论文信息 - Energy and entropy based switching algorithm for speech endpoint detection in varying SNR conditions

Energy and entropy based switching algorithm for speech endpoint detection in varying SNR conditions

In this work, we present an algorithm that switches between the energy and the entropy based voice activity detectors (VADs) to provide an improved performance under varying signal to noise ratio (SNR) conditions. The motivation for switching has come from the observed complementary behavior in the noise estimation performances of energy and entropy based voice activity detectors when evaluated over a range of −5 dB < SNR < 30 dB. At lower SNRs the entropy based voice activity detector outperformed the energy based one, whereas at higher SNRs interestingly the opposite trend was noted. The proposed online switching algorithm has a response time of 0.5 second and it achieves the optimal performance of either of the off-line VADs under fixed SNR conditions.

Rohit Sinha | Krishna Chaitanya | R. Sinha | K. Chaitanya

[1] Jeih-Weih Hung,et al. Robust entropy-based endpoint detection for speech recognition in noisy environments , 1998, ICSLP.

[2] Dirk Van Compernolle. Noise adaptation in a hidden Markov model speech recognition system , 1989 .

[3] Biing-Hwang Juang,et al. Fundamentals of speech recognition , 1993, Prentice Hall signal processing series.

[4] Andrzej Drygajlo,et al. Entropy based voice activity detection in very noisy conditions , 2001, INTERSPEECH.

[5] Ahmet M. Kondoz,et al. Digital Speech: Coding for Low Bit Rate Communication Systems , 1995 .

[6] K.-C. Wang,et al. Robust endpoint detection algorithm based on the adaptive band-partitioning spectral entropy in adverse environments , 2005, IEEE Transactions on Speech and Audio Processing.