Hearing Impairment Simulator Based on Auditory Excitation Pattern Playback: WHIS

Abstract

A new version of the Wakayama-University Hearing Impairment Simulator (WHIS) was developed based on the idea of auditory excitation pattern (EP) playback instead of direct simulation of loudness recruitment in conventional methods. The EPs of normal hearing (NH) and hearing impaired (HI) listeners were computed using a frame-based version of the gammachirp filterbank (GCFB), in which the cochlear input-output (IO) function was controlled by a parameter called the compression health α. WHIS synthesizes simulated hearing loss (HL) sounds to make the EPs of an NH listener sufficiently close to the EPs of a target HI listener by applying active and passive level reduction to the input signals. Active reduction can be simply formulated as the composite function of the IO function of HI and the inverse NH IO function of NH. Passive reduction was determined to maintain the hearing level shown in the audiogram of the HI listener. Two methods were proposed for sound synthesis: a direct time-varying filter method and a filterbank analysis-synthesis method. WHIS was compared with a Cambridge version of the HL simulator (CamHLS) in terms of differences in the IO functions and spectral distances in the EP spectrograms. WHIS yielded smaller distances than CamHLS, suggesting that EP playback is an effective method for HL simulations.

Effects of age and hearing loss on speech emotion discrimination

“…It is necessary to approach this issue with an auditory model that can simulate the analysis of speech dynamics by the older people. For example, a model with an auditory filterbank and a modulation filterbank can introduce the HL and the degraded temporal response 18,20,21 . The model would help to understand the possible dysfunction and to develop an effective algorithm to support emotion perception.…”

“…The HL simulator WHIS 18 , was applied to the 210 sounds to investigate the effect of peripheral HL on emotion discrimination in the YNH. We simulated the average hearing level of 80-year-old people 22 , i.e., 24,24,27,28,33,48, and 69 dB at 125, 250, 5/8 500, 1000, 2000, 4000, and 8000 Hz, respectively; the average between 500 Hz and 4000 Hz is 34 dB.…”

“…For this purpose, an HL simulator that provides an HL experience for normal-hearing (NH) listeners could be used for the experiments. Recently, WHIS 17,18 , as one such simulator, was designed to maintain relatively low distortion for perceptual experiments and has been used in several speech experiments [19][20][21] .…”