A Comprehensive Review of Conventional to Modern Algorithms of Speech Enhancement

Almost all of today’s communication techniques employ speech enhancement. It should go without saying that interference from the environment might reduce the quality of speech during transmission. Digital speech signal processing includes speech enhancement, which seeks to restore the clean speech from the contaminated signal. Different ways for speech enhancement are used depending on the kind of signal deterioration and noise present. The methods discussed in the review vary from single channel speech enhancement methods (statistical model-based approaches, and methods in time and transform domains), multi-channel speech enhancement methods (adaptive noise cancelation and multisensor beamforming), machine learning methods including Convolution Neural Networks, Deep Neural Networks, Residual Convolution Neural Networks, and speech enhancement methods with optimization techniques. The machine leaning algorithms have shown considerable improvement of 98% in enhancing the speech quality of signals.