Paulus, Maximillian; (2021) The influence of channel and source degradations on intelligibility and physiological measurements of effort. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text thesis-paulus.pdf - Accepted Version Download (6MB) | Preview

Abstract

Despite the fact that everyday listening is compromised by acoustic degradations, individuals show a remarkable ability to understand degraded speech. However, recent trends in speech perception research emphasise the cognitive load imposed by degraded speech on both normal-hearing and hearing-impaired listeners. The perception of degraded speech is often studied through channel degradations such as background noise. However, source degradations determined by talkers’ acoustic-phonetic characteristics have been studied to a lesser extent, especially in the context of listening effort models. Similarly, little attention has been given to speaking effort, i.e., effort experienced by talkers when producing speech under channel degradations. This thesis aims to provide a holistic understanding of communication effort, i.e., taking into account both listener and talker factors. Three pupillometry studies are presented. In the first study, speech was recorded for 16 Southern British English speakers and presented to normal-hearing listeners in quiet and in combination with three degradations: noise-vocoding, masking and time-compression. Results showed that acoustic-phonetic talker characteristics predicted intelligibility of degraded speech, but not listening effort, as likely indexed by pupil dilation. In the second study, older hearing-impaired listeners were presented fast time-compressed speech under simulated room acoustics. Intelligibility was kept at high levels. Results showed that both fast speech and reverberant speech were associated with higher listening effort, as suggested by pupillometry. Discrepancies between pupillometry and perceived effort ratings suggest that both methods should be employed in speech perception research to pinpoint processing effort. While findings from the first two studies support models of degraded speech perception, emphasising the relevance of source degradations, they also have methodological implications for pupillometry paradigms. In the third study, pupillometry was combined with a speech production task, aiming to establish an equivalent to listening effort for talkers: speaking effort. Normal-hearing participants were asked to read and produce speech in quiet or in the presence of different types of masking: stationary and modulated speech-shaped noise, and competing-talker masking. Results indicated that while talkers acoustically enhance their speech more under stationary masking, larger pupil dilation associated with competing-speaker masking reflected higher speaking effort. Results from all three studies are discussed in conjunction with models of degraded speech perception and production. Listening effort models are revisited to incorporate pupillometry results from speech production paradigms. Given the new approach of investigating source factors using pupillometry, methodological issues are discussed as well. The main insight provided by this thesis, i.e., the feasibility of applying pupillometry to situations involving listener and talker factors, is suggested to guide future research employing naturalistic conversations.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as XMLCSVJSON

Downloads by country - last 12 months

Export as XMLCSVJSON

Archive Staff Only

View Item