TWO-MICROPHONE VOICEACTIVITYDETECTIONBASEDON THE HOMOGENEITY OFTHE DIRECTIONOFARRIVALESTIMATES

Kentaro Ishizuka, Hiroshi Sawada, Shoko Araki, Tomohiro Nakatani

Abstract

VoiceActivity Detection (VAD)systems havebeentheobject ofcontinuous research during thelast three decades. While single microphone systems cannot takeadvantage ofcertain spatial properties ofspeech signals, microphone array systems consisting ofmanyelements based onbeamforming techniques canbedifficult toimplement inreality duetocostand complexity issues. Theaimoftheworkdescribed inthis paper was to achieve bothpractical feasibility and spatial discrimination ability. A newapproach isdeveloped fortwomicrophone VAD capable ofprofiting fromtheconcentration of speech energy intime, frequency andspace. Thealgorithm is implemented andcompared withseveral standard VAD algorithms, suchasAFE,AMR andG.729B, andother recently proposed systems, revealing promising results under real-world noise conditions. Themainadvantage oftheproposed approach isitscapacity tooutperform theabovemethods without the needforanyspatial orspectral constraints, which makesitboth versatile andcapable offurther improvement. IndexTerms-Speech processing, Acoustic signal detection, Robustness, Direction ofarrival estimation, Acoustic arrays
BibTeX
@inproceedings{ishizuka2007two,
  author = {Kentaro Ishizuka and Hiroshi Sawada and Shoko Araki and Tomohiro Nakatani},
  title = {{TWO-MICROPHONE VOICEACTIVITYDETECTIONBASEDON THE HOMOGENEITY OFTHE DIRECTIONOFARRIVALESTIMATES}},
  booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing},
  year = {2007},
  doi = {10.1109/icassp.2007.366930},
}

Loading related content...

Figure