Audio segmentation for speech/non-speech detection using wavelet transform

Abstract

In this project, the noisy audio signal is segmented into speech signal and non-speech signal (that is noise or silence). In our proposed algorithm both the time and frequency domain properties are used to segment the components signals. For simplicity and efficiency, linear time-frequency transformation is used. The wavelet transform is chosen as the segmentation method as it can handle a wide class of problems including those requiring high time resolution and those requiring high- frequency resolution by choosing appropriate wavelet parameters.