JP2020155793A - Echo canceller and IP phone - Google Patents

Abstract

To provide a technique capable of preventing deterioration of call quality by blocking learning for echo cancellation with voices (silence, hold sound, etc.) other than a voice to be originally cancelled by a line echo canceller.SOLUTION: An echo canceller including a line echo canceller having a function of detecting a line echo and canceling the line echo, is provided with voice detection means for detecting an input sound from a microphone of the other party and/or a hold sound. When the voice detection means detects silence and/or a hold sound, the echo canceller stops echo cancellation learning in the line echo canceller. An IP telephone includes the echo canceller.SELECTED DRAWING: Figure 1

Description

The present invention relates to an echo canceller and an IP telephone, and more particularly to an IP telephone including an echo canceller having a function of removing line echo and an echo canceller having a function of removing line echo.

As a technique for canceling an acoustic echo, Japanese Patent Application Laid-Open No. 10-41860 (Patent Document 1) is available. In this publication, "an analog subtractor that outputs a transmission signal by subtracting a pseudo echo signal from an analog input transmission signal on which an acoustic echo is superimposed, and a first A / that digitally converts the output of the analog subtractor. Residual from the D converter, the second A / D converter that digitally converts the analog input received signal from the other party, and the digital outputs of the first A / D converter and the second A / D converter. A tap coefficient calculator that estimates the impulse response of the echo and calculates the tap coefficient, and a plurality of tap outputs obtained by delaying the analog input / reception signal with a shift register are weighted by the tap coefficient obtained from the tap coefficient calculator. There is a description of "an echo canceller characterized by being provided with a pseudo echo generator that inputs a post-added weighted composite output as the pseudo echo to the analog subtractor."
Further, it is known that IP telephones generate line echo due to voice "delay" or "fluctuation".

Japanese Unexamined Patent Publication No. 10-41860

Conventionally, the line echo provided in a telephone is a connection portion between a 2-wire line that transmits and receives on two lines and a 4-wire line that uses a separate line for transmission and reception, that is, a 2-wire-4 line conversion part (hybrid). It occurs in the transformer). In other words, this line is also used for the line of the line with the doorphone, and a hybrid coil is used for the 2-wire-4 wire conversion part (hybrid transformer) at the connection between the 2-wire line and the 4-wire line. Be done. Therefore, a part of the received signal (audio signal) leaks into the transmission path due to the impedance mismatch at the connection point between the 2-wire line and the 4-wire line. The reflection caused by this leakage is called echo, and the caller talks while listening to his / her own echo with a delay, which makes it difficult to talk and the call does not proceed smoothly.
In particular, in the case of an IP telephone, voice is communicated by packetized voice data, and the delay is larger than that of an analog telephone. Further, when a VPN connection is made using an IP line, there is a problem that the delay becomes larger and the call cannot be made smoothly.

Conventionally, an echo suppressor has been used as in Patent Document 1, for example, to solve the problem of echo. However, since the echo suppressor suppresses the echo by the switch operation, a part of the voice of the other party (the other party) may be interrupted or the echo may leak, and deterioration in terms of call quality is unavoidable. .. Therefore, in order to solve these drawbacks in principle, an echo canceller has been used in recent years.
However, when the echo canceller is used, the line echo canceller should learn from the original voice to be canceled depending on the environment of the other party and the operation, but sounds other than the voice to be canceled, such as silence and hold sound, etc. In this state, a learning operation for echo cancellation may be performed, and in this case, residual echo is generated from the echo canceller, and there is a new problem that deterioration of call quality is unavoidable. A new solution is needed for that.

Patent Document 1 does not consider any of the above-mentioned problems in the line echo canceller. Therefore, the effect is limited to "reducing the circuit scale for generating a pseudo-echo with a large amount of arithmetic processing for canceling an acoustic echo having a long impulse response time, shortening the processing time, and reducing power consumption".

Therefore, an object of the present invention is to provide a technique capable of preventing echo cancellation learning for sounds other than the original voice to be echo-canceled (silence, hold sound, etc.) and preventing deterioration of call quality.

In order to solve the above problems, one of the representative echo cancellers and IP telephones of the present invention is
When the other party (speaking party) puts the input sound from the microphone into a silent state and / or a holding sound state, a means for stopping the echo cancellation learning in the line echo canceller is provided.

According to the present invention, as a line echo canceller, the echo canceller learning is performed for the voice originally desired to be canceled, and the echo canceller learning is stopped for other voices to maximize the characteristics of the echo canceller. It can be pulled out and the residual echo can be reduced.

The block diagram which shows the structural example (Example 1) of the echo canceller in the IP telephone of this invention. The block diagram which shows the other configuration example (Example 2) of the echo canceller in the IP telephone of this invention. The block diagram which shows the other configuration example (Example 3) of the echo canceller in the IP telephone of this invention.

Hereinafter, examples will be described with reference to the drawings.

FIG. 1 is a block diagram showing a configuration example of an echo canceller in the IP telephone of the present invention.

When operating the echo canceller, if the other party (talking party) mutes the input sound from the microphone or closes the hole in the microphone, during that time, the receiving system transmission path (hereinafter referred to as the receiving path) The received signal is in a "silent" state, that is, the sound input from the microphone is silent. In this case, the echo canceller performs echo canceller learning on the return voice (line echo) in which the received signal returns from the speaker to the microphone. In this "silence" state, when a silent value 0 (filter coefficient) is input to the echo canceller, the adaptive filter in the echo canceller learning for echo cancellation learns for incorrect echo cancellation. Will be done. Then, the residual echo that could not be removed by the adaptive filter of the echo canceller is output, and this residual echo leaks to the receiving path, and as a result, there is a problem that the characteristics of the echo canceller 1020 deteriorate.

In view of the above problems, the present invention considers the echo in the line echo canceling circuit 1620 of the echo canceller 1020 when the line echo canceling circuit 1620 cancels the line echo when a “silent” value is input from the earpiece. It stops learning for cancellation. Hereinafter, an IP telephone will be described as an example.

The IP telephone 1 includes a microphone 11, an A / D converter 12, a D / A converter 19, a speaker 21, a communication unit 14, and an echo canceller 1020. Although the IP telephone 1 is not shown, the IP telephone 1 includes an IP telephone processing unit, a main control unit, an interface unit, and the like that constitute the IP telephone 1, and since these are well known, detailed description thereof will be omitted.

The A / D converter 12 converts an utterance signal of a transmission system transmission line (hereinafter referred to as a transmission line), that is, an analog audio signal which is a microphone input signal from the microphone 11, into a digital audio signal, and echo canceller 1020. Output to.

The D / A converter 19 converts the received signal on the receiving path, that is, the digital audio signal which is the received signal received from the IP network of the audio communication system via the communication unit 14, into an analog audio signal, and converts it into an analog audio signal to the speaker 21. Output.

The microphone 11, the A / D converter 12, the arithmetic processing unit 13 of the echo canceller 1020, and the communication unit 14 form a transmission system transmission line, that is, a transmission path. Then, the utterance signal (microphone input signal) of the microphone 11 is transmitted to the voice communication system (IP network) via the A / D converter 12, the arithmetic processing unit 13 of the echo canceller 1020, and the communication unit 14.

The communication unit 14, the arithmetic processing unit 20 of the echo canceller 1020, and the D / A converter 19 form a reception system transmission line, that is, a reception path. Then, the received signal (microphone input signal of the communication partner) received by the communication unit 14 is output to the speaker 21 via the communication unit 14, the arithmetic processing unit 20 of the echo canceller 1020, and the D / A converter 19. ..

The echo canceller 1020 has a function of performing an echo canceller process, and is an acoustic echo canceling circuit 1315 that cancels the acoustic echo from the microphone 11, a line echo canceling circuit 1620 that cancels the line echo in the receiving path generated at the communication destination, and the receiving receiver. It has a voice detection / control circuit 1718 that controls the line echo canceling circuit 1620 to stop learning when the voice of the road is detected and, for example, a silent state in which the voice cannot be detected from the other party is detected.
The acoustic echo is a phenomenon that occurs when the output signal from the speaker 21 returns to the microphone 11. The line echo is an echo in which the microphone input signal from the microphone 11 returns via the speaker and the microphone of the other party, and this echo is a phenomenon caused by "delay" or "fluctuation" of voice.

The acoustic echo canceling circuit 1315 includes an acoustic echo canceller unit 15 and an arithmetic processing unit 13.

The acoustic echo canceller unit 15 is composed of, for example, a well-known adaptive filter, and is an acoustic echo based on an output signal (microphone input signal a) from the arithmetic processing unit 13 and an output signal (received audio signal) from the arithmetic processing unit 20. A signal a'(estimated acoustic echo component) corresponding to the acoustic echo for canceling is generated, and the estimated acoustic echo component is output to the arithmetic processing unit 13.

The arithmetic processing unit 13 includes a subtractor, and is a signal a'(a microphone input signal a including an acoustic echo component) from the A / D converter 12 and a signal a'(a) corresponding to the acoustic echo from the acoustic echo canceller unit 15. The estimated acoustic echo component) is subtracted (microphone input signal a-signal a corresponding to acoustic echo) to remove the acoustic echo component from the microphone input signal a, and the microphone input signal from the microphone 11 is sent to the communication unit 14 and the acoustic. Output to the echo canceller unit 15 and the line echo canceller unit 16.

The line echo canceling circuit 1620 includes a line echo canceller unit 16 and an arithmetic processing unit 20.

The line echo canceller unit 16 has a function of performing learning for echo cancellation. For example, the line echo canceller unit 16 is composed of a well-known adaptive filter, and is composed of an output audio signal (microphone input audio signal) of the arithmetic processing unit 13 and an output of the level determination unit 17. Based on the signal (level determination result), a signal b'(estimated line echo component) corresponding to the line echo for canceling the line echo is generated, and the estimated line echo component is output to the arithmetic processing unit 20.
The filter coefficient of the adaptive filter represents the transfer of the path, is called the transfer function, and is expressed as a function such as H (n).

The arithmetic processing unit 20 includes a subtractor, and is a digital audio signal from the other party received via the communication unit 14 (a reception signal b including a line echo component and a signal b corresponding to a line echo from the line echo canceller unit 16). '(Estimated echo component)) is subtracted (digital audio signal b-signal b corresponding to line echo) to remove the line echo component from the received signal b, that is, echo canceller processing is performed to obtain the received signal. Output to the D / A converter 19 and the acoustic echo canceller unit 15.

The voice detection / control circuit 1718 has a level determination unit 17 that detects the voice input from the microphone of the other party and determines the level of the detected voice. The level determination unit 17 detects the received signal (audio signal including the line echo) in the earpiece of the communication unit 14 and determines the level, that is, the level of the received signal (audio signal) in the earpiece is preset. It is determined whether or not the level is below a predetermined level, and the result is output to the line echo canceller unit 16.

Here, when the level of the received signal (voice signal) in the receiving path is equal to or lower than a predetermined level, the level determination unit 17 determines that the state is "silent".
At this time, the line echo canceller unit 16 receives the determination result of the level determination unit 17, and if it is in the “silence” state, the line echo canceller unit 16 stops learning for echo cancellation. In other cases, normal line echo canceller learning is performed.

According to the above-described embodiment, as an echo canceller, learn for the original voice to be canceled, in other words, do not learn for echo cancellation for the voice (silence state) that is not the voice to be canceled. By doing so, the characteristics of the echo canceller can be maximized, and the residual echo output from the line echo canceller 1620 can be reduced.

FIG. 2 is a block diagram showing another configuration example of the echo canceller in the IP telephone of the present invention.

In this embodiment, the following problems are corrected when the other party (the other party to call) is put on hold during a call.
When the other party (the other party to call) is put on hold during a call, the IP telephone 1 receives the hold tone from the other party.
At this time, similarly to the first embodiment, the echo canceller 1020 performs a canceller process of canceling the return voice (echo) in which the microphone voice of the speaker returns to the microphone 11 via the speaker 21. At this time, if a silent numerical value 0 (filter coefficient) is input to the line echo canceller unit 16 of the echo canceller 1020, the adaptive filter of the line echo canceller unit 16 learning for echo cancellation is incorrect as described above. Learn to cancel echoes. Therefore, the characteristics of the echo canceller 1020 (line echo canceller unit 16) deteriorate.

In order to correct such a problem, in this embodiment, the voice detection / control circuit 1718 detects the hold sound from the hold sound generation unit (not shown) of the other party, for example, music, in addition to the level determination unit 17. When the hold sound (music) detection unit 18'is provided and the hold sound (music) detection unit 18' detects the hold sound (music), the line echo canceller unit 16 stops learning for echo cancellation. ..

As a result, the same effect as in the first embodiment can be expected even when the other party is put on hold during the call.

FIG. 3 is a block diagram showing another configuration example of the echo canceller in the IP telephone of the present invention.

This embodiment corrects the following problems when the input voice from the speaker is small during a call.
When the input voice of the speaker input from the microphone 11 during a call, that is, the voice returned via the IP telephone of the other party is small, the fluctuation of the voice data as a parameter is small, so it is useful for learning for echo cancellation. It takes time. As a correction method, it is conceivable to provide a means for controlling the input voice so that the volume becomes a certain level or higher, or to automatically adjust it by using an automatic gain control circuit (AGC) or the like.
However, in these methods, there is a problem that even the inflection part of the voice is corrected. Therefore, with respect to the learning for echo cancellation in the line echo cancellation circuit 1620 of the echo canceller 1020, there is a new problem that the echo return is heard due to the delay in the learning for echo cancellation.

In order to correct the problem, in the present embodiment, when the receiving path is in a silent state and / or the input voice from the speaker is small during a call, the time including the time determination unit 18 for measuring and determining the time thereof is included. A measurement / determination circuit 1800 is provided.

The time determination unit 18 determines a received signal (voice signal) in the earpiece of the communication unit 14, that is, a voice time in which the voice from the speaker is smaller than a predetermined value set in advance during a call, and the result is a line. Output to the echo canceller unit 16.

At this time, if the speaker of the IP telephone 1 to be used can be identified by one person, the line echo canceller unit 16 holds the voice signal (voice data) at the time of the last call in a built-in memory (not shown) or the like. I will do it.

As a result, when the voice from the speaker is smaller than the specified value during a call, the past filter coefficient of the line echo canceller unit 16 (in order to calculate the filter, the past input value of the filter coefficient is saved. By starting the learning from), the learning for echo cancellation can be quickly converged, and the residual echo can be reduced. The filter coefficient (number of taps) is represented by h [N] when there are N, and indicates the expected output for a certain input.

The present invention is not limited to the above-described examples, and includes various modifications. For example, the above-described embodiment has been described in detail in order to explain the present invention in an easy-to-understand manner, and is not necessarily limited to those having all the described configurations. Further, it is possible to replace a part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. Further, it is possible to add / delete / replace a part of the configuration of each embodiment with another configuration.

11 Microphone 12 A / D converter 13 Arithmetic processing unit 14 Communication unit 15 Acoustic echo canceller unit 16 Line echo canceller unit 17 Level determination unit 17'Holding sound (music) detection unit 18 Time determination unit 19 D / A conversion unit 20 Calculation Processing unit 21 Speaker 1020 Echo canceller 1315 Acoustic echo cancel circuit 1620 Line echo cancel circuit 1718 Sound detection circuit 1800 Hours measurement / judgment circuit

In order to solve the above problems, one of the representative echo cancellers and IP telephones of the present invention is
Calling party (conversation partner) can silence the sound input from the microphone, when the及beauty holding tone state, is provided with a means for stopping an echo canceller training in the line echo canceler.

In order to correct such a problem, in this embodiment, the voice detection / control circuit 1718 detects the hold sound from the hold sound generation unit (not shown) of the other party, for example, music, in addition to the level determination unit 17. hold sound (music) detector 1 7 'and provided, music on hold (music) detector 1 7' when detecting a hold sound (music), the one that stops the learning for echo cancellation in the line echo canceler section 16 Is.

Claims (4)

In an echo canceller equipped with a line echo canceller having a function of detecting a line echo and canceling the line echo.
A voice detecting means for detecting the voice from the other party on the receiving path side of the line echo canceller and / or the holding sound is provided.
An echo canceller, characterized in that, when the voice detecting means detects silence and / or a holding sound of a voice from a other party, the learning for echo cancellation in the line echo canceller is stopped. In the echo canceller according to claim 1,
The line echo canceller has a line echo canceller unit and an arithmetic processing unit.
The voice detecting means has a level determination unit that determines the level of voice from the other party.
The line echo canceller unit
Based on the microphone input audio signal of the arithmetic processing unit and the output signal of the level determination unit, a signal corresponding to the line echo for canceling the line echo is generated, and the signal corresponding to the line echo is generated by the arithmetic processing unit. Output to
The level determination unit
The voice from the other party on the earpiece is detected, the level of the voice from the other party is determined, and the result is output to the line echo canceller unit.
The arithmetic processing unit
An echo canceller characterized in that a received signal including a line echo component of the other party's voice and a signal corresponding to the line echo from the line echo canceller unit are subtracted to remove the line echo component from the received signal. In the echo canceller according to claim 1,
Further, it has a time measuring / determining means including a time determining unit for measuring and determining the call time from the other party.
The time determination unit
It is determined that the voice from the other party on the earpiece is silent and / or the voice talk time is smaller than a preset predetermined value, and the result is output to the line echo canceller unit.
When the speaker of the IP telephone to be used can be identified, the voice data from the other party at the time of the last call is stored in the memory built in the line echo canceller unit, and the voice from the other party is specified during the call. If it is smaller than the value, the echo canceller is characterized in that learning for echo cancellation is started from the past filter coefficient of the line echo canceller unit. An IP telephone having a built-in echo canceller according to any one of claims 1 to 3.

Images