Speech processing model based on the cochlea temporal characteristics and neural networks with Gabor weights

摘要

Proposes a neurophysiologically based model for acoustical signal processing that employs temporal characteristics of the cochlea response and subsequent neural networks with fixed weights described by Gabor-like functions. With this model, which is based on an integral mechanism, one can naturally explain such effects as sharpening of the spectrum, lateral inhibition, and extraction of formant and pitch for a high-pass signal.