JPH0447720A - Echo canceller - Google Patents

Abstract

PURPOSE:To prevent increase in an estimate error of an echo path and increase in a residual echo attended therewith by multiplying a scaling factor with a filter coefficient so as to control the absolute value of the filter coefficient not in excess of the calculated maximum value and multiplying a reciprocal of the scaling factor with a pseudo echo. CONSTITUTION:A coefficient control circuit 4 is provided to fetch a filter coefficient revised by a coefficient revision circuit 3 thereby deciding a maximum value in the absolute values. Then when the maximum value of the absolute values of filter coefficients exceeds the calculation maximum value, a scaling factor is multiplied with the filter coefficient so as to control the filter coefficient not in excess of the calculated maximum value and a reciprocal of the scaling factor is multiplied with a pseudo echo from a FIR filter at multiplication. Thus, the tap coefficient of the FIR filter 1 does not exceed the calculated maximum value by the hardware and the calculation error due to overload is avoided. Thus, the increase in the estimate error of the echo path is prevented and the increase in the residual echo is suppressed.

Description

[Detailed Description of the Invention] [Summary] Regarding echo cancellers applied to video conferences, etc., even if the filter coefficient exceeds the maximum calculation value, the residual echo can be eliminated without deteriorating the echo path estimation accuracy. The purpose is to realize an echo canceller that can suppress the echo from the transmitted signal by estimating the characteristics of the echo path from the received signal and the transmitted signal, generate a pseudo echo, and subtract it from the echo component. an FIR filter that cancels the component, a coefficient memory that provides the FIR filter with filter coefficients corresponding to the characteristics of the echo path, and a coefficient update that updates the filter coefficients of the coefficient memory so that the magnitude of the residual echo is minimized. circuit and
When the maximum value of the absolute values of the filter coefficient exceeds the calculation maximum value, the filter coefficient is multiplied by a scaling value to control so as not to exceed the calculation maximum value, and at the time of the multiplication, the ig F I R filter is and a coefficient control circuit that multiplies the reciprocal of the scaling value by the reciprocal of the scaling value.

[Field of Application in Industry-1-] The present invention relates to an echo canceller, and particularly to an echo canceller applied to video conferences and the like.

Echo cancellers are used in various systems such as video conferences, long-distance telephones, and public address telephones to accurately transmit the near-end speaker's signal to the other party by reducing the echo that goes around from the speaker to the microphone. However, since the echo path from the speaker to the microphone fluctuates, normal operation is required even in such cases.

[Conventional technology]

FIG. 4 shows a conventionally used echo canceller. In this echo canceller, an FIR (finite impulse response) filter 1 receives a received signal Rin
is sampled and taken in, each time the received signal Rin is sampled, it is delayed by each delay element 25, the output is multiplied by the filter coefficient of the stamp 24, and the results of these multiplications are added in an adder 23. By convolution, the characteristics (impulse response) of the echo path from the speaker 11 to the microphone 12 are estimated, and a pseudo echo (echo replica) S' is generated for canceling the echo component S in the microphone human input Sln.

In estimating the characteristics of this echo path, the transmission signal 5out is fed back, the coefficient update circuit 21 updates the optimum filter coefficients, stores them in the coefficient memory 22, reads out the coefficients of each corresponding stamp 24, and stores the coefficients in the coefficient memory 22. 5out is updated by a well-known LMS algorithm so that the echo error component in the transmitted signal 5out is minimized.

Then, the absolute values of the filter coefficients 111 to fin become equal to the absolute value of the impulse response of the echo path in the converged state of the echo canceller, as shown in FIG. 5(a).

[Problem to be solved by the invention]

If we consider the case where the echo canceller is implemented using fixed-point arithmetic hardware, the range of numerical values that can be taken at each step of the calculation is finite, so the filter coefficients are also finite, and the maximum value of the calculation is As shown in Figure 5, it becomes a straight line ■.

Therefore, in a system where the echo path fluctuates, the gain of the echo path is large and the filter coefficients 111 to Iln
As shown in Figure (b), if the absolute value of I exceeds the maximum calculation value I, which is used as the tap coefficient of the filter, the calculation will not be performed correctly and the accuracy of echo path estimation will deteriorate. There was a problem in that the residual echo became large.

Therefore, the present invention provides an echo canceller that can suppress residual echo without deteriorating the echo path estimation accuracy even if the absolute value of the filter coefficient exceeds the maximum value of the calculation. The purpose is to realize this.

[Means to solve the problem]

FIG. 1 shows the principle of the echo canceller according to the present invention. In the present invention, the characteristics of the echo path are estimated from the received signal and the transmitted signal, a pseudo echo is generated, and this is subtracted from the echo component. an FTP filter 1 for canceling the echo component from the transmitted signal, and a coefficient memory 2 for providing the FIR filter 1 with filter coefficients corresponding to the characteristics of the echo path, so that the magnitude of the residual echo is minimized. a coefficient update circuit 3 for updating the filter coefficients in the coefficient memory; and a coefficient update circuit 3 for updating the filter coefficients in the coefficient memory; and a coefficient updating circuit 3 for multiplying the filter coefficients by a scaling value to obtain the calculation maximum value when the maximum value of the absolute values of the filter coefficients exceeds the calculation maximum value. a coefficient control circuit 4 that controls the scaling value so as not to exceed the value, and multiplies the pseudo echo from the FIR filter by the reciprocal of the scaling value during the multiplication;
By providing the following, the problem of -J two layers is solved.

[Function] That is, in the present invention, by newly providing the coefficient control circuit 4, the filter coefficients updated by the coefficient update circuit 3 are taken in, and the maximum value of these absolute values is determined.

Assuming that this maximum value is the filter coefficient 111 as in the example of FIG. 2, this filter coefficient old is the calculation maximum value ■ determined by hardware constraints.

exceeds.

Therefore, the coefficient control circuit 4 determines a scaling coefficient k (k<1) and multiplies each filter coefficient old to In in order to make the filter coefficient old less than or equal to the maximum calculation value. The taps of the FIR filter 1 are updated by giving .

If this is done, the convolution calculation output of the FIR filter 1 will also double and become a small value, so the coefficient control circuit 4 multiplies the output of the FIR filter 1 by the reciprocal of the scaling coefficient 1/k. It's being restored.

As a result, the tap coefficient of FTP filter 1 will not exceed the maximum calculation value on the hardware, and calculation errors due to overload will be eliminated, making it possible to prevent the echo path estimation error from increasing and reduce the residual echo. This means that the increase can be suppressed.

〔Example〕

FIG. 3 shows an embodiment of the echo canceller according to the present invention, and in this embodiment, the coefficient control circuit 4 is
Maximum value update circuit 41 connected to the output of coefficient update circuit 3
A maximum value memory (RAM) 42 stores the value of this maximum value updating circuit 41, and a scaling coefficient @k is given to the multiplication unit 31 from the output of the maximum value memory 42 and a predetermined maximum value, and its reciprocal 1 /k to the multiplication section 32. The rest of the structure is the same as that shown in FIG.

With this configuration, the maximum value update circuit 41 always sequentially compares the filter coefficient output from the coefficient update circuit 3 for each sampling with the value stored in the maximum value memory 42 (which may be zero initially). and set the larger value to the maximum value memory 4
By repeating this process, the maximum value Hmax of the absolute values of the filter coefficients in all samples is stored in the maximum value memory 42.

Therefore, in the scaling calculation circuit 43, using the maximum filter coefficient Hmax from the maximum value memory 42, k
By calculating =L/Hmax, the scaling coefficient and its reciprocal 1/k can be obtained, and are used for calculations in the multipliers 31 and 32, respectively.

〔Effect of the invention〕

As described above, according to the echo canceller according to the present invention, when the maximum value of the filter coefficients to be updated exceeds the maximum calculation value on the hardware, the filter coefficient is multiplied by the scaling value and the maximum value of the calculation is Since the configuration is configured such that the pseudo echo is controlled so as not to exceed , and the pseudo echo is multiplied by the reciprocal of the scaling value, the filter coefficients will not be overloaded, thereby increasing the estimation error of the echo path and the resulting residual echo. It is possible to prevent an increase in

[Brief explanation of drawings]

FIG. 1 is a block diagram of the principle configuration of an echo canceller according to the present invention, FIG. 2 is a diagram for explaining the present invention in detail, and FIG. 3 shows an embodiment of the echo canceller according to the present invention. Block diagram: Figure 4 is a block diagram showing a conventional example; Figure 5 is a diagram showing fluctuations in filter coefficients to be updated;
It is. In FIG. 1, 1...FIR filter, 2...Coefficient memory, 3...Coefficient update circuit, 4...Coefficient control circuit. In the figures, the same reference numerals indicate the same or corresponding parts.

Claims (1)

[Claims] FI that estimates the characteristics of the echo path from the received signal and the transmitted signal, generates a pseudo echo, and subtracts it from the echo component to eliminate the echo component from the transmitted signal.
an R filter (1), a coefficient memory (2) that provides the FIR filter (1) with filter coefficients corresponding to the characteristics of the echo path, and a coefficient memory (2) that applies filter coefficients of the coefficient memory so that the magnitude of the residual echo is minimized. Coefficient update circuit (3
), when the maximum value of the absolute values of the filter coefficient exceeds the calculation maximum value, the filter coefficient is multiplied by a scaling value to control the calculation so that it does not exceed the calculation maximum value, and at the time of the multiplication, the FIR An echo canceller comprising: a coefficient control circuit (4) that multiplies the pseudo echo from the filter by the reciprocal of the scaling value.