High robustness to quantification effect of an adaptive filter based on genetic algorithm

Fixed-point arithmetic operations for adaptive filter present a great advantage in VLSI implementation such as low power consumption, low integration area and high computational speed. However, the processing characteristics (convergence, precision, asymptotic error, stability) are affected by the bit-wordlength used to compute the filter coefficients. This paper presents a method based on a genetic algorithm (GA) increasing the robustness to quantification effect. The method is studied to identify the nonlinear system parameters in comparison with most popular adaptive methods, LMS and RLS adaptive filtering. Simulation results with linear and nonlinear systems are shown to evaluate the robustness of the method in fixed-point arithmetic.