| Variability of speaker accent is a challenge for human communication as well as speech technology, including automatic speech recognition, and accent identification. The motivation of this thesis is to contribute to a deeper understanding of accent variation from a cognitive perspective. The goal of this study is to provide perceptual assessment of accent variation in native and nonnative English. Two main issues are: (1) how different components of prosodic affect accent perception, and (2) how listener's accent backgrounds affect accent perception and comprehensibility of the speech. Two sets of perceptual experiments were performed to examine these issues. The results show that listeners can differentiate US southern vs. non-southern accent, based only on prosody (i.e., low-pass filtered, unintelligible speech) but only when the content of the speech is provided (e.g., access to transcripts provides meaning, word sequence, possible phoneme sequence). The trends also indicate that listeners are sensitive to prototypical, conceptual representation of accents, such as "people from New York talk fast". These observations suggest that listeners use both the bottom-up process, based on the acoustic input, and the top-town process, based on their conceptual representation of accents. It is also shown that the listeners' accent background impacts their ability to categorize accents (e.g., native-nonnative accent detection, native accent classification). This finding supports the observation that listeners use both bottom-up processing and bottom-down processing in accent perception. When listeners are familiar with the accent, their mental images of the accent more accurately represents the actual accent characteristics, and therefore listeners are better able to correctly categorize speaker accents. In addition, the results indicate that the comprehensibility of the speech affects accent perception. This suggests that listeners not only use both bottom-up and top-down processes in accent perception, but these processes are multi-dimensional (i.e., language comprehension as well as speech perception). Taken together, these findings contribute to a deeper understanding of the cognitive aspects of accent variation, and its application for speech technology and related fields. The outcomes also point to the complex nature of accent perception, and encourage future studies on accent variation to focus more on the cognitive processes involved. |
