ISCA Archive Interspeech 2008
ISCA Archive Interspeech 2008

Voice activity detection using modified Wigner-ville distribution

Lakshmish Kaushik, Douglas O'Shaughnessy

This paper introduces a new method of voice activity detection (VAD) using modified Wigner-Ville distribution. Modified Wigner-Ville distribution can track the speech regions efficiently in noisy environments. Fourier Transform (FT) and Discrete Cosine Transform (DCT) variants of Wigner-Ville distribution based, voice activity detection schemes are presented. The efficient time frequency spectral representation of Wigner-Ville is exploited for increasing the accuracy of VAD decision. The techniques are tested in three different standard noisy conditions (babble, gaussian, vehicle) with different levels of degradation. The proposed technique significantly outperforms the existing techniques in literature.

doi: 10.21437/Interspeech.2008-638

Cite as: Kaushik, L., O'Shaughnessy, D. (2008) Voice activity detection using modified Wigner-ville distribution. Proc. Interspeech 2008, 2574-2577, doi: 10.21437/Interspeech.2008-638

  author={Lakshmish Kaushik and Douglas O'Shaughnessy},
  title={{Voice activity detection using modified Wigner-ville distribution}},
  booktitle={Proc. Interspeech 2008},