The Research Of Blind Source Separation Algorithm Based On Negentropy And Gaussian Moments Noisy ICA

Blind source separation (BSS) means that it can reconstruct the original signals from some observed signals without any prior knowledge of the mixing system and source signals. Now, the main method to solve the problem of BSS is independent component analysis (ICA). The purpose of ICA is to reach a separate matrix which allows the separated signals to be statistically independent each other. With the deeply research, the ICA is widely used in many fields such as speech signal process, image recognition, biological medicine signal process, communication and so on. In recent years, ICA has been received considerable attention from the mind signal processing community and the neural network community.The major contribution of this paper is summarized as follow:1 . This paper analyzes systematically the research propose and tendency of BSS algorithms, expatiates on the foundational principles of BSS, discusses several applications of BSS algorithms in some fields, discusses three mathematical models of BSS and the typical performance ??indexes evaluating the effect of its algorithms.2. Main problems of the research of ICA are discussed. Several classical cost functions of BSS algorithms based on ICA and their derivations are introduced. This paper unifies them in the information-theoretic framework.3. In the paper, we analyze the fixed-point algorithm based on kurtosis of BSS , we introduce a fixed-point algorithm based on maximize negentropy, which is robust and converges rapidly. It's proved that the new algorithm is effective on the computer.4. This paper expatiates on the foundational principles of BSS on noise, analyzes the fixed-point algorithm, uses the method of bias-deleted and introduces a new algorithm called noisy-ICA which is based on Gaussian moments of function. It's proved that the new algorithm is correct on the computer.