FPGA Implementation of Adaptive Filters using Partial Update Algorithm

Abstract

Adaptive filtering is a vast area of research in present decade in the field of noise cancellation and acoustic echo cancellation. Adaptive Noise Cancellation algorithms are Least Mean Square (LMS), Recursive Least Square (RLS) and Lattice algorithms. RLS algorithm is used in adaptive filter because it provides better convergence. The adaptive filters may require to have large number of coefficients to model an unknown physical medium with ample accuracy. The number of filter coefficients decides the computational complexity of adaptation algorithms. This reveals that, for longer adaptive filters, the adaptation process becomes highly expensive, precluding economic implementation on digital signal processors. In this paper, partial update RLS (PURLS) algorithm is developed for denoising Electrocardiogram (ECG) and speech signal efficiently by reducing computational load in adaptive filter implementations. Conventional RLS and PURLS filters are designed in the MATLAB/Simulink and ModelSim, then the designs are validated by implementing them in FPGA KIT and design vision of Synopsys tool.