Researches of Real-Time Signal Processing for Blind Source Separation in Convolutive Mixing Environment and Real-Time Signal Processing for Independent Component Analysis

点击翻译按钮获取中文摘要

(1) We present a novel algorithm for independent component analysis (ICA) based on gradient learning with simultaneous perturbation stochastic approximation (SPSA). This algorithm can work well in on-line mode, in a dynamic mixing environment. It converges very fast even for non-stationary, and/or non-identically independent distributed (non-I.I.D.) signals, so that the algorithm is very suitable for most real-time applications.(2) We present an approach for blind separation of acoustic sources produced from multiple speakers mixed in realistic room environments. We first transform recorded signals into the time-frequency domain. We then separate the sources in each frequency bin based on an ICA algorithm. We choose the complex version of fixed point iteration (CFPI) as the algorithm.(3) We proposed an algorithm for real-time signal processing of convolutive blind source separation (CBSS). We applied an overlap-and-save strategy, and considered the issue of separating sources in the fr … More equency domain. We introduced a modified correlation matrix and performed CBSS by diagonalization of the matrix. We proposed a method that could diagonalize the modified correlation matrix by solving a so-called normal equation for CBSS. A real-time separation of the convolutive mixtures of sources can be performed.(4) CBSS that exploits the sparsity of source signals in the frequency domain was addressed. We proposed a novel natural gradient method for complex sparse representation. Moreover, a new CBSS method was further developed based on complex sparse representation. The developed CBSS algorithm works in the frequency domain.(5) This research presents a new type of algorithm for solving ICA problems. This new algorithm was based on an effective updating scheme in which learning updating acts as a series of orthonormal matrix transformations. One attractive feature of the algorithm is that it does not include any predetermined parameters, such as a learning step size, as do gradient-based algorithms. Less