A novel algorithm of blind separation for delayed sources

In this paper, a novel algorithm for blind separation of delayed sources is proposed. We build a group of extended observations which can be approximated by the mixtures of the original sources and their respectively first-order derivatives, according to that the first-order Taylor approximation holds when the delay is small. Unlike most existing algorithms, which need more observations than sourcesthe proposed algorithm only needs observations as many as sources. Second-order blind identification (SOBI) approach is used to solve the new blind separation problem. A simulation analysis was conducted to evaluate the proposed algorithm. We compared the performance of our algorithm with Ning Jiang's algorithm for blind separation of delayed sources, which need twice observations as many as sources. The results show that the proposed algorithm can achieve good performance even under large delay circumstances.