Speech enhancement for noisy signals using adaptive algorithms

Abstract

Speech enhancement is a signal processing technique used to improve the quality and intelligibility of speech recordings that contain noise or interference. Its main goal is to eliminate unwanted background noise while preserving the clarity and naturalness of the speech signal. This paper provides a comprehensive analysis of three widely used adaptive filtering algorithms, Least Mean Square (LMS), Normalized Least Mean Square (NLMS), and Affine Projection Algorithm (APA). The limitations of LMS, such as slow convergence and sensitivity to input variations, are addressed in this study. By incorporating normalization, NLMS improves convergence speed and robustness to input power levels. The Affine Projection Algorithm (APA) is known for its exceptional performance in non-stationary environments, achieved through subspace projection to estimate optimal filter coefficients, resulting in faster convergence and improved tracking capabilities. In this paper, the algorithms are compared using Signal-to-Noise Ratio (SNR), Mean-Square Error (MSE), and Root-Mean-Square-Error (RMSE) values.