A Generalized Mixed-Norm Adaptive Echo Cancellation Method Against Saturation Distortion

Abstract

A generalized hybrid norm adaptive echo cancellation method with anti-saturation distortion, the main steps are: A. remote signal acquisition to obtain input vector X(n); B. echo signal estimation to obtain echo signal estimated value y(n) ), y(n)=X T (n)W(n); C, echo cancellation, near-end signal d(n), estimated value of echo signal y(n), get error signal estimated value e(n); D, filter tap weight vector update : Through the calculation of the probability density of the echo signal, the cumulative value P(n) of the probability density of the echo signal of the estimated value vector Y(n) of the echo signal, the probability density value p(n) of the echo signal, and the probability density value of the negative value of the echo signal are obtained. p * (n), and then obtain the echo signal corrected for saturation distortion and the estimated value of the error at the next moment; then update the tap weight vector according to it; E. Repeat the above process until the call ends. The method has strong anti-saturation distortion performance, small steady-state error, fast convergence speed and good echo cancellation effect, and is especially suitable for strong voice communication scenarios.

Description

technical field [0001] The invention relates to an adaptive echo cancellation method in a voice communication system. [0002] technical background [0003] The echo phenomenon in the voice communication system refers to the phenomenon that the sound signal picked up by the near-end microphone is delayed or deformed during the transmission process and then transmitted back to the near-end and played by the earpiece (that is, people can hear their previous sound during the call). momentary speaking voice phenomenon). Echo can seriously affect the quality of people's calls, therefore, the elimination of echo has become the focus of people's attention. The communication echo can be adaptively eliminated through the system identification model: the identified system is an echo channel, and the output of the system identification is the estimate of the echo signal; Then, it will be played by the earpiece at the near end, so that the echo can be eliminated at the near end. [00...

AI Technical Summary

Problems solved by technology

[0005] However, the existing adaptive echo cancellation methods are all established under the assumption that the speech signal picked up by the microphone is complete

When the near-end voice signal is strong, the near-end microphone will produce saturation distortion, which will cause saturation and distortion of the picked-up near-end signal; resulting in the assumption that the voice signal is complete, and the estimated error signal will be severely distorted, which will lead to algorithm Low accuracy and large steady-state error, cannot be used in strong voice communication scenarios

Images