Echo cancellation method, system and device

Abstract

The invention provides an echo cancellation method, system and device. The method comprises the followings steps: A, dividing input time-domain signals into M sub-signals, wherein the sub-signals are subband signals or frequency-domain signals, and M is an frequency domain which is more than 1; and B, for each sub-signal, using far-end signal energy and error signal energy of the sub-signal at the existing sampling time and filtering parameters at the previous sampling time to set filtering parameters of an adaptive filter at the existing sampling time; and then carrying out the echo cancellation on the sub-signal at the existing sampling time in accordance with the set filtering parameters. According to the invention, the effects of the echo cancellation can be improved.

Description

technical field [0001] The invention relates to the technical field of signal processing, in particular to an echo cancellation method, system and device. Background technique [0002] In voice communication systems, echoes are often generated due to the coupling of speakers and microphones. In order to improve the performance of voice communication system, echo cancellation has become an important topic in the field of voice communication. However, in voice communication systems, the double talk problem becomes the bottleneck of echo cancellation. [0003] The so-called double-end sounding specifically refers to: in the context of hands-free two-way conversation, both ends speak at the same time, which is equivalent to inputting noise into the adaptive filter, which affects the normal update of the adaptive filter and causes the adaptive filter to fail. If the filter coefficient of the device diverges, the opposite end can hear its own echo, and the call experience is not...

AI Technical Summary

Problems solved by technology

However, it is often difficult to meet this condition in the actual application of echo cancellation. In this way, there will be the following problems: when the dialogue starts, the probability of double talk is higher than that during the conversation. If the adaptive filter is still During the convergence process, use the above-mentioned energy comparison method or correlation calculation method to perform double-talk detection and control the adaptive filter to stop iterations, which will lead to a vicious cycle of echo cancellation performance deterioration and detection errors

[0029] In addition, no matter what method is used to detect double-ended voice, there will be a certain probability of false detection

If the iteration of the adaptive filter is stopped after the false detection, it may also cause the detection and the adaptive filter to fall into a state that cannot work normally. Controlling the adaptive filter to stop iterating after double-ended sounding will cause the iterative freezing of the filter coefficients, resulting in a vicious cycle of deterioration of the echo cancellation performance, which seriously reduces the echo cancellation effect

Images