JPH08163009A - Echo canceller and learning method thereof - Google Patents

Abstract

(57) [Abstract] [Purpose] To provide an echo canceller and a learning method therefor capable of learning at any time even if the talker of the other party does not speak. In an echo canceller for canceling an echo generated when a sound emitted from a speaker 1 returns to a microphone 2 through a space, an adaptive filter 3 simulating a transmission path formed by the microphone 2 including the space from the speaker 1. And a learning signal input terminal 7 for inputting a voice signal other than the reception voice signal as a learning signal. The learning signal input to the learning signal input terminal 7 and the reception voice signal input terminal 4 are provided.
It is equipped with an adder 8 for adding the audio signal input to the output to the adaptive filter 3 and obtaining the difference between the echo audio signal picked up by the microphone 2 and the simulated echo audio signal output by the adaptive filter 3. An echo canceller having a subtractor 6 and a learning method thereof.

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 its learning method, and more particularly to an echo canceller for inputting a voice signal other than a received voice signal as a learning signal and its learning method.

[0002]

2. Description of the Related Art As a conventional example of an echo canceller used in a loudspeaker device having a speaker and a microphone such as a loudspeaker type telephone used in a communication conference, a chirp as a learning signal when a loudspeaker device is installed and adjusted. signal,
Alternatively, there is a method of inputting white noise for learning.
In addition, when learning the echo canceller, there is also a method in which the voice signal of the other party's speaker, that is, the received voice signal is used as a learning signal for learning during a call.

A conventional example of a loudspeaker having an echo canceller will be described with reference to FIG. In FIG. 2, 1 is a speaker, 2 is a microphone, 3 is an adaptive filter, 4
Is a receiving voice signal input terminal, 5 is a transmitting voice signal output terminal,
Reference numeral 6 is a differentiator. The received voice signal is input from the input terminal 4 and is emitted by the speaker 1 in a loud voice. The received voice that has been emitted loud is collected by the microphone 2 as an echo through the space. The adaptive filter 3 is a filter that simulates a transmission path formed by the microphone 2 including the space from the speaker 1. The filter coefficient of the adaptive filter 3 is learned by an adaptive algorithm so that the difference between the echo picked up by the microphone 2 and the output signal of the adaptive filter 3 is obtained by the differentiator 6. By using the adaptive filter 3 learned in this way, the echo at the transmission voice signal output terminal 5 is eliminated.

[0004]

In the above-described embodiment, even if a device for generating a learning signal such as a chirp signal or white noise is specially prepared in a loudspeaker and a learning signal is input. You can always learn the echo canceller. By the way, in the case of a loudspeaker having a learning signal generator in the loudspeaker, the call mode / measurement mode selector switch is set to the call mode during a call. Therefore, when adjusting the echo cancellation amount during a call, the call mode / The adjustment must be made after switching the measurement mode switch to the measurement mode. In this way, the call will be cut off.

The propagation characteristics of the space in the vicinity of the loudspeaker change due to the movement of speakers participating in the communication conference or the temperature change in the operating state of the loudspeaker, and the echo cancellation amount is caused by this change. May decrease. But,
With this method, even if an attempt is made to adjust the echo cancellation amount to the optimum state, this is cut off during a call, so it is not possible to immediately respond to the adjustment of the echo cancellation amount.

In the next embodiment, the adaptive filter 3 is learned and the echo cancellation amount is adjusted at any time when there is a received voice signal even in a use operation state other than when a loudspeaker is installed and adjusted. It can be adjusted and held in an optimum state. However, if the other party's speaker is not transmitting and there is no received voice signal, learning cannot be performed and the amount of echo cancellation decreases.

The present invention provides an echo canceller and a learning method therefor which can be learned even when there is no received voice signal in a telephone operating state other than when a telephone is installed to adjust the echo canceller. To do.

[0008]

In an echo canceller for canceling an echo generated when a sound emitted from a speaker 1 returns to a microphone 2 through a space, a speaker 1
To the learning signal input terminal 7 for inputting a voice signal other than the received voice signal as a learning signal. The learning signal input terminal 7 has an adaptive filter 3 for simulating a transmission path formed by a microphone 2 including a space. The microphone 2 is provided with an adder 8 for adding the learning signal to be input and the voice signal input to the reception voice signal input terminal 4 and outputting the result to the adaptive filter 3.
The echo canceller is provided with the difference device 6 for obtaining the difference between the echo sound signal picked up by and the simulated echo sound signal output from the adaptive filter 3.

Then, an echo canceller using the audio signal output from the electronic device as a learning signal is constructed. In addition, in the learning method of the echo canceller that cancels the echo generated when the sound emitted from the speaker returns to the microphone through the space, the learning method of the echo canceller that inputs a voice signal other than the received voice signal as a learning signal is configured. did.

Further, a learning method of an echo canceller using a voice signal output from an electronic device as a learning signal is constructed.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT An embodiment of the present invention will be described with reference to FIG. 1, 1 is a speaker, 2 is a microphone, 3 is an adaptive filter, 4 is a receiving voice signal input terminal, 5 is a transmitting voice signal output terminal, and 6 is a differentiator. Absent. The present invention has the configuration of FIG.
It is configured by adding a learning signal input terminal 7 and an adder 8.

The echo canceller of the present invention is a voice signal of a voice, a musical sound, or an environmental sound, an off-talk communication sound, or other sound emitted from a television, a radio, a CD player or other electronic equipment for receiving a broadcast, that is, a partner side. An audio signal other than the speaker's audio signal is input to the learning signal input terminal 7 as a learning signal at all times or when the other speaker is not transmitting. Specifically, televisions, radios, CD players, BGMs, and other electronic devices that emit music or environmental sounds are individually equipped with speakers. These electronic devices usually also include an output terminal of a speaker. The voice signals output from these electronic devices are input to the learning signal input terminal 7 as a learning signal at all times or when the other speaker is not transmitting. In this case, the speaker 1 of the loudspeaker can be commonly used as the speaker of the electronic device by collectively connecting the audio signal output terminals of these electronic devices to the learning signal input terminal 7 through an appropriate filter. Therefore, it is not always necessary to separately install a speaker in the electronic device.

The learning signal input to the learning signal input terminal 7 and the voice signal input to the reception voice signal input terminal 4 are added and output by the adder 8. When there is no received voice signal, only the learning signal is output from the adder 8.
The learning signal output from the adder 8 is radiated loudly in the speaker 1. The sound radiated from the speaker 1 goes around to the microphone 2 through the space and picks up the sound as an echo.

The difference between the echo sound signal picked up by the microphone 2 and the simulated echo sound signal output through the adaptive filter 3 which simulates the transmission path including the space is calculated in the differentiator 6. . The adaptive filter 3 is a difference unit 6
The filter coefficient is learned so that the difference becomes zero, and as a result, the echo voice signal at the transmission voice signal output terminal 5 is erased.

As described above, the echo canceller according to the present invention is a sound signal of a sound, a musical sound, an environmental sound, an off-talk communication sound, and other sounds radiated by a television, a radio, a CD player, or other electronic equipment for receiving a broadcast, that is, A voice signal other than the voice signal of the other party's speaker is also used as a learning signal. A voice signal other than the voice signal of the other party speaker is input as a learning signal constantly or when the other party speaker is not transmitting, and the echo canceller learns. Therefore, even when the other party's speaker is not transmitting, the television, radio, or CD that receives the broadcast
There are audio signals of sounds emitted from players and other electronic devices, musical sounds, environmental sounds, off-talk communication sounds and other sounds,
Since these voice signals are input to the learning signal input terminal 7, learning of the echo canceller is always possible without any talk by the talker of the other side even in an operating state other than the installation adjustment state of the loudspeaker. You can

[0016]

As described above, according to the present invention, a voice signal of a sound other than the received voice signal is input as a learning signal at all times or when the other party's speaker is not transmitting the voice signal. Even when the speaker is not transmitting, the echo canceller can be learned at any time, and the echo cancellation amount can always be adjusted and held in a large state.

Further, the speaker of the loudspeaker can be commonly used as the speaker of the electronic device by collectively connecting the audio signal output terminal of the electronic device to the learning signal input terminal 7 through an appropriate filter. Therefore, it is not always necessary to install a speaker for each electronic device,
The number of speakers can be reduced.

[Brief description of drawings]

FIG. 1 is a diagram illustrating an example.

FIG. 2 is a diagram illustrating a conventional example.

[Explanation of symbols]

 1 Speaker 2 Microphone 3 Adaptive filter 4 Received voice signal input terminal 5 Transmitted voice signal output terminal 6 Difference device 7 Learning signal input terminal 8 Adder

Claims (4)

[Claims] 1. An echo canceller for canceling an echo generated when a sound emitted from a speaker returns to a microphone through a space, comprising an adaptive filter simulating a transmission path formed by the microphone including the space from the speaker. , A learning signal input terminal for inputting a speech signal other than the received speech signal as a learning signal is provided, and the learning signal input to the learning signal input terminal and the speech signal input to the reception speech signal input terminal are added to obtain an adaptive filter. An echo canceller, comprising: an adder for outputting to an output, and a differencer for obtaining a difference between an echo sound signal picked up by a microphone and a simulated echo sound signal output from the adaptive filter. 2. The echo canceller according to claim 1, wherein the audio signal output from the electronic device is used as a learning signal. 3. A learning method of an echo canceller for canceling an echo generated when a sound emitted from a speaker returns to a microphone through a space, wherein a voice signal other than a received voice signal is input as a learning signal. How to learn an echo canceller. 4. The learning method of the echo canceller according to claim 3, wherein a voice signal output from the electronic device is used as a learning signal.