Underdetermined Blind Source Separation of Dynamic Sources and Mixing Matrix

A new kind of improved two-step approach for underdetermined blind separation of dynamic sources and mixing matrix is proposed in the paper. Firstly, the change time of the mixing matrix is estimated based on planar minimum offset angle of sparse domain. The mixing matrix is then estimated and the dynamic number of sources is determined. Secondly, the source signals are recovered by pseudo extraction vector method, and the underdetermined blind separation of source signals is achieved by applying linear programming method. Results show that not only the proposed method can accomplish underdetermined blind separation of dynamic sources and mixing matrix but also its velocity of separation is ten times of the methods based on linear programming.