Within- versus between-speaker acoustic variability in Thai

Abstract

This paper continues our studies of acoustic variability in voice between and within speakers. Previous work indicates that acoustic variability is characterized by the balance between high-frequency harmonic and inharmonic energy in the voice (measured using cepstral peak prominence) and by formant dispersion, regardless of the speaker’s sex, native language, or speaking style. Our recent investigation of the language effect on voice variability surveyed three languages, English, Korean, and White Hmong, which form an organized subset with respect to their linguistic use of F0 and the phonological status of phonation quality. The cross-linguistic comparisons revealed a second tier of language-specific variability that reflects features of the phonology of each language. The present paper adds a tone language, Thai, to the mix. Samples of read speech from fifty Thai speakers (33F, 17M) were evaluated as in our previous studies. Results revealed that variability in Thai voices is primarily accounted for by the same biologically relevant measures observed globally, and further shaped by the language-specific use of F0. These data further support our hypothesis that acoustic variability in voice is governed first by biological factors, secondly by language-specific, cultural factors, and finally by idiosyncrasies of an individual’s anatomy and personal speaking style.