Blind Source Separation Based on Variational Bayesian Independent Component Analysis
Chunli Wang, Yan Xu, Minan Tang, Lei Wang · 2018
While the traditional independent component analysis methods were used to separate speech signals, the inherent structural features of the speech signal and the interference of the noise in the hybrid system were not considered, since the observed signal and the known prior information were not fully utilized, the effect of the separation were not ideal. However, the advantages of ICA has been taken and the full use of the prior information of the hybrid system has been made in the algorithm of variational Bayesian independent component analysis (VBICA), the source signal under various noise conditions has been separated, the results indicated that the algorithm of VBICA was more advantageous than the traditional ICA algorithms.