Detecting Prolonged Vowel in Spontaneous Speech

K. Niwa, Y. Sagawa, and N. Sugie (Japan)


spontaneous speech, disfluency, filled pause


Spontaneous speech includes grammatical, semantical and contextual disfluency which speaker does not intend. It was reported that voice including disfluency cause recognition error in speech recognition system. To recognize the voice including disfluency correctly, this paper proposes the technique for detecting and removing one kind of disfluency (filled pause) including the prolonged vowel, such as "a-, e-" in Japanese.

Important Links:

Go Back