Acoustic modeling for speech recognition based on spotting of phonetic units | IEEE Conference Publication | IEEE Xplore