Emotion perception in speech and music in listeners with hearing impairment
Tim Metcalfe
For his PhD research, Tim investigated the perception of emotion in speech and music in listeners with a cochlear implant or hearing aid. His research included experimental testing and computational modeling of perception of emotion in hearing impaired listeners and was conducted in collaboration with Prof. Guy Brown (Computer Science) and Dr Harriet Crook (Audiology, Sheffield Teaching Hospitals). Tim visited Starkey in Michigan, USA for a research placement during his PhD project. Tim is working researcher in a business context in Cambridge.
Thesis abstract
The everyday tasks of perceiving speech, music and emotional expression via both of these media, are made much more difficult in the case of hearing impairment. Chiefly, this is because relevant acoustic cues are less clearly audible, owing to both hearing loss in itself, and the limitations of available hearing prostheses. This thesis focussed specifically on two such devices, the cochlear implant (CI) and the hearing aid (HA), and asks two overarching questions: how do users approach music and speech perception tasks, and how can performance be improved? The first part of the thesis considered auditory perception of emotion by CI users. In particular, the underlying mechanisms by which this population perform such tasks are poorly understood. This topic was addressed by a series of emotion discrimination experiments, featuring both normal-hearing (CI-simulated) participants and real CI users, in which listeners heard stimuli with processing designed to systematically attenuate different acoustic features. Additionally, a computational modelling approach was utilised in order to estimate participants' listening strategies, and whether or not these were optimal. It was shown that the acoustic features attended to by participants were a compromise of those generally better-preserved by the CI, and those particularly salient for each stimulus. In the latter half of the thesis, the nature of assessment of music perception by hearing-impaired listeners was considered. Speech perception has typically taken precedence in this domain which, it is argued, has left assessment of music perception relatively underdeveloped. This problem was addressed by the creation of a novel, psychoacoustical testing procedure, similar to those typically used with speech. This paradigm was evaluated via listening experiments with both HA users and CI-simulated listeners. In general, the results indicated that the measure produced both valid and reliable results, suggesting the suitability of the procedure as both a clinical and experimental tool. Lastly, the thesis considered the consequences of the various findings for both research and clinical practice, contextualising the results with reference to the primary research questions addressed, and thereby highlighting what there is left to discover.
