In recent years, with the advent of the era of Big data, many building intercom systems have correspondingly entered the ranks of transformation. The traditional dual wire four wire intercom system is slowly entering the aging phase, and the emerging Ethernet transmission network is once again popular. But in the process of transformation, engineers will also face a new challenge - Echo suppression and cancellation! Echo "is a common problem encountered in the actual use of communication products and accessories. Objectively speaking, whether it is analog communication or digital communication, there must be echoes during use. Therefore, echo canceller products have become an endless topic in the communication industry.

When designing a "Echo suppression and cancellation" product or modular circuit, designers should first understand the mechanism of "echo" generation, and then start from the actual conditions to choose the appropriate product scheme. The following discussion is limited to products under conventional usage conditions in the video conferencing industry. The generation of echo originated from people shouting in an open canyon and hearing their own voice multiple times. This phenomenon is called "acoustic echo", which refers to the emission of sound waves on the surface of a certain object after the sound source is generated, forming a "secondary sound source". If the sound wave is reflected multiple times, it will form the effect of shouting in the canyon. The Echo Wall of the Temple of Heaven in Beijing, China is a historical scenic spot constructed by artificially adopting this echoing principle.

After the phone appeared, people found that they would hear what they were saying after a certain brief delay during the call. This echo phenomenon, we call it "network echo", especially when using a two wire telephone system, which carries bidirectional voice signals on two copper wires. After the radio wave delay, a "secondary signal" will appear.

If the echo in communication causes "multi harmonics", it will cause "self-excited howling" and affect the communication effect. However, in telephone communication, a certain level of "network echo" (side tone) is beneficial for both parties to communicate. The echo discussed in the current building intercom also includes two factors: the side echo caused by signal delay in the circuit and the acoustic echo caused by the venue environment. The following is mainly caused by the acoustic echo. In the figure below, the reason for the occurrence is explained: in communication, indoor unit users and local users form a communication loop, and a two-way communication line forms a closed loop.

As shown in the figure, the voice signals of indoor unit users are collected by the microphone and transmitted to the outdoor unit equipment through communication lines in the form of data signals, which are played out through speakers; The sound played and the sound spoken by the outdoor unit user simultaneously enter the microphone, forming a mixed signal, which is then transmitted to the indoor unit user through a communication line.

After this process, the sound signal heard by indoor unit users from their speakers includes the sound of the local user speaking and the sound of their own speaking, forming an "echo". The occurrence of echo problems affects communication effectiveness, and in severe cases, it can cause "howling" and interfere with the call process. At present, a considerable number of software products use the software Echo suppression and cancellation source code from Skype to improve the quality of products. The way of software Echo suppression and cancellation is also to increase the digital processing process of audio through the AC97 audio standard to achieve the effect of Echo suppression and cancellation. The transmission of sound will cause delay in this process, and at the same time, in multi user communication with more than 4 users, it will affect the integrity and authenticity of the voice. Moreover, using software to eliminate echoes is quite difficult, and most companies simply cannot obtain reliable algorithms to simply adjust the sensitivity of MIC and speaker volume. Some even sacrifice full duplex call indicators to achieve the purpose of echo cancellation, which is a cure for symptoms rather than the root cause. So is there a relatively perfect way to solve the problems encountered by the current building intercom system? The answer is yes. We all understand that sound waves, as a type of conductive wave, include two parameters: the phase of the wave and the amplitude of the wave. In the logical relationship of waves, the result of the logical sum of inverse and equal signals is zero. The basic principle of Echo suppression and cancellation is to use a signal wave intervened by one person to eliminate the echo signal generated in the communication process, while retaining other normal voice signals to achieve normal use of communication.

At present, the schemes of echo canceller products of various brands are basically to obtain an audio signal synchronously at the output end of the sound source in the meeting place. This signal is shifted+inverted with a certain delay. At the same time, the amplitude of the signal is amplified to the average amplitude range of the "secondary sound source" according to different use conditions. The processed signal is logically added to the input signal of the venue sound source (microphone) to cancel out the echo signal. The other voice signals in the venue were input into the system normally because there was no cancellation signal. This is a relatively simple and cost-effective method, as shown in the following figure: after simplifying the above figure, we can obtain two signals: one is a mixed blue and red signal 1, which is the voice stream formed by mixing the actual speech to be sent with the actual unwanted echo; The other is signal 2 with a dashed line, which is the original voice that causes echoes. So everyone will say, oh, so echo cancellation is so simple, why don't we just subtract the 2 of this dashed line from mixed signal 1? Please note that there is a difference between the dashed signal 2 obtained and the echo echo, and directly subtracting will make the speech completely unrecognizable. We call mixed signal 1 the near end signal ne, and dashed signal 2 the far end reference signal fe. Without this signal fe, echo cancellation is an impossible task. Although the reference signals fe and echo are not exactly the same and have differences, they are highly correlated, which is why echo refers to them as echoes. At least, the semantics of echoes and reference signals are the same and understandable, but if you say one sentence and immediately hear your own words coming back, it is quite uncomfortable. Since fe and echo are highly correlated and echo is caused by fe, we can express echo as a mathematical function of fe: echo=F (fe). The function F is called the echo path. In acoustic echo cancellation, the function F represents the physical process of multiple reflections of sound on surfaces such as walls and ceilings; In circuit echo cancellation, the function F represents the coupling process of the two to four wire matching of the electronic circuit. Obviously, our next task is to solve the function F. By obtaining the function F, echo can be calculated from fe, and then echo cancellation is achieved by subtracting echo from mixed signal 1.