Investigating spoken emotion : the interplay of language and facial expression

This thesis aims to investigate how spoken expressions of emotions are influenced by the characteristics of spoken language and the facial emotion expression. The first three chapters examined how production and perception of emotions differed between Cantonese (tone language) and English (non-tone language). The rationale for this contrast was that the acoustic property of Fundamental Frequency (F0) may be used differently in the production and perception of spoken expressions in tone languages as F0 may be preserved as a linguistic resource for the production of lexical tones. To test this idea, I first developed the Cantonese Audio-visual Emotional Speech (CAVES) database, which was then used as stimuli in all the studies presented in this thesis (Chapter 1). An emotion perception study was then conducted to examine how three groups of participants (Australian English, Malaysian Malay and Hong Kong Cantonese speakers) identified spoken expression of emotions that were produced in either English or Cantonese (Chapter 2). As one of the aims of this study was to disambiguate the effects of language from culture, these participants were selected on the basis that they either shared similarities in language type (non-tone language, Malay and English) or culture (collectivist culture, Cantonese and Malay). The results showed that a greater similarity in emotion perception was observed between those who spoke a similar type of language, as opposed to those who shared a similar culture. This suggests some intergroup differences in emotion perception may be attributable to cross-language differences. Following up on these findings, an acoustic analysis study (Chapter 3) showed that compared to English spoken expression of emotions, Cantonese expressions had less F0 related cues (median and flatter F0 contour) and also the use of F0 cues was different. Taken together, these results show that language characteristics (n F0 usage) interact with the production and perception of spoken expression of emotions. The expression of disgust was used to investigate how facial expressions of emotions affect speech articulation. The rationale for selecting disgust was that the facial expression of disgust involves changes to the mouth region such as closure and retraction of the lips, and these changes are likely to have an impact on speech articulation. To test this idea, an automatic lip segmentation and measurement algorithm was developed to quantify the configuration of the lips from images (Chapter 5). By comparing neutral to disgust expressive speech, the results showed that disgust expressive speech is produced with significantly smaller vertical mouth opening, greater horizontal mouth opening and lower first and second formant frequencies (F1 and F2). Overall, this thesis provides an insight into how aspects of expressive speech may be shaped by specific (language type) and universal (face emotion expression) factors