基于麦克风阵列的近场声源定位子阵算法研究  

Research of Speech Source Localizations Sub-array Algorithms in Near Field Based on Microphone Arrays

作　　者：居太亮[1] 彭启琮[1] 邵怀宗[1] 

机构地区：[1]电子科技大学通信与信息工程学院

出　　处：《电子测量与仪器学报》2006年第5期50-55,共6页Journal of Electronic Measurement and Instrumentation

基　　金：华为高校科技基金资助项目(编号:YJCB2004013MU)。

摘　　要：基于麦克风阵列的声源定位技术在通信、语音处理等领域得到广泛的应用。结合远场窄带信号的子阵算法和麦克风阵列信号处理的特点,提出了基于麦克风阵列的声源三维定位子阵算法。根据室内环境噪声分布特点,把阵列分成两个位置不同的子阵,调节子阵的位置,使阵列接收相同的语音信号和不相关的方向噪声,利用两个子阵的互相关矩阵,实现声源定位。该算法能够抑制混响、干扰等非高斯百噪声,提高声源定位系统的性能。仿真结果表明,该算法比声源三维定位MU-SIC算法有更好的定位性能。The speech source localization technology based on microphone array has been extensively applied to communications and speech signal processing. In this paper, a near field speech source three-dimensional（3D） localization sub-array algorithm has been proposed by combining the characteristics of microphone array signal processing and subarray methods based on far field narrowband signal. The array is divided into two sub-arrays accord- ing to the noise＇s location. The sub-array＇s location can be adjusted so that the speech signals captured by the arrays are the same, and the directional noises are non-correlative. The cross correlation matrix about the two sub-arrays can be calculated. The speech source location is solved according to MUSIC or focusing methods. This algorithm can restrain non-Gauss white noise such as reverberation and interfere etc. , and improve the performance of the speech source localization system. Simulation results show the performance of the localization is better than threedimensional MUSIC of speech source localization when this algorithm used.

关 键 词：麦克风阵列 近场DOA估计 子阵算法 子空间方法 语音信号处理 

分 类 号：TN911.7[电子电信—通信与信息系统]