The human auditory system not only perceives the amplititude, pitch, tone and spatial locations of auditory source but also distills the focused source mixed with complex environmental sounds. We try to use computer technology to model the auditory processing. Neural oscillatory networks are applied to the speech segregation of auditory scene and the directional orthogonal wavelet is applied to the location of the sound source. The main contents of this thesis are expressed as followers: Firstly, the parameters of the gamma tone filters and the weight of the group layer are effectively modified based on concerned speech signal, which get improved result. Secondly, we utilize periodic cardinal interpolatory wavelet as directional orthogonal wavelet basis, which is sensitive to the direction, to locate the sound source, the result is presented.
|