Investigating the perception of relative speaker size using synthetic talkers

Abstract

Several previous experiments have investigated the perception of speaker size by presenting listeners with acoustic stimuli that differ in average f0 and/or apparent vocal tract length (i.e., higher formant frequencies overall), and asking listeners to make judgments of relative or absolute speaker size. Typically, these experiments use stimuli with a fixed phonetic content so that the acoustic characteristics of the stimuli may be compared directly. In this experiment, listeners were presented with pairs of vowels produced by synthetic speakers with different apparent vocal tract lengths and the same f0, and were asked to make judgments of relative size. However, listeners were presented with either the same, or different vowels produced by the two speakers. In some cases, differences associated with varying vocal tract lengths were in conflict with differences arising from the formant patterns associated with the differing vowel categories (e.g., lower F1 and F2 for /u/ vs /e/). Results suggest that judgments of relative size are affected by both the vowel categories presented and the apparent vocal tract lengths of the synthetic speakers. That is, more than a simple (category-corrected) vocal tract length estimate is involved in making size judgments for an unknown speaker.