基于负熵最大的FastICA语音信号分离算法

语音信号分离是现代信号处理的热点问题,针对未知信号源个数的情况,提出一种基于负熵最大的FastICA(Fast Independent Component Algorithm)语音信号盲分离算法,有效解决了源信号数目估计、语音信号分离及复原等问题。改进的算法增加了源信号数目估计环节,放宽了算法适用条件,即在源信号数目未知的情况下,也能够实现信号盲分离功能。并将其成功应用于运用信号分选过程中,最终复原语音时域波形,完成信号分选任务。仿真实验中,详细讨论了该方法在不同信噪比以及不同源信号数目情况下的分选能力,证明了方法的有效性和优越性。

Research of FastICA Speech Signal Separation Algorithm Based on Negative Entropy Maximum

Speech signal separation is a hot topic in modern signal processing problems,aiming at the condition of the number of unknown source,a kind of separation algorithm based on negative entropy maximum fast independent component the correlation algorithm is proposed in this paper,this algorithm can effectively solve the problem of unknown-number of source signals. The improved algorithm increases a link of number estimation of speech signal,and it relaxed algorithm applicable conditions,namely,in the case of a number of unknown source signals,also can realize blind signal separation function. The proposed method has successful applied in the process of the use of signal sorting,the time domain waveform signal sorting task. This algorithm is discussed in detail in different SNR and number of cases of different source signal separation ability,to prove the validity of the method and superiority in simulation experiments.