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

Area of Science:

• Phonetics and Phonology
• Computational Linguistics
• Speech Science

Background:

• Cantonese possesses a complex tone system with nine lexical and three changed tones, exceeding Mandarin's complexity.
• The voice fundamental frequency (F0) contour is the primary acoustic feature distinguishing Cantonese tones.

Purpose of the Study:

• To develop and validate an extended command-response model for accurately representing Cantonese F0 contours.
• To provide a quantitative phonological description of Cantonese tones and analyze tonal phenomena.

Main Methods:

• Utilizing an extended command-response model for F0 contour analysis.
• Employing analysis-by-synthesis with fixed carrier frames embedding target syllables.
• Statistical analysis of tone command patterns (polarity, timing, amplitude) in unconstrained utterances.

Main Results:

• Each Cantonese tone type is characterized by a distinct pattern of tone commands.
• The model achieves high accuracy in approximating F0 contours and reveals quantitative distinctions between tones.
• Systematic alignment between tone command timing and syllable rhyme was identified.

Conclusions:

• The extended command-response model offers a novel phonological description of Cantonese tones.
• The model accurately generates F0 contours, validated by perceptual evaluations of intelligibility and naturalness.
• Quantitative analysis of tone command parameters provides insights into Cantonese tonal phenomena.