TY - GEN

T1 - Voice Activity Detector (VAD) based on long-term phonetic features

AU - Barabanov, Andrey

AU - Kocharov, Daniil

AU - Salishev, Sergey

AU - Skrelin, Pavel

AU - Moiseev, Mikhail

N1 - Conference code: 7

PY - 2016

Y1 - 2016

N2 - We propose a VAD using long-term 200 ms Mel frequency band statistics, auditory masking, and pre-trained two level decision tree ensemble based classifier, which allows capturing syllable level structure of speech and discriminating it from com-mon noises. Proposed algorithm demonstrates almost 100% acceptance of clear voice for English, Chinese, Russian, and Polish speech and 100% rejection of sta-tionary noises independently of loudness

AB - We propose a VAD using long-term 200 ms Mel frequency band statistics, auditory masking, and pre-trained two level decision tree ensemble based classifier, which allows capturing syllable level structure of speech and discriminating it from com-mon noises. Proposed algorithm demonstrates almost 100% acceptance of clear voice for English, Chinese, Russian, and Polish speech and 100% rejection of sta-tionary noises independently of loudness

KW - Voice Activity Detector

KW - classification

KW - decision tree ensemble

KW - auditory masking

KW - phonetic features

M3 - Conference contribution

SN - 2529-1092; 978-960-466-161-9

SP - 33

EP - 36

BT - Proceedings of the 7th Tutorial and Research Workshop on Experimental Linguistics ExLing 2016

PB - National and Kapodistrian University of Athens

CY - Athens

T2 - 7th Tutorial and Research Workshop on Experimental Linguistics

Y2 - 27 June 2016 through 2 July 2016

ER -