VOL. 55, BOOK 1, PART B, 2017, pp. 60 – 72 Full text (En)
Author: Erwan Pépiot
Affiliation: University Paris 8
Abstract
The present study focuses on gender identification obtained from a speech experiment. It was conducted jointly on 25 Parisian French native listeners with French stimuli, and 25 American English listeners with English stimuli. Extracts from (C)VCV words and pseudo-words were presented to the participants, with the use of the gating paradigm. The listeners had to identify the speaker’s gender and indicate their degree of certainty. In both languages, the percentages of correct identification were significantly above chance for initial voiceless consonants, and close to 100 % with initial vowels. An acoustic analysis performed on the stimuli showed that American English and French listeners did not use identical strategies: mean f0 and voice quality (H1-H2) seemed to have more influence on American English listeners’ judgements than on those of French listeners who appeared to rely more on vowel formant frequencies and f0 range.
Key words: Gender identification from speech, cross-gender acoustic differences, cross-language variations, Parisian French listeners, American English listeners
