1000-word speaker-independent continuous-speech recognition using hidden Markov models | IEEE Conference Publication | IEEE Xplore