Third International Conference on Spoken Language Processing (ICSLP 94)

Yokohama, Japan
September 18-22, 1994

Perception for VCV Speech Uttered Simultaneously or Sequentially by Two Talkers

Kazuhiko Kakehi (1), Kazumi Kato (2)

(1) Graduate School of Human Informatics, Nagoya University, Nagoya, Japan
(2) NTT Basic Research Laboratories, Kanagawa, Japan

We investigated robust perception of speech using different talkers VCVs uttered simultaneously or sequentially. We conducted two experiments: one to investigate the perception of the number of talkers and the other to investigate the perception of intervocalic consonants. The experimental results are as follows : (1) Even when two talkers are perceived in a VCV stimulus, the extracted phoneme features are perceptually integrated regardless of the talker source information. (2) Phoneme features in a pre-closure part in double talk contribute to intervocalic stop consonant perception regardless of talker source information. This means that the talker source information is used to determine speech portion where phoneme features are extracted, and that these features are integrated to perceive a phoneme regardless of the talker source information.

Full Paper

Bibliographic reference.  Kakehi, Kazuhiko / Kato, Kazumi (1994): "Perception for VCV speech uttered simultaneously or sequentially by two talkers", In ICSLP-1994, 1411-1414.