Underdetermined Blind Source Separation in Single Mixtures Signal

Abstract

The paper proposes a new method based on the layered technique and the independent sub-band functions for overcoming some difficult problems of the underdetermined blind source separation in single mixtures signal. The new method needn't use any training data to separate the multisource signals from the single mixed-signal. Based on one-way mixed-signal separation model, the paper brings forward the layered principles of one-way mixed-signal processing, and discuss how to do this layered work and how many layers should be separated into? It introduces the ways for selecting the number of independent component, and then through a combination of independent sub-wave functions entry into the one-way mixedsignal, thus we can realize the blind separation. Through the separation experiments this method is confirmed to be validity and the feasibility.