The Improved Method Of Blind Signal Seperation Algorithm And Its Applications

Blind Signal Processing (BSP) is mentioned more and more in the latest decade of last century. It can be divided into several subfields which are relative but oriented distinct objectives, as Blind Signal Separation (BSS), Multichannel Blind Deconvolution (MBD), blind deconvolution, blind equalization etc. In fact, BSP has become a hot topic for research and development in many areas, especially communication systems, speech enhancement, pattern recognition, biomedical engineering, medical imaging, remote sensing, exploration seismology, geophysics, econometrics, data mining, etc.Blind source separation (BSS) is a very important topic of BSP. BSS refers to the problem of recovering signals from several observed linear mixtures. The strong point of the BSS model is that only mutual statistical independence among the source signals is assumed and no other priori information about, e.g., the characteristics of the source signals, the mixing matrix or the arrangement of the sensors is needed. Independent Component Analysis (ICA) is a significant method of BSS, which is aimed to separate signals in a statistical way as independent as possible.This thesis is about the research of blind signal separation. At first, we introduce the mathematical principle of blind signal separation and the basic model. We give the relation between non-gaussianity and statistical independence. Based on linear instantaneous mixture model, we study the algorithm of BSS, and briefly show the development, applications and state of ICA, particularly treat the basic principle and implementation of ICA. We carefully studied the FastICA algorithm with one unit and several units. Focus on the shortcoming of FastICA algorithm; we propose our novel algorithm based on cumulant tensor which is called improved method of FastICA. Then, a performance comparison study on these two approaches is conducted through the simulations on some standard benchmark data. The experimental results demonstrate that the IM-FastICA achieves higher performances on unmixing error and signal noise ratio while appreciably increasing computation cost. Finally, we give the applications of BSP and ICA methods in communication systems.