Source Separation Technology Research

Blind Signal Separation (BSS) is mentioned more and more in recent decade. The research can be divided into two classes: 1) the separation of instant mixed signals, 2) the separation of dynamic mixed or mixed signals obtained after convolution of original signals. Because the blind processing techniques do not need a prior information of the original source signals and the channel, they have significant value in theory study and broad foreground in practical applications in the fields of digital communication, array signal processing, biomedical signal processing, speech dignal processing, etc.This dissertation primarily discusses the linear mixture model of the sources in BSS and blind equalization, it mainly contains two components: instantaneous mixture and convolutive mixture model. The main contributions are as follows: The blind signal extraction approach may have several advantages over simultaneous blind separation, such as the specified order, the flexible, local and simple algorithms, etc. Blind signal extraction of temporally correlated sources based on the second order statistics (SOS) assumes that source signals have some temporal structure. Results show the validity of proposed method In this dissertation, we present using various models and assumptions, relatively simple and efficient, adaptive algorithms for blind deconvolution and equalization for single-input multiple-output (SIMO) and multiple-input multiple-output (MIMO) dynamical minimum phase and non-minimum phase systems. The natural gradient algorithms with nonholonomic constraints are derived by incorporating a nonholomic constraint into the natural gradient algorithms. It overcomes the drawback of the Bussgang algorithm. The result shows that proposed method coverges very quickly and is more stable than Bussgang algorithm.