PERFORMANCE OF NOVEL RLS ADAPTIVE FILTERING FOR SPEECH ENHANCEMENT USING FREQUENCY MODULATION

Abstract

Speech recognition is one of the most challenging applications of signal processing. It is an adaptive filter which can adjust itself its transfer function corresponding to the best adaptive algorithm conveyed by an error or corrupted signal. The Recursive Least Squares (RLS) algorithm has established itself as the “ultimate” adaptive filtering algorithm in the sense that it is the adaptive filter exhibiting the best convergence behaviour. In this paper, we propose a speech imprudent based on Recursive Least Squares (RLS) adaptive filter of speech signals. Experiments were performed on noisy data which was prepared by adding AWGN, to clean speech samples at -2dB, 0dB, 5dB, 10dB and 15dB SNR levels with Frequency modulation. We then compare the noise cancellation performance of proposed RLS algorithm with existing NLMS algorithm in terms of Mean Squared Error (MSE), Pick to Signal to Noise ratio (PSNR) and SNR Loss. Based on the performance evaluation through the simulation Matlab