Separation Of Overlapping Speech Signals Based On Auditory Scene Analysis

This paper explains the conception of auditory scene analysis (ASA) and its effect in speech signals processing. We propose a method about separation of overlapping speech signals using wavelet transform based on auditory scene analysis in this paper. This algorithm comprises of overlapping speech analysis stage and speech signals reconstructed stage. Wavelet transform is used for harmonics analysis firstly. Next, we observe and update the probability density function (PDF) of the pitch using non-parametric Kalman filter in order to get the optimum pitch of each sound stream from the maximum of PDF. Each speech signal is reconstructed lastly. A system for separation of overlapping speech signals has been constructed based on above algorithm and some experiments have been done on several sound streams. Other algorithm has been used to compare. We can draw a conclusion based on the results of the experiments that the algorithm presented in mis thesis is quite efficient.