A biophysical model of the human cochlea for speech stimulus using STFT

Abstract

This paper presents a new approach to an auditory model which matches closely the response patterns in physiological data for single tone inputs. Including the biophysical complexity of the wave motion in the cochlea and considering all terms of the motion equation, the new model can evaluate the auditory spectrum using the Basilar membrane (BM) displacement and the inner hair cells' (IHCs) firing rate for all input signals, including speech. We employ the partial differential motion equations of the BM and its parameters measured for the human auditory system, and design an algorithm which uses short time Fourier transform (STFT) to compute the output for speech stimulus. The idea is to isolate the input signal in the vicinity of a time-window and try to follow the changes in its frequencies and their influences on the signal perceived by the auditory system. The new model includes the nonlinearity action of outer hair cells (OHCs) and provides a new auditory spectrum for speech inputs in the real time domain which reflects a proper view of propagating signal in the cochlea. Despite most of the previous models this model can track the effects of high formant frequencies in the human cochlea as well. This model is a new signal processing tool for studying the response of the auditory system to transient signals which is highly demanded in various speech enhancement and audio coding algorithms