Neural feedback mechanisms for speech-in-noise listening

Abstract

The human auditory system is exceptional at comprehending an individual speaker even in complex acoustic environments. The neural mechanisms involved in such speech-in-noise listening remain, however, poorly explored. This is partly due to the complexity of natural speech as well as of many real-world background sounds. However, methods from statistical learning have recently been employed to measure neural responses to real-world stimuli such as speech. Such studies have, for instance, found a role of subcortical structures in selective attention to speech. Nevertheless, it remains unclear whether these neural processes may inform on “hidden” forms of hearing impairment, such as cochlear synaptopathy, that may cause difficulty with speech-in-noise listening.

In this thesis we investigate the subcortical processing of naturalistic speech and its relation to speech-in-noise performance, using non-invasive clinically applicable measurements as well as computational modelling. We find that the speech-evoked activities in the auditory brainstem (speech-ABR) and in the cochlea (speech-OAE) are larger when a speech signal is attended than when it is ignored. The variability in the modulation of the brainstem response to speech relates to the ability of the subjects to understand speech in noise but does not correlate with proposed measures for cochlear synaptopathy. The computational model results confirm that the speech-ABR is dominated by neural responses in the inferior colliculus and that the response is shaped by nonlinearities and interaction between different harmonics. Regarding the role of the inner ear, or cochlea, we develop a method to record otoacoustic emissions related to speech (speech-OAEs). We show that these speech-OAEs are modulated by selective attention to speech, and that this modulation in turn correlates with a subject’s ability to understand speech in noise.

Our work shows that neural processing in subcortical structures contributes to speech-in-noise listening and can inform on speech-in-noise deficits.