Speech Enhancement-Adaptive Algorithms

This study tackles the issue of signal distortion brought on by undesired background noise, which is frequently encountered in real-world situations such as VoIP, cell-phone conversations, teleconferencing, and voice recognition. At the receiving end, the intelligibility or perceived quality of the audio transmission is decreased because of this acoustic noise, which is automatically added to the signal and detected by the microphone. Therefore, when only noisy speech is available, methods for improving speech degraded by uncorrelated additive noise have been extensively investigated in the past and remain an important area of research. Using hearing aids with integrated noise reduction technology or using specific speech enhancement algorithms at the receiver side to improve perceived sound quality.This work presents many adaptive algorithms—namely, Least Mean Square, Normalized Least Mean Square, and Recursive Least Square, that have been developed for noise cancellation in the last few years. Next, each algorithm’s ability to cancel out noise is evaluated by comparing the proportion of noise that is removed from the restored signal. The RLS method produces the best subjective and objective outcomes, but at the expense of significant computational complexity and memory requirements, as demonstrated by the simulation results on MATLAB.