ISCA Archive WSLP 2003
ISCA Archive WSLP 2003

Begin-end detection using vowel onset points

S. R. Mahadeva Prasanna, Jinu Mariam Zachariah, B. Yegnanarayana

This paper proposes a method for detecting begin and end points of a speech ntterance using the knowledge of Vowel Onset Points (VOPs). VOP is defined as the instant at which the onset of vowel takes place. An algorithm for VOP detection in continuous speech is discussed. VOP helps in overcoming the difficulties present in coming up with multiple thresholds followed in most of the existing begin-end detection algorithms. The VOP of the first vowel is used as an anchor point for further analysis to detect the begin of the speech utterance. Similarly, the VOP of the last vowel is used as an anchor point for detecting the end point. The performance of the proposed begin-end detection algorithm is compared with the existing energy-based approach by conducting text-dependent speaker verification experiments. The speaker verification system nsing the knowledge of VOP for begin-end detection shows a significant improvement in the performance.

Cite as: Prasanna, S.R.M., Zachariah, J.M., Yegnanarayana, B. (2003) Begin-end detection using vowel onset points. Proc. Workshop on Spoken Language Processing, 33-39

  author={S. R. Mahadeva Prasanna and Jinu Mariam Zachariah and B. Yegnanarayana},
  title={{Begin-end detection using vowel onset points}},
  booktitle={Proc. Workshop on Spoken Language Processing},