Investigating within and between talker variability in the optical characteristics of American English vowels.

Abstract

Previous research investigating vowel production provides evidence of substantial within and cross-talker variation in acoustic characteristics. The current study is designed to investigate whether similar within and cross-talker variation exists in the optical characteristics of American English vowels. Additionally, potential shared acoustic-optical variation across tokens will also be examined. Measures taken from three-dimensional motion data for a set of 13 markers positioned around the lips, cheeks, and chin sampled at 100 frames/s with millimeter spatial resolution along with simultaneous video (50 frames/s) and audio for each of ten repetitions of 11 vowels spoken by multiple talkers will be reported. Preliminary analysis of motion and acoustic data from12 repetitions of 11 vowels spoken by one male talker provides evidence of substantial within vowel variation in both vertical and horizontal lip separations measured at the midpoint of the vowel as determined by analysis of the acoustic signal. Correlations were observed between horizontal lip separation and second formant frequency (r = 0.48) and third formant frequency (r = 0.66) at the vowel midpoint. No correlations were obtained with vertical lip separation measure. Results will be discussed in terms of potential implications for talker intelligibility in visual only presentation conditions.