Neural coding of overlapping speech and the effects of hearing loss

Smith, Samuel S. (2020) Neural coding of overlapping speech and the effects of hearing loss. PhD thesis, University of Nottingham.

[img] PDF (Thesis - as examined) - Repository staff only - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Download (8MB)

Abstract

How listeners identify speech that is masked by an overlapping speaker remains a key problem in auditory neuroscience. A mild hearing impairment can profoundly alter a listener’s capacity to identify speech amongst interfering talkers. Existing theories propose that the auditory system uses cues (e.g. pitch, onset/offset timing) to segregate neural representations of overlapping speech. Historically, much attention has been paid to the benefit of fundamental frequency cues to the identification of concurrently presented vowels, but segregation-based models fail to adequately explain human behaviour. Proposed here is a model void of segregation but that embodies an optimal strategy of predicting the combined representations of vowels. The model was able to quantitatively replicate listeners’ identification of concurrent-vowel pairs, an improvement on segregation-based models. This was the case for neural representations simulated from a filter recreation of the auditory periphery, or recorded from multi-channel electrodes in the guinea pig midbrain. The compressive nature of auditory encoding appeared to facilitate a strategy of predicting the combined representation of speech. In addition, neural coding did not support previously proposed segregation cues such as temporally encoded pitch and harmonic beats. Neural recordings were also gathered in response to overlapping vowel-consonants, at variable onset delays, from animals with normal hearing or a noise induced hearing loss. Poorer speech recognition was predicted for the hearing loss data, compounded by an inability to optimally utilise knowledge of interfering speech sounds. Overall, investigation of the auditory system’s goals seems a promising approach to understanding the neural processes underlying speech identification when masked by a competing talker.

Item Type: Thesis (University of Nottingham only) (PhD)
Supervisors: Sumner, Christian J.
Wallace, Mark N.
Akeroyd, Michael A.
Keywords: speech, hearing loss, auditory midbrain, neural coding, computational modelling
Subjects: R Medicine > RC Internal medicine > RC 321 Neuroscience. Biological psychiatry. Neuropsychiatry
Faculties/Schools: UK Campuses > Faculty of Medicine and Health Sciences > School of Medicine
Item ID: 59942
Depositing User: Smith, Samuel
Date Deposited: 28 Oct 2020 11:25
Last Modified: 28 Oct 2020 11:30
URI: http://eprints.nottingham.ac.uk/id/eprint/59942

Actions (Archive Staff Only)

Edit View Edit View