A Maximum Entropy Echo Cancellation Method Based on Zero Attraction and Variable Kernel Width

Abstract

A kind of maximum entropy echo cancellation method of zero-attraction variable nuclear width, its step is: A, obtain signal, the speech signal sampling that far-end transmits obtains the far-end signal discrete value x (n) of current moment n; Simultaneously, to The echo signal collected by the near-end microphone is sampled to obtain the expected signal d(n) at the current time n. B. The output of the filter, C. Echo cancellation, the error signal e(n) is obtained by subtracting the near-end signal d(n) from the output signal y(n), D. The update of the weight coefficient vector: D1. According to the current moment The error signal e(n) of n, calculate the error signal D2 of the current time n to remove the impulse interference, and calculate the kernel width σ of the current time n n ; D3, using the kernel width σ of the previous time n n Update to obtain the tap weight vector W(n+1) at the next moment n+1; E, let n=n+1, repeat steps A, B, C, and D until the call ends. The method has strong identification ability for the telephone communication system, fast convergence speed, low steady-state error, and obvious echo cancellation effect.

Description

technical field [0001] The invention belongs to the technical field of adaptive echo cancellation for voice communication. Background technique [0002] In communication systems, the interference of noise and echo cannot be ignored. Among them, the most important impact on the quality of voice calls is the acoustic echo. Acoustic echo refers to the ability for users to hear their own voice repeatedly during communication. Short echo delays are imperceptible and can be understood as a form of spectral distortion. On the contrary, if the delay exceeds tens of milliseconds, the echo can be detected alone. Since the human ear is extremely sensitive to echoes, research on methods for canceling acoustic echoes is still a hot topic. [0003] In the adaptive echo cancellation method, the widely used methods mainly include the traditional least mean square (LMS) algorithm and the affine projection (APA) algorithm. But in the echo channel, most of them are sparse channels, and on...

AI Technical Summary

Problems solved by technology

However, since the algorithm does not consider the sparsity of the system, the performance will decrease when solving the sparse system problem

Images