1. Home
    Item seperator
  3. Item seperator
    On Non-stationary Hidden Markov Modeling of Speech Signals

On Non-stationary Hidden Markov Modeling of Speech Signals

New Image

This paper describes the reestimation algorithm and its performance when applied to recognize spoken versions of the English alphabet and the digits. In our experiments, the system was trained using speech material from four talks (two males and two females) and tested using different utterances from the same speakers. Preliminary results show that the average recognition accuracy obtained with the proposed system is comparable to that in "Mixture Autoregressive Hidden Markov Models", IEEE Trans. Acoust. Speech, Signal Processing, Vol. ASSP-33, No. 6, pp. 1404-1413, Dec. 1985.