Novel fast confluence adaptive algorithm for independent component analysis

Abstract

Independent component analysis (ICA) is a statistical signal processing technique which is used for separation of original signals from their mixtures. In this paper, a novel adaptive optimization algorithm is proposed to increase the convergence speed of the ICA algorithm. Adaptively changing the weight vector based on the fitness value increases the convergence speed of the algorithm. The use of Shuffled frog Leap optimizations (SFLO) ensures the convergence of the algorithm to a global optimum. The proposed Fast confluence Adaptive Evolutionary ICA algorithm (FCA-ICA) is compared with ICA based on SFLO and Fast ICA. Performance comparison shows that FCA-ICA shows improved performance over SFLO-ICA and Fast ICA. The complexity of these ICA techniques is reduced by using modularity, hierarchy and parallelism concepts. Floating point IEEE single-precision representation is used for all the ICA manipulations for improving the accuracy and dynamic range of the signal. The proposed FCAICA processor separates the sub-Gaussian signals from their mixtures with maximum operating frequency of 2.91 MHz. The area power and timing analysis are done with ALTERA FPGA and reports are discussed.