Higher-order Statistics-based Blind Signal Separation

Blind sources separation is becoming the problem that gets more and more focus because of broad application in many fields, such as communication, speech recognition and so on. We offer a model for blind sources separation which is based on high-order statistics.In this paper we just consider the condition under which many source signals are mixed instantaneously. First, we offer a method through that we can count the number of source signals , which uses the knowledge of matrix , the high-matrix and physical information of mixed signals. Then it is to be given how to choose the contrast function which is established on the knowledge of information. A new algorithm is to be given ,which is based on the contrast function and to be proved efficacious. Finally we design a system that can simulate blind sources separation. The result shows thatthe model and algorithm are effective.