Voice activity detection based on augmented statistical noise suppression
Yasunari Obuchi, Ryu Takeda, Naoyuki Kanda · Asia-Pacific Signal and Information Processing Association Annual Summit and Conference · 2012
A new voice activity detection (VAD) algorithm using augmented statistical noise suppression is introduced. Statistical noise suppression is an effective tool for speech processing under noisy conditions. It achieves the best VAD performance when the noise suppression is augmented in various ways. The speech distortion, which is usually a severe side effect of strong noise suppression, does not affect the VAD performance, and the correctly estimated signal power provides accurate detection of speech. The performance of the proposed algorithm is evaluated using CENSREC-1-C public database, and it is confirmed that the proposed algorithm outperforms other algorithms such as the switching Kalman filter-based VAD.