基于加权次梯度投影算法的数字助听器自适应声反馈抑制  

Feedback Cancellation based on Weighted Adaptive Projection Subgradient Method in Hearing Aids

作　　者：王青云[1,2] 赵力[1] 乔杰[1] 邹采荣[1] 

机构地区：[1]东南大学信息科学与工程学院,江苏南京210096 [2]南京工程学院通信工程学院,江苏南京211167

出　　处：《信号处理》2009年第4期519-525,共7页Journal of Signal Processing

基　　金：国家自然科学基金项目(No:60472058);教育部博士点基金(No:20050286001);教育部新世纪优秀人才支持计划(No:NCET-04-0483)

摘　　要：本文提出了一种利用加权自适应次梯度投影算法(Weighted Adaptive Projection Subgridient Method,WAPSM)进行声反馈抑制的方案。WAPSM算法来自于自适应次梯度投影算法(Adaptive Projection Subgridient Method,APSM),它以次梯度投影的超平面作为搜索区域来进行松弛投影。本文提出的算法将估计系统的先验知识以权重因子一在很多应用中为指数衰减一的方式加入APSM算法中提高算法性能。以WAPSM算法应用于助听器声反馈抑制的大量仿真实验表明,算法相比传统的NLMS算法和APSM算法在收敛速度、稳定性和精度方面取得了显著的进展。进一步的实验表明,算法在以实际语音作为数字助听器输入信号时取得了优异的性能,并且在低信噪比条件下具有较强的鲁棒性。Weighted adaptive projection subgradient method （WAPSM） is proposed in this paper for feedback cancellation in hearing aids. WAPSM is derived from the technique of adaptive projection subgradient method（ APSM）, which utilizes the subgradient projection hyperplanes as the searching area in the process of relaxed projection. In this paper weight factor - which is exponential decayed in most implementations - is added to APSM to incorporate a priori information. We applied this WAPSM algorithm for acoustic feedback cancellation in hearing aids. Numerical experiments demonstrate that notable improvements are achieved including speed, stability and accuracy of convergence compared to the traditional NLMS algorithm and APSM algorithm. Another exciting conclusion by further experiments is that WAPSM achieves excellent performance in the situation that real speech segment as input in hearing aids, and WAPSM is more robust for low SNR compared to other algorithms.

关 键 词：声反馈抑制 加权自适应次梯度投影算法 助听器 

分 类 号：TN912[电子电信—通信与信息系统]