Performance Evaluation and Analysis of Thresholding Techniques for Speech Denoising and Enhancement

This paper evaluates and analyse the performance of numerous thresholding techniques in context of speech denoising and enhancement. The performance of each technique is critically compared across a spectrum of thresholds - Hard, Soft, Hard-soft, Super soft, semi soft, Universal, Minimax, Adaptive, and Bayesian thresholding methods. Performance comparisons were based on metrics such as signal to noise ratio exploring the efficacy of each thresholding method. Results obtained from each method recap the strengths and limitations of each thresholding methodology, and thus the applicability of each with respect to noise environments or signal characteristics. This study provides important information for selecting an appropriate thresholding method for denoised and enhanced effective speech signal.