fastica的改进算法及应用研究电路与系统专业论文.docxVIP

ABSTRACTIndependent ABSTRACT Independent Component Analys i s(ICA)i s a data process ing method which search for its instrinsiC statistical independence and non—gauss’S factor from multivariate data．ICA can separate source signals which iS statistically independent when nothing iS known about source signals and hybrid matrix．ICA has very important appl ication value in many aspects due to that the independence assumptions can be satisfied in the most practical problems，such as speech signal processing， image processing，face recognition，pattern recognition，data mining，and medical signal processing and SO on．The research of its theory and algorithm also draws the attention of broad scholars．ICA comes to mature with nearly 20 years’development， but there are still some problems to be further perfect at present，including how to improve the convergence of the al gori thm to better apply to the pract ice． This paper focuses on the relatively mature and widely used fast independent component analysis algorithm based on negat ive entropy(FastICA)，The Fast ICA algorithm based on the——fifteen order Newton iterative correction form iS improved by modifying kernel iterative process aiming to solving the III 万方数据 problem problem that basic FastICA algorithm iS sensitive to initial values．Simulation experiments show that initial value sensitivity iS improved with more faster and stable convergence speed when the convergence of the results i S ensured． In this paper，the improved FastICA algorithm iS introduced to the moving object detection of video sequences，each video frame i s regarded as mixtures of the independent sources such as background image and the foreground image wh i ch cons i st s of the moving object．With the improved FastICA algorithm，the moving object image and the background image could be est imated， SO the moving object could be detected and extracted from the video frames．Experimental results show that the moving object detect ion method usi