Is Phoneme Level Better than Word Level for HMM Models in Limited Vocabulary ASR Systems? | IEEE Conference Publication | IEEE Xplore