a learning identification method, then, echo control is performed by continuing the sequential identification processing.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is used in a communication line or a room sound field control device, and removes an echo component in which a signal of a receiving path appears through a transmitting path by using an adaptive algorithm having a faster convergence speed. The present invention relates to an acoustic echo canceller.

[0002]

2. Description of the Related Art Generally, an acoustic echo canceller removes reflection caused by impedance mismatch of a two-wire to four-wire converter in a long distance telephone line using communication hygiene and a submarine cable, and a voice call system such as a video conference system. In the above, it can be roughly classified into one that removes the reverberation due to the acoustic coupling of the speaker's voice, and is composed of a correction amount calculation circuit 7, a variable coefficient filter 3 and a subtraction circuit 5 that generate a pseudo echo. The basic operation of the acoustic echo canceller will be described below.

FIG. 2 is a diagram showing the basic structure of an acoustic echo canceller. The reception signal input terminal 1 is the reception signal output terminal 2
The received signal at the received signal input terminal 1 is branched and supplied to the variable coefficient filter 3 to generate a pseudo echo. The transmission signal from the transmission signal input terminal 4 and the pseudo echo that is the output of the variable coefficient filter 3 are input to the subtraction circuit 5, the echo component in the transmission signal is removed, and the output of the subtraction circuit 5 is the transmission signal. The signal is output to the signal output terminal 6. The output of the transmission signal output terminal 6 and the signal of the reception signal input terminal 1 are input to the correction amount calculation circuit 7, and the filter coefficient of the variable coefficient filter 3 is corrected by the output of the correction amount calculation circuit 7. In the variable coefficient filter 3, the reception signal is the reception signal input register 8
The sum of products of the reception signal of the reception signal input register 8 and the pseudo impulse response of the pseudo impulse response register 9 is obtained by the sum of products circuit 10, and the output of the sum of products circuit 10 is output as a pseudo echo. The reception signal output terminal 2 and the transmission signal input terminal 4 are two lines 4 in the case of a long distance telephone line.
The line converter is connected to a speaker and a microphone in the case of a public telephone system.

Assuming that the signal propagation characteristic of the echo path is linear and represented by an FIR type digital filter, if the impulse response h (t) and the input received signal X (t) are used, the sampling time interval is is T, the echo y _k at time kT is represented by ^{_{y k = h T X k (}} 1). However, T: Transpose of vector.

On the other hand, 'if _k, the estimated value y of y _k' estimates of h at time kT h _k is given by ^{_{y k = h 'k T X}} k (3). In the acoustic echo canceller, when there is a voice signal at the reception signal input terminal 1 and there is no voice signal at the transmission signal input terminal 4 but only an echo exists, the echo elimination operation is performed as an adaptive operation state. A learning identification method is generally adopted for this adaptive operation algorithm. The sequential correction of h ′ _k by the learning identification method is performed by h ′ _{k + 1} = h ′ _k + α (X _k e _k ) / X _k ^T X _k (4) (6) (7). However, e _k = y _k −y ′ _k 0 <α ≦ 1 (5), and e _k is called a residual echo. Such an arithmetic operation is processed in the correction amount arithmetic circuit 7. The contents of the pseudo impulse response register 9 include the variable coefficient h '.
_k is stored.

However, the learning identification method has an extremely slow convergence speed with respect to a colored signal having a very strong correlation such as a voice signal as compared with white noise. The echo duration is comparatively short at several tens of msec like a long-distance telephone line, but in the case of a loudspeaker telephone system, it is about several hundred msec, and the identification convergence speed needs to be ten and several sec. Therefore,
It is considered that the speech signal is made uncorrelated to improve the identification convergence speed. One of these methods is the decorrelation method. FIG. 3 shows a basic configuration diagram of an acoustic echo canceller that employs the correlation canceling method. The present method does not use the received signal as it is for the coefficient correction operation for correcting the coefficient of the variable coefficient filter in the learning identification method, but rather the phase of the received signal at the received signal input terminal 1 and its primary delayed signal 11 The number of relations is calculated by the first-order autocorrelation coefficient calculation circuit 12, the signal obtained by multiplying the first-order delayed signal 11 by the first-order autocorrelation coefficient is subtracted from the received signal by the decorrelation circuit 13, and the decorrelation circuit 1
3 is input, the reception signal and the output error signal of the subtraction circuit 5 are input, and the coefficient of the variable coefficient filter 9 is corrected so that the echo component does not appear at the transmission output terminal 4. The decorrelation adaptive algorithm is as follows.

If the first-order autocorrelation coefficient is c ₁ ^T and is calculated from the input received signal sequence X, then c _1k = X _k ^T X _k-1 / X _k-1 ^T X _k-1 (7). Using this correlation coefficient, the chromaticity of the input received signal is removed.

Z _k = X _k −c _1k X _k−1 (8) The correction amount for updating the coefficient of the impulse response sequence from the decorrelated processed input signal sequence z is as follows. .

DH _k = α (z _k e _k / X _k z _k ) (9) Therefore, the coefficient sequence successively estimated from this is h ′ _{k + 1} = h ′ _k + dH _k ( 10) is obtained. At this time, α is set corresponding to the line state and noise as in the learning identification method. The use of the de-correlation method is expected to improve the convergence speed in the input of a colored signal such as a voice, so for signals with weak correlation such as a white signal, both methods are superior or inferior. There is no. The simulation results of the identification convergence speed for the input signals of the learning identification method and the correlation removal method are shown in FIGS. The vertical axis represents the number of data steps, and the horizontal axis represents the correction error for the echo model as an evaluation function. β is a correlation coefficient of the input signal, which indicates that the value is 0, the non-correlation signal, and the closer to 1 the chromaticity increases. In the uncorrelated signal, that is, in the white signal, the learning identification method and the correlation removal method show the same convergence, but when β = 0.8, the correlation removal method is 2 more than the learning identification method.
To about 3 times faster. From this, it can be seen that the decorrelation method has a better effect on a color signal such as a voice.

[0010]

The adaptive algorithm based on this decorrelation method has a very excellent convergence characteristic even for colored signals. Therefore, it is possible to cope with a system having a long echo path such as a public address telephone system. However, it is stable for a relatively large echo path length such as a large video conference system, but is a system in which the echo path length changes greatly like a personal small video conference system. On the other hand, it is unstable, and there is a problem that it is impossible to completely eliminate echoes, resulting in intermittent echo generation and oscillation. Therefore, it is an object of the present invention to provide an acoustic echo canceller that eliminates the above problems, eliminates instability with respect to variations in echo path characteristics, and improves followability during correction operation.

[0011]

DISCLOSURE OF THE INVENTION The present invention is intended to solve these problems and comprises a reception signal input terminal, a reception signal output terminal, a transmission signal input terminal, and a transmission signal output terminal. , A delay circuit for delaying a signal arriving at the reception signal input terminal, a primary autocorrelation coefficient calculation circuit for calculating a correlation coefficient with the delay signal, and a primary delay signal from the reception signal A decorrelation circuit for subtracting the signal multiplied by the first-order autocorrelation coefficient, a coefficient correction circuit for correcting the coefficient of the variable coefficient filter so that an echo component does not appear at its transmission output terminal, and A pseudo echo generation circuit that generates a pseudo echo signal by inputting the signal at the reception signal input terminal and the correction information, and a calculation time that calculates the difference between the pseudo echo signal and the signal at the transmission input terminal to obtain a residual echo. And a judgment circuit for judging whether or not the difference is at a zero level, and when the judgment circuit detects that the difference is at a zero level, the operation of the decorrelation circuit is interrupted, and when it is not at a zero level, the operation is stopped. An acoustic echo canceller that operates a correlation circuit.

[0012]

With this structure, the difference between the signal at the reception signal input terminal, the pseudo echo signal output by the correction information, and the signal at the transmission signal input terminal is calculated to obtain the residual echo. Processing by the primary autocorrelation coefficient calculation circuit and the decorrelation circuit at the time when it is detected by the judgment circuit for judging whether the difference is zero level that the power of the error signal converges to zero in a certain section. The operation is aborted. That is, the echo can be promptly removed by using the adaptive algorithm having a faster convergence speed in the initial stage of the identification operation. After that, since it is possible to switch to an adaptive algorithm having a high variation tracking property, very high operational stability can be realized by adding a minimum improvement without increasing hardware.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the configuration of the acoustic echo canceller of the present invention.

The acoustic echo canceller of the present invention is shown in FIGS.
The reception signal input terminal 1, the reception signal output terminal 2, the variable coefficient filter 3, the transmission signal input terminal 4, the subtraction circuit 5, and the transmission signal output terminal 6, which are the conventional acoustic echo cancellers described in 1.
Coefficient correction circuit 7, reception signal input register 8, pseudo impulse response register 9, sum of products circuit 10, first-order delay signal 1
1, primary autocorrelation coefficient calculation circuit 12, decorrelation circuit 13
The acoustic echo canceller is composed of the above and a convergence determination circuit 14 which is a feature of the present invention. Figure 1
The operation will be described based on the acoustic echo canceller.

At the start of the identifying operation, the signal arriving at the reception signal input terminal 1 is delayed by the primary delay circuit 11, and the primary autocorrelation coefficient calculation circuit 12 calculates the correlation coefficient between the delayed signal and the received signal.

The correlation coefficient is calculated by the first-order autocorrelation coefficient calculation circuit 12, and the signal obtained by multiplying the first-order delayed signal 11 by the first-order autocorrelation coefficient is subtracted from the reception signal by the decorrelation circuit 13, and the non-correlation is obtained. The output of the digitization circuit 13, the received signal, and the output error signal of the subtraction circuit 5 are input, and the coefficient of the variable coefficient filter 9 is sequentially corrected so that the echo component does not appear at the transmission output terminal 4. When the power of the error signal output to the transmission output terminal 4 is detected to be zero in a certain section by this sequential correction processing, the convergence determination circuit 14 interrupts the decorrelation processing to correct the filter coefficient. Echo control is performed so that the echo component does not appear at the transmission output terminal 4 by switching to the adaptive algorithm that employs the learning identification method that uses the received input signal as it is and continues the sequential identification processing. When the power of the error signal is not zero, the convergence determination circuit 14 operates the decorrelation processing.

[0017]

As described above, the present invention determines whether the difference between the signal at the reception signal input terminal and the pseudo echo signal output by the correction information and the signal at the transmission signal input terminal is zero level. However, when the level is zero, the processing operation of the decorrelation circuit is terminated. When the level is not zero, the decorrelation circuit can be operated, so that the following effect is obtained.

(1) Complete echo control can be performed because the echo can be removed by using an adaptive algorithm having a faster convergence speed in response to a rapidly changing echo path state.

(2) Since the tracking controllability is improved, intermittent echo generation and oscillation can be suppressed, and the stability of the identification operation is improved.

(3) The acoustic echo can be surely removed only by adding the convergence determination circuit without largely changing the configuration of the apparatus, so that the hardware can be greatly simplified and the cost can be reduced.

[Brief description of drawings]

FIG. 1 is a block diagram of an acoustic echo canceller according to the present invention.

FIG. 2 is a diagram showing a basic configuration of an acoustic echo canceller using a conventional general learning identification method.

FIG. 3 is a diagram showing a basic configuration of an acoustic echo canceller using a conventional general correlation canceling method.

FIG. 4 is a diagram showing a simulation result of an identification convergence speed for an input signal of the learning identification method.

FIG. 5 is a diagram showing a simulation result of an identification convergence speed with respect to an input signal of the correlation removal method.

[Explanation of symbols]

 1 reception signal input terminal 2 reception signal output terminal 3 variable coefficient filter 4 transmission signal input terminal 5 subtraction circuit 6 transmission signal output terminal 7 coefficient correction circuit 8 reception signal input register 9 pseudo impulse response register 10 multiply-accumulation circuit 11 first-order delay Signal 12 primary autocorrelation coefficient calculation circuit 13 decorrelation circuit 14 convergence determination circuit

Claims (1)

[Claims] 1. A reception signal input terminal, a reception signal output terminal, a transmission signal input terminal, a transmission signal output terminal, and a delay circuit for delaying a signal arriving at the reception signal input terminal, A primary autocorrelation coefficient calculation circuit for calculating a correlation coefficient with a delayed signal, and a decorrelation circuit for subtracting a signal obtained by multiplying the primary delayed signal by the primary autocorrelation coefficient from the received signal, A coefficient correction circuit for correcting the coefficient of the variable coefficient filter so that an echo component does not appear at the transmission output terminal, and a pseudo echo for generating a pseudo echo signal by inputting the signal at the reception signal input terminal and the correction information. A generation circuit; a calculation circuit that calculates a residual echo by calculating the difference between the pseudo echo signal and the signal at the transmission input terminal; a judgment circuit for judging whether the difference is zero level; Acoustic echo removing apparatus when when it is detected that a zero level by the constant circuit not zero level to interrupt the operation of the decorrelation circuit, characterized in that operating the decorrelation circuit.