Blind Signal Processing Techniques Applications In Underwater Target Separation

Blind signal processing is a newly presented and lively area of signal processing domain in recent years. It estimates sources given mixed signals without prior knowledge such as sources number, location and mixing process. It has many potential exciting applications in science and technology, especially in wireless communication signal processing, speech signals processing, image processing and underwater acoustic signal processing.This dissertation studies the basic principle and method of BSS(Blind signal separation, BSS), in detail which are conception theorem, structure, several representative algorithm function for evaluating the BSS algorithm's performance, and some application and so on.Generally, on one hand, BSS approaches can classified into two groups according to the mixing channels:approaches for instantaneous mixtures and that for convolutive mixtures. However instantaneous mixtures are just a special case of convolutive mixtures. On the other hand, BSS approaches can be categorized into three classes with respect to statistical information exploited:information theory, second-order statistics (SOS) and parameters defined in high-order statistics(HOS). In addition, BSS approaches can also be time-domain approaches and frequency-domain approaches according to their implementation manner. Under methods of instantaneous blind source separation, we studied Informax (Information Maximization, Informax) algorithm, natural gradient algorithm and FastICA (Fast Independent Component Analysis, FastICA) algorithm, them were parts of the information theory algorithms. Then research the AMUSE (Algorithm for Multiple Unknown Signals Extraction, AMUSE) algorithm and SOBI (Second-Order Blind Identification, SOBI) algorithm which base on second-order statistics algorithms. At the same time try to separation the mixture noise signals.Through the simulation show those algorithms's performance.Present dissertation addresses the blind separation of underwater acoustic signals. Methods of instantaneous BSS and convolutive BSS were used to separate underwater mixing signals. Above the reviewe of the classic algorithms, use the FastICA algorithm to separation the underwater signal. At the beginning, use the time-domain FastICA algorithm to separation the underwater signal, which mixture approaches was instantaneous. Then we found that convolutive mixtures were better for separation the real underwater signal. So we have deduced the fast fixed-point frequency domain algorithm. And we have modified the fast fixed-point frequency domain algorithm. Besides we have resolved the problem of permutation. Finally, use the real underwater signal to verify the performance of the modified fast frequency domain algorithm. The separation results verify the validity of the modified fast frequency domain algorithm.