Using multiple vector quantization and semicontinuous hidden Markov models for speech recognition | IEEE Conference Publication | IEEE Xplore