Skip to Main Content
A vocoder based segmentation algorithm for audio signal is presented. The signal to be detected consists of multi-components, which are closely located in time and buried under high level of noise. While detecting the signal components, this method can also estimate the duration of the narrow gaps between them. The combination of a vocoder and binary Walsh transform is the fundamental concept of this method. The location of audio segments and gaps are determined from the output waveform of the vocoder. It is found that using Walsh transform at the synthesis stage of a vocoder provides high capability in separating the signal components from narrow gaps. Although this method is computationally less expensive, the experimental results are satisfactory in the absence of any a priori knowledge about the signal and noise.