基于经验模态分解的子带自适应声学回声消除算法

在语音通信中,声学回声消除技术用于消除扬声器与麦克风之间耦合产生的回声干扰.在声学回声抵消系统的实现过程中,可以通过子带技术来提高系统的性能,并减小算法本身的运算量.常见的子带算法多是基于组合滤波器、小波变换实现的.本文基于经验模态分解提出一种新的自适应回声消除算法EMD-APNLMS,它克服了基于组合滤波器算法收敛慢的缺点以及基于小波变换算法需要选择小波基的问题.计算机仿真结果表明,该算法实现了回声的消除,收敛速率较快,非常适用回声这种非平稳信号的处理.

Sub-band Adaptive Filtering Algorithm Based on Empirical Mode Decomposition in Acoustic Echo Cancellation

In voice communications,acoustic echo cancellation technology is used to eliminate echo caused by coupling between speaker and microphone.In acoustic echo cancellation system implementation process,the sub-band technology can improve system performance and reduce the computational algorithm itself.Common sub-band algorithm is based on the combination of multiple filters,wavelet transform.This article proposes a new adaptive echo cancellation algorithm——EMD-APNLMS which based on empirical mode decomposition.It could overcome the slow convergence of the algorithm based on the combination of multiple filters;and it could overcome the drawback of the algorithm based on wavelet transform which need to choose the wavelet basis.The computer simulation results show that the algorithm achieved to overcome the echo elimination;it owns faster convergence rate;so the algorithm is fit for processing this non-stationary echo signal.