Speaker adaptation using semi-continuous hidden Markov models | IEEE Conference Publication | IEEE Xplore