Analysis of tones in Cantonese speech based on the command-response model.

As one of the major Chinese dialects, Cantonese has a tone system consisting of nine lexical tones and three additional changed tones, which is considerably more complex than that of Mandarin. The most important acoustic feature characterizing these tones is the contour of the voice fundamental frequency (the F(0) contour). In this article we present an approach to modeling F(0) contours of Cantonese utterances, based on an extension of the command-response model. Analysis-bysynthesis of F(0) contours of the utterances with a fixed carrier frame, in which a target syllable with each tone type is embedded, shows that each tone type can be represented by a specific pattern (polarity, timing, and amplitude) of tone commands. These patterns are found to be essentially maintained in F(0) contours of the utterances with unconstrained text. With the definition of these tone command patterns, the command-response model not only provides a novel phonological description of tones, but also gives high accuracy of approximations to F(0) contours of Cantonese utterances and allows one to analyze various tonal phenomena in quantitative terms. Quantitative distinctions between various tones are then revealed by statistical analysis of the timing and amplitude of tone commands. Especially, systematic alignment in timing is found between the onsets/offsets of tone commands and the rhyme of a syllable, and hence a set of constraints can be introduced, which together with those on tone command amplitudes and phrase command parameters, is then applied for generating F(0) contours of Cantonese utterances. The validity of the approach is verified by perceptual evaluation of the synthetic speech stimuli with model-generated F(0) contours, both on the intelligibility of tones and on the naturalness of prosody.