Authors : S. Karthikeyan, P. Ganesh Kumar and K. Saran
Abstract: The proposed research in the denoising of speech signal is based on the concepts of Wavelet Thresholding by imposing quantum parameters. The idea of signal denoising is to preserve the signal features while reducing the noise level. Various denoising approaches exist in which wavelet-based point wise thresholding approaches are extensively adopted in many application fields like speech processing applications, medical applications, etc. For signal denoising based on wavelet thresholding, there are two decisive aspects, namely, the use of a proper thresholding function and the estimate of the noise standard deviation. Both greatly manipulate the quality of the denoised signal. In this study, a simple wavelet-based denoising approach is performed for real time speech signal which uses the modified linear expansion of thresholds based on Stein’s Unbiased Risk Estimation (SURE) and the noise standard deviation estimation depending on the number of vanishing moments of the wavelet transform. Investigational results demonstrate that higher Signal to Noise Ratio (SNR) with lower mean square error.
S. Karthikeyan, P. Ganesh Kumar and K. Saran, 2013. A Noise Reduction Approach for Speech Signal Based on Stein’s Unbiased Risk Estimate. Asian Journal of Information Technology, 12: 154-159.