Application of Markov chains to speech recognition

The authors describe how discrete Markov chains may be applied to speech recognition. In this application, a spectral vector is modelled by a state of a Markov chain, and an utterance is viewed as a sequence of observed states. Experiment showed that a speech recogniser based on this Markov model not only outperforms the HMM recognisers tested for comparison but also offers a saving in computation time.