Objective estimation of loudness growth using tone burst evoked auditory responses

Abstract

Human perception of loudness as function of a wide range of stimulus intensity is hypothesized to vary in complex ways that cannot be accounted for simply by the listener's threshold and discomfort level. The ability to estimate loudness growth as function of stimulus intensity using an objective procedure could potentially allow for customized non-linear hearing aid fitting for patients unable to perform standard subjective auditory tests. Some researchers have investigated the feasibility of using evoked auditory brainstem responses (ABR) and otoacoustic emissions (OAE) as physiological predictors of loudness growth in humans. Results from OAE studies were promising, but lacked an investigation into the selection of its parameters and a measure of the methodological performance in hearing impaired listeners. Previous studies using ABR, on the other hand, did not use frequency specific stimuli and were based on subjective segmentation techniques that did not take into account the residual noise levels in the averaged waveform. The work in this thesis aims at investigating and improving the use of simultaneously recorded tone burst OAE and ABR for assessing frequency specific loudness growth in normal-hearing and hearing-impaired listeners. First, an optimal set of parameters for the OAE estimator is selected based on mean square error criteria. Second, a new signal-to-noise-ratio (SNR) estimation procedure is developed that attempts to take into account non-stationary noise sources present in ABR recordings. Third, several objective ABR segmentation and noise control procedures are tested for estimating loudness growth. Additionally, SNR estimates are compared with ABR loudness growth estimation errors in an attempt to examine dependencies between residual noise levels in averaged ABR waveforms and estimation performance. Results show that for normal listeners and with 1 kHz tone bursts, both OAE and ABR loudness growth estimators can perform within range of standard psychoacoustical procedures. With hearing impaired listeners, or with 4 kHz tone bursts, the performance of the OAE estimator produced a higher mean-square error with respect to a psychoacoustical measure of loudness than most of the ABR estimators. The performance of ABR estimators with hearing-impaired listeners is close to, but not as accurate as the standard psychoacoustical procedures.