基于听觉仿生和压缩感知的助听器声源定位方法  

Acoustic source localization based on compressed sensing and auditory bionics for hearing aids

作　　者：梁瑞宇[1,2] 赵力[2] 奚吉[1] 张学武[1] 王青云[2] 

机构地区：[1]河海大学计算机与信息学院,南京210098 [2]东南大学信息科学与工程学院,南京210096

出　　处：《仪器仪表学报》2011年第6期1390-1395,共6页Chinese Journal of Scientific Instrument

基　　金：国家自然基金(60472058;60975017);江苏省自然科学基金(BK2008291);中央高校基本科研业务费专项资金(2009B32614)资助项目

摘　　要：在声源定位方面,由于麦克风阵列利于评估麦克风对间的声源定位参数,因此其性能比单麦克风优越。但是当麦克风对间距过小时,获得足够的定位参数信息变得非常困难。针对这个问题,在研究奥米亚棕蝇听觉系统定位机制的基础上,提出采用差分微麦克风信号构建包含声源信息的信号模型。同时为了减低计算量,采用多通道联合压缩感知(compressed sensing,CS)进行声源采样。根据构建的CS模型,采用自适应次梯度投影算法重构包含声源信息向量,通过评估重构信号的能量峰值获得声源位置。理论分析和仿真结果表明,该方法同其他声源定位算法相比,具有定位精度高、抗噪声鲁棒性强,计算量小等优点。Compared with single microphone,microphone array helps to access acoustic location parameters between the microphone pairs more easily.However,when the space between the microphones in the microphone pairs is too small,it has become very difficult to acquire adequate positional parameters.Aiming at this problem,this paper proposes a signal model,which contains source information and is constructed using differential microphone signals based on the localization ability of Ormia ochracea.Moreover,to decrease computation complexity,we adopt multi-channel compressed sensing（CS） to obtain acoustic source data.Then,according to the constructed CS model,adaptive projection subgradient method is used for signal reconstruction.Finally,acoustic source location is obtained by estimating the energy in the reconstructed signal.Theoretical analysis and simulation results conclude that the proposed approach exhibits a number of advantages over other source localization techniques,which include increased resolution,improved robustness to noise,less computation task,and etc.

关 键 词：麦克风阵列 压缩感知 听觉仿生 数字助听器 

分 类 号：TN912[电子电信—通信与信息系统]