Discrete Fourier Transform, Minimum mean-square error, Noise estimation, Noise reduction, Signal to Noise Ratio, Speech enhancement, Spectral Subtraction,

In an everyday life the speech communication is vivid uses for the hearing impaired and numerous other applications. Speech is a fundamental means of human communication. After over thirty years of research throughout the world, no perfect solution exists to this problem. The objective of our work is to implement a novel speech enhancement algorithm, which offers superior noise reduction over current methods. All speech enhancement systems suffer from distortion or residual noise due to imperfect noise removal. It is often necessary to perform denoising in speech processing system operating in highly noisy environment. Wavelet transform is one of the most promising techniques used in signal processing, due to its ability to decompose signals and to reduce noise having non- stationary characteristics. Wavelet Packet Transform is the advantageous over wavelet transform that it also resolves higher frequency component of the signal. In the present work wavelet thresholding and wavelet packet thresholding algorithm has been used to reduce the noise from the speech signal. A simple SURE threshold method is proposed to compute the optimum threshold value. Mean square error at different values of SNR is computed to evaluate the performance of the proposed method like traditional spectral subtraction, Weiner Filtering method, Spectral Subtraction with MMSE etc. The result obtained is compared with the other speech enhancement algorithms given in various reference papers. In comparison to other reference papers we get improved results in terms of SNR and MSE. Simulation is done on MATLAB platform. RESEARCH PAPER