Voice quality is due in part to patterns of vibration of a speaker's vocal folds inside the larynx. In some languages different voice qualities can distinguish word meanings, so each speaker must control multiple patterns of vibration. This project studies how speakers of different languages produce such voice contrasts when speaking, and how they perceive them when listening. Physiological recordings, acoustic measurements, and perceptual responses are analyzed to uncover the overall multi-dimensional phonetic space for voice, and to determine which subpart of that overall space each language uses.
The interdisciplinary research team for the project offers a unique approach to understanding linguistic voice quality, combining linguistic phonetics, voice science, and electrical engineering. Semi-automation of acoustic measurements of voice samples, with new algorithms for recovering voice properties from standard audio recordings, allows large-scale data collection and analysis. These tools will be made available publicly for use by other researchers. The cross-language perspective of the project takes advantage of the many languages spoken in the Los Angeles area. The results of the project will be relevant for (1) describing languages, including documenting endangered languages; (2) high-quality synthesis of a wide range of normal voice qualities; (3) use of voice information in automatic speech recognition; (4) speaker (voice) recognition; (5) sociolinguistic study of voice quality variation as a function of linguistic/regional/social groups, including in bilingual populations; (6) study of voice quality variation with affect, emotion, etc.; (7) second-language learning of language-specific voice quality.
