Echo Cancellation Technology of VoIP technology

"On the side of a PBX or bureau switch, a small amount of electrical energy is not fully converted and returned along the original path to form an echo." If the caller is not far from the PBX or switch, the Echo returns quickly and the human ear cannot hear it, which is irrelevant. But when the echo returns more than 10ms, the human ear can hear the obvious echoes. In order to prevent echoes, the echo cancellation technique is generally needed, and special software code in the processor listens for the echo signal and removes it from the hearer's voice signal. For IP telephony devices, the echo cancellation technology is very important, because the general IP network delay can easily reach 40~50ms. ”
　　
I. Characteristics of echoes in Internet voice communication
Compared with the traditional telephone, the real-time transmission of voice on the internet has its fatal weakness, that is, the poor voice quality, the factors affecting the quality of the internet voice are manifold, one of the most critical factors is the influence of ECHO. Therefore, in order to improve the voice quality of the Internet, it is necessary to process the echo in the voice transmission of the Internet, that is to say, the IP telephone gateway, as the Voice access device of the Internet, must have the function of eliminating echo. Because the voice transmission of the internet is a kind of brand-new telecommunication service which realizes by the packet switching technology, transmits the speech signal to encode, the compression, the packing and so on a series of processing, this not only causes the echo path the delay to be big, moreover the delay jitter is also big. Therefore, in the Internet voice transmission process, the echo problem is particularly prominent, and has the following characteristics.
　　
1, Echo Source complex
　　
In the traditional telephone system, there is a so-called "circuit comeback." The main origin of the Echo is the conversion of 2-4 lines in the system. Completion of the 2-4 conversion of the mixer due to impedance matching, resulting in "leakage", resulting in "circuit echo." From the Internet IP Telephony Gateway connection way can be seen, IP telephony gateway one end of the PSTN connected to the other end of the Internet.
　　
Although the circuit echoes are generated in the PSTN, they also transmit the IP telephony gateway, which is one of the echo sources in the Internet voice transmission, and the second echo source in the Internet voice transmission is the so-called "acoustic echo". Acoustic echo is the sound of the Speaker's voice is picked up by the microphone to send back to the far end, which allows the remote speakers to hear their own voice. Acoustic echoes are divided into direct echoes and indirect echoes. Direct echo refers to the sound of the speaker playing out without any reflection directly into the microphone. The echo delay is the shortest, and it is related to the voice energy of the remote speaker, the distance between the speaker and the microphone, the angle, the volume of the speaker, and the pickup sensitivity of the microphone. An indirect echo is a collection of echoes generated by the sound of a speaker's voice when it is reflected through a different path, once or more, into the microphone. Because the changes in the surrounding objects, such as the movement of human beings, will change the return path of the Echo, because this echo is characterized by multipath, time-varying. In addition, background noise is one of the factors that produce echoes.
　　
2, the delay of the Echo path is large
　　
In the voice transmission of Internet, there are three kinds of delay sources: compression delay, packet transmission delay and processing delay. Speech compression latency is a major delay in generating echoes, for example, in the g.723.1 standard, the maximum delay for compressing one frame (30ms) is 37.5ms. Packet transmission Latency is also an important source, and tests show that the maximum end-to-end transmission delay is up to 250ms. Processing delay is the encapsulation delay of speech packet and its buffer delay.
　　
3, the Echo path delay jitter is big
　　
In the process of voice transmission in the Internet, there are many uncertain factors, such as echo path, speech compression delay, packet transmission routing and so on, and the range of fluctuation is large, usually between 20~50ms.
　　
II. structure and correlation algorithm of acoustic echo Canceller
With the development of ECHO-eliminating technology, the focus of the current study of ECHO cancellation has been shifted from "Circuit echo" to "acoustic echo".
　　
(1) Treatment of surrounding environment
　　
To analyze the mechanism of acoustic echo generation, it can be known that the simplest control method of acoustic echo is to improve the ambient environment of the loudspeaker, and to minimize the reflection of the speaker's playback sound. For example, you can attach a layer of sound absorbing material to the surrounding walls, or add a layer of padding to increase the scattering, ideally in an ambient environment where the echo time or RT-60 (the time required for sound attenuation 60dB) is between 300ms~600ms. Because such an environment can control reflection, it can not make the speaker feel uncomfortable. The improvement of the environment can effectively inhibit the indirect acoustic echo, but the direct acoustic echo is powerless.
　　
(2) Echo suppressor echo suppressor is an early echo control method. The echo suppressor is a nonlinear echo cancellation. It uses a simple comparer to compare the level of the sound that is ready to be played by the speakers with the sound that the current microphone picks up. If the former is above a certain threshold, it is allowed to pass to the speaker, and the microphone is turned off to prevent it from picking up the sound that the speaker is playing and causing a remote echo. If the microphone picks up the sound level above a certain threshold, the speaker is banned to eliminate the echo. Because the echo suppression is a nonlinear echo control method, it will cause the speaker to play discontinuous. Effect of Echo cancellation, with the advent of high-performance Echo Canceller, the echo suppressor has been used by few people.

(3) Acoustic echo Canceller Another method for acoustic echo cancellation is the use of acoustic echo Canceller (aec:acoustic echo Chancellor), which is a voice model of the remote signal based on the correlation of the speaker signal and the multipath echo produced by it, By using it to estimate the echo and modify the coefficients of the filter continuously, the estimated value is more approximate to the real echo. The echo estimate is then subtracted from the microphone's input signal to eliminate the echo, and AEC also compares the microphone input with the speaker's previous values, eliminating the acoustic echoes that extend the multiple reflections of the delay. The AEC can eliminate all kinds of delay echoes by how much the output value of the past speakers is stored in the memory.