Blind Source Separation Algorithms Based On Degree Of Cyclostationary And Joint Appoximate Diagonalization

Blind Source Separation (BSS) is a method which aimed at recovering and extracting desired signals from observed signals, for the premise of the channel priori information and the source signals unknown. As a new branch of digital signal processing, BSS has important theoretical significance and practical value, which has been widely used in many fields such as wireless communications, biomedical signal processing, geological exploration, speech recognition, image signal processing and etc.. It has received considerable attention in signal processing and neural network academic circles.The traditional blind source separation algorithm was proposed on the basis of stationary signal. But most of the natrual signals have the characteristics of cyclostationary (CS). This paper processed the mixed signal based on cyclostationary theory, and extract the useful information from it to reduced the computation complexity.The major contribution of this paper are as follow:1. Analyzed the cyclostationary theory and its application in the field of signal separation. Introduced the concepts, principles and techniques, and the classification of blind source separation algorithm, mainly analyzed the advantages and disadvantages of these algorithms.2. To improve the separation result, deduced the second order degree of cyclostationary (DCS) based on CS. The newly proposed BSS algorithm based on second order DCS can extract the useful information from part of the separated signal, which is effectively to surpress the stationary noise. The simulation results showed that this algorithm convergent speed faster and separated better.3. Combined DCS criterion and joint approximate diagonalization (JADE) algorithm to overcome the DCS algorithm ineffectiveness of the multi CS signals, which can separate the source signals with same cycle frequency breaking through the limitation of other BSS algorithms separating the mixed signals with different cycle frequencies. This new algorithm suit the noise environment superior to the traditional JADE algorithm suitable for the non-noise condition. Simulation results showed that the new algorithm has better separation result with greater fault tolerance.