달팽이관 필터뱅크 및 ZCPA 청각모델에 기반한 두 귀의 시간차 추정기법

Abstract

This study presents a method for estimating interaural time difference (ITD) by modeling the human auditory system. The proposed model simulated the response of the human inner ear by signal processing. The response of the cochlea consists of the mechanical vibration of the basilar membrane and the neural transduction of the inner hair cells. The traveling waves on the basilar membrane were represented by a cascade of digital filter sections that function as a series of bandpass filters. We mimicked the neural firing pattern from the inner hair cells by the zero-crossings with peak amplitudes (ZCPA) auditory model in which frequency information of the signal is obtained by zero-crossing intervals. Zero-crossings have been used to find noise-robust speech features. The ITD between two channels was estimated by obtaining the ZCPA pattern of each channel and then calculating an interaural cross-correlogram by center frequencies between the two patterns. We examined the feasibility of the proposed method by simulation, in which binaural speech signals were mixed with normally distributed noise. As a result, the proposed method was able to provide an accurate estimate of ITDs and was robust to Gaussian noise.