摘要
在混响时间较长的情况下,一般的频域盲源分离算法可能不完全收敛,导致分离信号中包含部分干扰信号,降低分离算法性能。根据语音信号在时频域分布的稀疏特性,估计分离信号中目标信号和干扰信号的功率谱密度,提出一种改进的频域维纳滤波后处理算法。仿真实验结果证明,与原有频域维纳滤波算法相比,该算法在不增加计算复杂度的前提下,抑制的干扰信号增加了1dB～2dB,是一种有效的频域盲源分离后处理算法。
Under long time reverberation circumstances, general frequency-domain blind source separation algorithm may converge incompletely, which result in interference signal in separation signali and the algorithm performance decrease. This paper proposes an improved Wiener filtering postprocessing algorithm according to the sparse properties of speech signal in time-domain and frequency-domain, and by estimating power spectrum density of separation signal and interference signal in separation signal. Experimental results show that compared to the original Wiener filtering algorithm, this algorithm can increase interference signal by 1 dBN2 dB without increasing computation complexity.
出处
《计算机工程》
CAS
CSCD
北大核心
2009年第11期202-204,共3页
Computer Engineering
基金
温州市科技计划基金资助项目"语音的盲分离和说话人定位技术研究"(G20060102)
关键词
盲源分离
维纳滤波
后处理
blind source separation
Wiener filtering
postprocessing
