Back to list

Evaluation and Extension of a Binaural Loudness-Scaling Method for Cochlear-Implant Listeners

Authors Loss, T.
Year 2016
Thesis Type Master's thesis
Topic Audio Signal Processing
Abstract Cochlear implants (CIs) are hearing aids converting acoustic information into electrical signals, which are then used to stimulate the neurons within the cochlea. CIs are successful in giving back auditory perception to the deaf. The link between the stimulating electrical signals and the perceived loudness is complex and not yet fully understood with relation to bilateral stimulation. This master thesis aimed at evaluating and extending an existing method for binaural loudness scaling in CI listeners, which models the subjectively perceived loudness as a function of current levels at binaural electric stimulation. In order to detect possible improvements, data from previous CI loudness studies was examined. Sequential effects were found to play a major role and influenced variability in CI loudness experiments. Large differences between succeeding stimuli increased variability of responses and non-randomly collected pre-test data deviated from data of the main procedure. Thereby, goodness of fits of final loudness growth functions was impaired. An adapted procedure was developed taking these findings into account. Step sizes were restricted to certain percentages of the dynamic range (DR) in order to reduce variability of responses. Additionally, pre-test data was discarded for the final loudness growth function fit. The adapted procedure was then evaluated by means of a normal-hearing (NH) procedure, in which a Gaussian-enveloped tone (GET) vocoder was used to simulate CI stimulation. Results showed that major improvements in the goodness of fits of loudness growth functions can be achieved by discarding pre-test data and using a robust fit and 40 % DR step size restriction. Additionally, the effect of step size restriction is most significant for sequential test settings, in which electrodes for both ears are tested separately. The NH procedure developed in this thesis is suitable for simulating CI signals and doing check-ups of possible improvements. Although the exact progression of the loudness growth function cannot be simulated due to a highly variable loudness perception of CI listeners, the simulation can successfully mimic the DR of subjects and can be used to simulate psychoacoustic effects on a cognitive level. By using a GET vocoder the temporal structure of the CI stimulation signal can be taken into account, however, the vocoder is only suitable for low pulse rates as used in the loudness-scaling method. All in all, the adapted loudness-scaling method includes major improvements which lead to a more precise loudness scaling.
Supervisors Majdak, P.