Research On Scientific Computing Free Software Scilab And Blind Signal Separation Problems

The thesis is concerned with the scientific computing software Scilab and the problem of Blind Signal Separation (BSS).Currently, commercial scientific computing software such as Matlab, Mathematica and Maple has been widely used around the world. Scilab, which can provide an equivalent function as Matlab for engineering and scientific applications, is an open source scientific computing software developed by INRIA. By modifying the source code, Chinese localization of Scilab is realized in this thesis, thereby laying a fine foundation for its further study, research and application.BSS has attracted a great deal of attentions from the signal processing community recently. The thesis not only deals with the BSS models, realization method and performance index, but also reviews some existed typical BSS algorithms. The problem of blind separation of sources with mixed kurtosis signs is considered. A new algorithm is presented based on the Kullback-Leibler (K-L) divergence principle, and a new method to construct Score Function is also derived, which can perform blind separation of mixed sub-Gaussian and super-Gaussian sources. The over-determined BSS problem is considered, and the number of sources is even assumed unknown. A new algorithm based on Principal Component Analysis (PCA) and information maximization is developed. Computer simulations verify the validity of the algorithm framework. In addition, BSSLAB toolbox for simulating BSS algorithms based on Scilab is developed. The toolbox is easy to use and provides a good environment for the research on BSS problem.