JPH04271622A - Echo canceller - Google Patents

Abstract

PURPOSE:To prevent reduction in echo cancelling quantity caused by background noise in the echo canceller in an on-vehicle hand-free telephone set. CONSTITUTION:An echo signal 8 and a pseudo echo signal generated by a digital filter 1 are added by an adder 2 inversely to each other to suppress an echo signal. In this case, a power level of an echo signal component (error signal) remaining in an output of the adder 2 is calculated by an error power calculation circuit 5. A convergence coefficient used to revise the coefficient of a digital filter 1 is controlled adaptively by a coefficient revision circuit 3 in response to the calculated level. Thus, even for a background noise not detected by a double talk detection circuit 4, the conversion coefficient is decreased when the background noise is increased to limit the revision of the tap coefficient. Thus, the disturbance of the tap coefficient going to be converged due to background noise is prevented and stable echo path is estimated and the reduction in the echo cancelling quantity is prevented.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an echo canceller, and more particularly to an echo canceller used in a loudspeaker telephone capable of making hands-free calls in a car telephone.

0002

2. Description of the Related Art FIG. 4 is a block diagram showing an outline of a conventional echo canceller. First, the signal flow will be explained. The level of the received signal is adjusted by the amplifier 49, and the level is adjusted by the A/D converter 4.
8, the analog signal is converted into a digital signal.

The digital signal is input to an ADF (adaptive digital filter) 50, a double talk detection circuit (DTD) 4, and a D/A converter 47. The D/A converter 47 converts the received signal from a digital signal to an analog signal. The received signal output from the D/A converter is sent to the amplifier 46.
The level of the signal is adjusted and input to the speaker 45.

[0004] The voice of the person speaking in the car is input to the microphone 40,
An amplifier 41 adjusts the signal level. The microphone input signal adjusted by the amplifier 41 is converted from an analog signal to a digital signal by the A/D converter 42. The digitally converted microphone input signal is added to the pseudo echo signal generated by the ADF 50 by the adder 2, and the echo signal is suppressed.

[0005] Echo is a signal generated from a speaker.
This is the signal that leaks into the microphone input signal due to acoustic coupling between the speaker and microphone. The output of adder 2 is D/A
A converter 43 converts the digital signal into an analog signal. The analog-converted signal is level-adjusted by an amplifier 44 and transmitted.

The echo canceller is a circuit that suppresses the echo signal by adding to the echo signal an opposite phase signal of the pseudo echo signal generated by the ADF 50 that simulates the echo path from the speaker 45 to the microphone 40.

A signal obtained by suppressing this echo signal will be called an error signal. For example, the ADF 50 updates the filter coefficients using an updating algorithm called the learning identification method, which is expressed by the following equation, so that the error signal is zero, and simulates the echo path.

[0008] Hn+1 = Hn + α * (Xn /Xn Xn T
)*enHere, Hn is the ADF tap coefficient row vector at time n, Xn is the received signal row vector at time n, e
n is the echo cancellation error signal at time n. α is called a convergence coefficient and has a value in the range of 0<α<2. T represents the transpose of the vector.

[0009] This ADF 50, when the signal input to the microphone 40 is only an echo signal (single talk),
It is possible to successively modify the tap coefficients of the filter to generate the same characteristics as the echo path. However, when the signal of the person talking in the car is input to the microphone input signal (double talk state), the signal of the person talking in the car is not erased, and the converged tap coefficient of the ADF 50 is disturbed by the signal.

[0010] In order to solve this problem, a double talk detection circuit 4 is required which controls the updating of the tap coefficients to be stopped when a signal from a person talking in the car is input.

In FIG. 4, the ADF 50 is constituted by an adaptive digital FIR type filter (hereinafter simply abbreviated as FIR). For this FIR, an updating algorithm called the learning identification method as shown in the above-mentioned formula is generally used, and the coefficients are updated at any time during a call period other than the coefficient update prohibition period such as double talk.

In the updating algorithm of the learning identification method, in order to update the filter coefficients so as to make the error signal zero as described above, when the near-end speaker's voice is input, the signal becomes a disturbance; This disturbs the coefficients of a filter that is about to converge.

[0013] Therefore, the double talk detection circuit 4
When the near-end speaker's voice is input, either stop updating the coefficients or update the convergence coefficient α to a sufficiently small value, and perform the FIR.
This prevents the tap coefficients from being disturbed.

In such a conventional echo canceller, when a sufficiently high level signal such as the voice of a near-end speaker is input, the double talk detection circuit 4 is activated and controls the convergence coefficient as described above. There is a function to prevent filter coefficients.

However, sufficient operation of the double talk detection circuit 4 cannot be expected for stationary low-level signals such as background noise. Therefore, there is a drawback that this noise becomes a disturbance and the amount of echo cancellation decreases.

[0016]

OBJECTS OF THE INVENTION An object of the present invention is to provide an echo canceller that can prevent a decrease in the amount of echo cancellation caused by background noise.

[0017]

[Structure of the Invention] The echo canceller according to the present invention includes an FIR type adaptive digital filter that simulates an echo path between a speaker and a microphone, a negative phase signal of a pseudo echo signal generated by the digital filter, and a signal that leaks from the speaker to the microphone. an adding means for adding an echo signal obtained by suppressing the echo signal; a double talk detecting means for detecting a double talk state based on a signal in which the echo signal is suppressed (hereinafter referred to as an error signal) and a received signal; The present invention is characterized in that it includes coefficient updating means for adaptively controlling convergence coefficients used when updating coefficients of the digital filter according to the power level of the error signal at the time of detection.

[0018]

Embodiments Next, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of the present invention, and parts equivalent to those in FIG. 4 are designated by the same reference numerals. Note that FIG. 1 discloses only the digital signal system part.

The received signal 6 is input to the digital filter 1, and a pseudo echo signal is generated by the digital filter 1. This pseudo echo signal and the echo signal from the microphone input 8 are added in opposite phases to each other in the adder 2, and the echo signal is suppressed.

The output of the adder 2 includes an error signal after addition suppression, and this error signal is input to the error signal power calculation circuit 5 and detected.

In the coefficient updating circuit 3, the coefficients of the digital filter 1 are updated and controlled by an updating algorithm called the learning identification method and expressed by the above-mentioned formula so as to make the error signal zero.

[0023] The double talk detection circuit 4 is disclosed in Japanese Patent Application Laid-Open No. 1986-
As disclosed in Japanese Patent No. 56526, an echo signal, an error signal, and a received signal are input, and when the level difference between the received signal and the error signal becomes smaller than a certain threshold, the near-end speaker Assuming that the voice is input, double talk is detected and double talk information "1" is output.

In the case of single talk in which the voice of the near-end speaker is not input, the double talk detection circuit 4 outputs double talk information "0".

The error signal power calculation circuit 5 calculates the power level of the error signal at the time of single talk based on the double talk information from the double talk detection circuit 4. Based on this error signal power level, the value of the convergence coefficient α of the coefficient update circuit 3 is changed as shown in FIG. 2, for example.

If the error signal power level is always high when the near-end speaker's voice is not input, it is considered that the background noise is large. If this error is used to update the tap coefficients,
As in the case where the near-end speaker's voice is input, this results in disturbing the Tabb coefficients that are about to converge.

Therefore, as shown in FIG. 2, by changing the convergence coefficient α according to the error signal power level, when the background noise is large, updating of the tap coefficients is restricted and stable echo paths can be identified. It is.

FIG. 3 is a block diagram showing a specific example of the coefficient updating circuit 3. A register 1 inputs the error signal e from the terminal 10, temporarily stores it in the register 11, and stores the reciprocal of the norm of the received signal (1/Xn Xn T ).
2 is input to the multiplier 13 and the multiplication result (e/Xn
n T ) is temporarily stored in the register 14. The error signal power level is input from the terminal 15 and temporarily stored in the register 16. The value of the convergence coefficient α is referenced from the ROM 17 using the value of the error signal power level stored in the register 16 as an address. The ROM 17 stores the value of the convergence coefficient α corresponding to FIG.

When the double talk information "0" is input to the terminal 18, the logic is inverted by the NOT circuit 19, the gate 19 is opened, and the value of α referred to in the ROM 17 is input to the register 21. When double talk information “1” is input,
The value stored in the register 22 when the gate 21 opens becomes the convergence coefficient α.

In this manner, the convergence coefficient α is selected using the double talk information and is temporarily stored in the register 23. register 2
3 and the value temporarily stored in the register 14 in the multiplier 24, and the result (αe/Xn
temporarily memorize it.

The received signal X is input from the terminal 26 and the RAM 2
7, the received signals for the number N of taps of the digital filter are stored. The latest received signal X1 stored in the RAM is transferred to the register 28. Received signal X1 stored in register 28 and αe/Xn Xn stored in register 25
The value of T is multiplied by the multiplier 29 and the result is stored in the register 3.
Temporarily memorize at 0.

The multiplication result stored in the register 30 is the correction amount of the tap coefficient, and is added to the tap coefficient stored in the RAM 31 to correct the tap coefficient. The tap coefficient H1 stored in the RAM 31 is temporarily stored in the register 32. AL the values stored in register 32 and register 30
Addition is performed using U, and the result is temporarily stored in register 34. The modified tap coefficients stored in the register 34 are stored in the RAM 3.
1 and the tap coefficient H1 is updated.

Similarly, X2 received one sample before the received signal X1 is transferred from the RAM 27 to the register 28, and the multiplier 29 multiplies the values stored in the registers 28 and 25. The tap coefficient H2 is transferred from the RAM 31 to the register 32 and added to the value of the register 30 to correct the tap coefficient. The modified tap coefficient H2 is temporarily stored in the register 34. Save the value of register 34 to RAM
31 to update the tap coefficients. Similarly, N tap coefficients are updated.

[0034]

Effects of the Invention As explained above, the present invention can solve the problem of backgrounds that cannot be detected by a double talk detector by adaptively controlling the convergence coefficients used in updating the adaptive digital filter coefficients according to the error signal power level. Regarding noise, when the background noise becomes large, the convergence coefficient is set to a small value to limit updating of the tap coefficients, thereby preventing the tap coefficients that are about to converge from being disturbed. Therefore, it is possible to estimate a stable echo path, and it is possible to prevent a decrease in the amount of echo cancellation caused by background noise.

[Brief explanation of the drawing]

FIG. 1 is a system block diagram of an embodiment of the present invention.

FIG. 2 is a diagram showing the relationship between error signal power level and convergence coefficient.

FIG. 3 is a block diagram showing a specific example of the coefficient update circuit 3 in FIG. 1;

FIG. 4 is a block diagram of a conventional echo canceller.

[Explanation of symbols]

1 Digital filter 2 Adder 3 Coefficient update circuit 4 Double talk detection circuit 5 Error signal power calculation circuit

Claims (1)

[Claims] 1. An FIR type adaptive digital filter that simulates an echo path between a speaker and a microphone, an inverse phase signal of a pseudo echo signal generated by this digital filter, and an echo signal leaking from the speaker to the microphone are added together. addition means; double talk detection means for detecting a double talk state using a signal in which the echo signal is suppressed (hereinafter referred to as an error signal) and a received signal; An echo canceller comprising: coefficient updating means for adaptively controlling convergence coefficients used when updating coefficients of the digital filter according to a power level.