Abstract Views :169 |
PDF Views:2
Authors
Affiliations
1 Department of Electronics and Telecommunication, Rajarambapu Institute of Technology, Rajaramnagar, Sakharale (MS), IN
Source
Digital Signal Processing, Vol 2, No 7 (2010), Pagination: 76-83
Abstract
Speech denoising is very essential for many applications in this era as most of the communication is wireless which is often prone to noise. Noise is unwanted signal and interference of which is unavoidable in any kind of communication systems. Due to presence of noise, speech signal loses its intelligibility and it is very difficult to process such corrupted signal. Main aim of denoising algorithm is to eliminate noise from noisy speech signal without losing its intelligibility. But various algorithms implemented so far by the researchers are not that much effective as they either loose data or generate artificial noise called musical noise. Denoising algorithms experimented so far are not effective as they try to remove noise with same threshold for different frequency bands. Wavelet transform which is recent tool is suitable for non stationary signal such as speech as it facilitates use of different window size and threshold for different frequency bands. In this paper we have implemented wavelet transform for speech denoising. Here focus is on eliminating additive white gaussian noise from speech signal. Denoising is achieved by implementing Symlet with hard and soft threshold methods where threshold selection is done using rigorous sure technique. Results obtained are compared for SNR (Signal to noise ratio) and MSE (Mean square error). Results show that Symlet is successfully implemented for denoising the speech signal.
Keywords
Additive White Gaussian Noise, Denoising, Symlet Wavelet, Thresholding.
