To advance the understanding of electric hearing, from both a theoretical and practical perspective, the present study employs an engineering approach to examine whether a fundamental stochastic link exists between neural stimulation and perception. Through the use of custom-developed psychophysics software, temporal gap-detection experiments were carried out and compared with simulation results of a theoretical model. The results are informative, and the suggested modeling principles may be a step forward to a clearer understanding of how the hearing system perceives temporal stimuli. To enable the implementation of psycho-electric experiments involving cochlear implants, a software framework was developed for Matlab version 6.5, called the Psychoacoustics Toolbox, which can present stimuli either acoustically or (for interfacing with cochlear implants) using Cochlear Ltd. hardware. This toolbox facilitates easy setup of experiments based on extensible markup language (XML) templates, and allows for both adaptivestaircase procedures and presentation of a fixed set of stimuli to a participant. Multi-track interleaving of stimuli is also supported, as put forward by Jesteadt (1980), to allow for capturing of subjective responses (such as loudness perception). As part of this research, experiments were performed with three subjects, with a total of four cochlear implants. For the temporal gap-detection experiments, the rate of electrical stimulation varied over a range from 100 to 2700 pulses per second; both periodic stimulus sequences and stimuli reflecting a dead-time-modified Poisson process were used. Also, three spatially distinct stimulation sites were used with each implant to allow comparison among basal, central and apical cochlear responses. A biologically plausible psychophysical model (in contrast with a phenomenological one) was developed for predicting temporal gap-detection thresholds in electric hearing. The model was applied to both periodic and Poisson stimuli, but can easily be used with other kinds of stimuli. For comparison with experimental results, model predictions were made over the same range of stimulus rates. As a starting point, the model takes the neural stimuli, runs them through a neural filter, and then draws statistical interspike-interval (ISI) distribution data from the generated spikes. From the ISI statistics, psychometric curves can be calculated using the principles of Green and Swets (1966), from which predictions can be made for threshold measurements based on the percentage-correct mark for the specific experimental setup. With a model in place, simulations were executed to compare the model results with experimental measurements. In addition to the simulations, mathematical equations for the periodic types of stimuli were derived, given that numerical calculations could be made with higher computational e ciency for this kind of stimulus. These equations allowed for an investigation into the implications of varying the values of different neuron-model parameters. Clear similarities were found between the shapes of gap-threshold curves for experimental and modeled data, and qualitative links have been identified between model parameters and features recognized in threshold curves. For periodic stimuli, quantitative predictions of gap thresholds are close to experimental ones, although measured values cover a larger range. The results of experimental measurements using Poisson stimuli are generally somewhat larger than model predictions, although the shapes of the curves show resemblance. A possible explanation is that participants may find decision tasks involving Poisson stimuli, as opposed to periodic stimuli, confusing. Overall, model predictions and experimental results show close correspondence, suggesting Department of Electrical, Electronic and Computer Engineering. University of Pretoria. ii that the principles underlying the model are fundamentally correct. Copyright 2007, University of Pretoria. All rights reserved. The copyright in this work vests in the University of Pretoria. No part of this work may be reproduced or transmitted in any form or by any means, without the prior written permission of the University of Pretoria. Please cite as follows: Geldenhuys, TA 2007, Temporal gap detection in electric hearing : modelling and experiments, MEng dissertation, University of Pretoria, Pretoria, viewed yymmdd < http://upetd.up.ac.za/thesis/available/etd-02232012-131459 / > E1091/gm
Dissertation (MEng)--University of Pretoria, 2012.