基于增益最小的语音盲分离算法

blind source separation based on minimum gain ia a new signal separation algorithm of the instantaneous linear mixing signal which has low computational complexity. This covariance matrix function of source signal and noise signal is expressed as a generalized eigenvalue problem. The algorithm do not need any iteration when solve separation matrix through generalized eigenvalue problem. Computer simulation proves that the algorithm can separate a mixture of super-Gaussian and sub-Gaussian source signals, and it can separate speech signal effectively.