Acoustic modeling of subword units for speech recognition | IEEE Conference Publication | IEEE Xplore