Chen, Yue; (2025) Computational simulation of tone and focus perception ------ Sequential processing based on direct phonetic perception at syllable level. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text Yue Chen - PhD Thesis.pdf - Accepted Version Download (5MB) | Preview

Abstract

This thesis uses computational simulations of perception of lexical tone and prosodic focus to address a number of fundamental questions about the basic mechanisms of speech perception: (a) Is it necessary to extract intermediate featural representations of speech prior to perceptual categorisation? (b) Is it necessary to process variability through normalisation? (c) Is it possible to process different communicative functions independently even if they are encoded on the same acoustic dimension? And (d) Is it possible to sequentially process a global prosodic function syllable by syllable in perception? The first study focused on the perception of lexical tone and investigated the need for tonal feature extraction for Mandarin tone recognition. Computational tone recognisers were built with and without feature extraction. Several sets of pitch features were chosen as hypothetical sub-tonal features. The results show that direct tone recognition not only yields better performance than any of the feature extraction schemes, but also requires less computational complexity. It is suggested that prior feature extraction is unlikely to be the operational mechanism of speech perception. The second study focused on the concurrent perception of lexical tone and prosodic focus. Computational models were developed to simulate different perceptual strategies for both tone and focus. The results demonstrate that the neural network-based syllable-by-syllable processing of focus can generate the similar recognition patterns to human perception. With respect to the speech variability, the results show that pitch normalisation schemes do not benefit either tone or focus recognition. Furthermore, there is little difference between recognising tone and focus independently or as a tone-focus combination, suggesting that the two functions are sufficiently separated from each other in their f0 coding, despite sharing the same acoustic dimension.

Download activity - last month

Export as JSONCSVXML

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item