A cross-linguistic investigation of acoustic voice spaces

Abstract

This study examines the extent to which the phonological structure of a language impacts acoustic variation in voice spaces for individuals and populations of speakers. Our recent work on two typologically different languages, American English and Seoul Korean, showed striking similarities in the acoustic spaces derived from Korean and English voices, but also notable differences in the importance of F0 and variability in vowel quality that were related to phrasal intonation patterns present in the Seoul dialect, but not in English. The present study examined eight speakers (5F) of Hmong, a language with tone and phonation contrasts. We performed principal component analysis on scaled values of F0, formants, spectral noise, source spectral shape, and their variability, measured from vowels, and compared results to Korean and English (which lack tone and phonation contrasts). Results revealed both substantial similarities and differences across languages. F0 commonly emerged only for Hmong and Korean voices, and H1-H2 (correlated with phonation) accounted for substantial variation only in Hmong voices. Our findings suggest that both biologically and phonologically relevant factors shape acoustic voice spaces, and point to a potential mechanism for the “own language” advantage in speaker perception. [Work supported by NIH/NSF.]