A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2017; you can also visit the original URL.
The file type is
This paper presents a new method to detect speech/nonspeech components of a given noisy signal. Employing the combination of binary Walsh basis functions and an analysis-synthesis scheme, the original noisy speech signal is modified first. From the modified signals, the speech components are distinguished from the nonspeech components by using a simple decision scheme. Minimal number of Walsh basis functions to be applied is determined using singular value decomposition (SVD). The maindoi:10.1155/2007/39546 fatcat:s24cgf4i45hibckhzaen7dtnhu