Unvoiced speech recognition using tissue-conductive acoustic sensor

We present the use of stethoscope and silicon NAM (nonaudiblemurmur) microphones in automatic speech recognition. NAMmicrophones are special acoustic sensors, which are attachedbehind the talker's ear and can capture not only normal (audible)speech, but also very quietly uttered speech (nonaudible murmur).As a result, NAM microphones can be applied in automatic speechrecognition systems when privacy is desired in human-machinecommunication. Moreover, NAM microphones show robustness againstnoise and they might be used in special systems (speechrecognition, speech transform, etc.) for sound-impaired people.Using adaptation techniques and a small amount of training data,we achieved for a 20 k dictation task a 93.9% wordaccuracy for nonaudible murmur recognition in a cleanenvironment. In this paper, we also investigate nonaudiblemurmur recognition in noisy environments and the effect of theLombard reflex on nonaudible murmur recognition. We also proposethree methods to integrate audible speech and nonaudible murmurrecognition using a stethoscope NAM microphone with very promisingresults.