结合主分量分析与DOA估计的语音盲分离  

Blind speech separation combining with principal component analysis and DOA estimation

作　　者：王国鹏[1] 刘郁林[1] 罗颖光[1] 

机构地区：[1]重庆通信学院DSP研究室,重庆400035

出　　处：《声学技术》2009年第5期624-628,共5页Technical Acoustics

基　　金：国家自然科学基金(60672157;60672158);重庆市自然科学基金(CSTC2005BB4219)

摘　　要：在欠定语音盲分离中,W-分离正交性假设通常使问题简化,但这种简化是以降低分离性能为代价。在语音信号满足近似W-分离正交性的假设下,提出利用主分量分析(PCA)检测只有一个源信号存在的时频点,检测出的时频点均满足W-分离正交性,因此提高了混合矩阵的估计精度。通过从混合矩阵中估计源信号的波达方向,可以较好地解决置换模糊问题。仿真结果表明,提出的方法与经典的DUET方法相比具有更优的性能,平均信干比提高了2.77dB。The assumption of W-disjoint orthogonality （W-DO） can simplify the problem of blind separation for under-determined mixed speech signals at the cost of decreasing the separation performance. A method based on principal component analysis （PCA） is proposed to detect the timefrequency cells where only one source exists, under the assumption of approximate W-DO of speech signal. All the detected time frequency cells satisfy the W-DO, so that the estimation precision of the mixing matrix is improved. The direction of arrival （DOA） of sources is estimated from the mixing matrix and is exploited to solve the permutation ambiguity problem. Simulation results demonstrate that the proposed method outperforms the typical DUET method, the average signal to interference （SIR） is improved by 2.77dB.

关 键 词：语音盲分离 主分量分析 波达方向 混合矩阵估计 

分 类 号：TN911.7[电子电信—通信与信息系统]