A joint synchrony/mean-rate model of auditory speech processing.  

Oleh:

SENEFF, A

Jenis: Article from Journal - ilmiah internasional
Dalam koleksi: Journal of Phonetics vol. 16 no. 1 (Jan. 1988), page 55 - 76.

Ketersediaan

Perpustakaan PKBB Nomor Panggil: 405/JOP/16 Non-tandon: tidak ada Tandon: 1

Lihat Detail Induk

Isi artikel

This paper describes a speech processing system that is based on properties of the human auditory system. A bank of critical-band filters defines the initial spectral analysis. Filter outputs are processed by a model of the nonlinear transduction stage in the cochlea, which accounts for such features as saturation, adaptation and forward masking. The parameters of the model were adjusted to match existing experimental results of the physiology of the auditory periphery. The output of this model is delivered to two parallel channels, each of which produces spectral representations appropriate for distinct subtasks of a speech recognition system. One path yields an overall energy measure for each channel that can be identified with the average rate of neural discharge. The outputs of this path appear to be useful for locating acoustic events and assigning segments to broad phonetic categories. In the other path, the extent of dominance of periodicities at each channel's center frequency is captured by a synchrony measure, which yields a spectral representation with enhanced spectral contrast, relative to the mean-rate spectrogram. The outputs of this stage show distinct formant peaks during sonorant regions, with smooth transitions over time, as well as preserving spectral prominences in the high-frequency region for fricatives and stops.

Opini Anda

Klik untuk menuliskan opini Anda tentang koleksi ini!

Kembali