Computer Engineering and Applications ›› 2013, Vol. 49 ›› Issue (14): 212-216.

Previous Articles     Next Articles

Underdetermined blind source separation of speech signals

YANG Meiyu, LIU Qinghua   

  1. School of Information and Communication, Guilin University of Electronic Technology, Guilin, Guangxi 541004, China
  • Online:2013-07-15 Published:2013-07-31

病态混叠下语音信号的盲源分离

杨美钰,刘庆华   

  1. 桂林电子科技大学 信息与通信学院,广西 桂林 541004

Abstract: For the defects that blind source separation potential function method requires too many parameters and the number of the source signal needs to be known as priori condition in the clustering algorithm, the potential function method based on Laplacian model is used to estimate the number of source signals and the mixing matrix. Then the mixed signals are re-clustered, and the covariance matrix of each type of signal is solved with the singular value decomposition. The mixing matrix is estimated more precisely, and then the source signals are also estimated more precisely. Through computer simulation, it demonstrates the superiority of the proposed algorithm.

Key words: singular value decomposition, mixing matrix, sparse signal, potential function, clustering

摘要: 针对稀疏信号盲源分离势函数法需要过多参数,以及聚类算法需要知道源信号个数的缺陷,采用基于拉普拉斯模型的势函数法估计源信号数目和混合矩阵。将混合信号重新聚类,对每一类信号的协方差矩阵进行奇异值分解,混合矩阵得到更精确的估计,进而源信号也得到更精确的估计。通过计算机仿真,表明了该算法的优越性。

关键词: 奇异值分解, 混合矩阵, 稀疏信号, 势函数, 聚类