ISCA Archive ICSLP 2000
ISCA Archive ICSLP 2000

On auditory-phonetic short-term transformation

René Carré, Liliane Sprenger-Charolles, Souhila Messaoud-Galusi, Willy Serniclaes

In a previous experiment, we showed that the vowel-vowel token [ai] is perceived by adult French listeners as /ai/ for transition durations between the 2 vowels lesser than 200 ms (50% of the response) and as /aEi/ for larger durations. Recall that the /ai/ formant trajectory in the F1-F2 plane crosses the region of the vowel /E/. Further, the 200 ms /ai/-/aEi/ perceptual boundary can be related to syllable duration. The same experiment was tested with children of 6.5, and 13 years old. Six tokens [ai] with different duration transitions (50, 100, 150, 200, 250, 300 ms) were randomly presented 10 times each. The question was: "do you hear 2 or 3 sounds?". Six and half years old children predominantly perceived 2 vowels and their responses were only slightly affected by transition duration. This response pattern decreased progressively with age for large duration transition tokens. At 13 years, the results were closed to the adults ones. At this point of our research, we may suppose that, at the beginning of the acquisition process, the perception is holistic, i.e. global; then, progressively with age and linguistic environment (probably also through reading acquisition), an auditory-phonetic working short term memory of syllabic duration is set up according to the syllabic structure of the French language. This memory could be used to transform the speech signal into symbolic representation (phonemes, features gestures, ...).

Cite as: Carré, R., Sprenger-Charolles, L., Messaoud-Galusi, S., Serniclaes, W. (2000) On auditory-phonetic short-term transformation. Proc. 6th International Conference on Spoken Language Processing (ICSLP 2000), vol. 3, 937-940

  author={René Carré and Liliane Sprenger-Charolles and Souhila Messaoud-Galusi and Willy Serniclaes},
  title={{On auditory-phonetic short-term transformation}},
  booktitle={Proc. 6th International Conference on Spoken Language Processing (ICSLP 2000)},
  pages={vol. 3, 937-940}