FPGA implementation of novel fast confluence adaptive independent component analysis for mixture of sub and supergaussian signal

Abstract

Independent component analysis (ICA) is a technique that separates the independent source signals from their mixtures by minimizing the statistical dependence between components. This paper presents a low-power FPGA implementation of a novel 2-channel fast confluence adaptive independent component analysis (FCAICA) technique for mixture of sub-gaussian and super-gaussian signals. The proposed FCAICA consumes less power and provides the high convergence speed. The reduction in power is achieved by hardware optimization and high convergence (confluence) speed is achieved by a novel optimization scheme that adaptively changes the weight vector based on the kurtosis value. To increase the number precision and dynamic range of the signal, the floating-point (FP) arithmetic units are used. To validate the performance of the proposed ICA, simulation and synthesis are performed with sub and super-gaussian mixtures and experimental results are compared with Fast ICA and SFLO ICA (Shuffled frog Leap Optimization ICA). The proposed low power ICA processor separates the mixture of super and sub-Gaussian signals with maximum operating frequency of 2.91MHz.The FCA ICA, Fast ICA and SFLO ICA algorithms converge to optimal solution at 300ps, 200ps and 500ps, with power consumption of 246.94 mW, 270.76mW and 307.27 mW respectively.