Noise Reduction Algorithm for Speech Enhancement

Abstract

Speech communication involves transmitting information through speech between individuals or between individuals and machines in different areas such as speaker identification and automatic speech recognition. However, background noise can hinder effective communication by interfering with speech signals. Therefore, it is necessary to improve speech signals to minimize external disturbances. The process used to generate a more precise voice synthesis from an unclear audio source is called speech enhancement, which employs different algorithms to enhance speech quality. Wavelet transform is used to remove background noise from the messy audio while retaining essential speech information. To eliminate noise from the signal and achieve a clear signal, a semi-soft thresholding approach is employed, which removes chaotic coefficients from the wavelet. The primary objective of this paper is to use semi-soft thresholding to eliminate noise from signal and produce a clear signal. Noise reduction is a critical aspect of speech enhancement that has various applications, including speaker identification, prosthetic devices, VoIP, telepresence, and mobile devices.