Title :
Interframe dependent hidden Markov model for speech recognition
Author :
Ming, Ji ; Smith, F.J.
Author_Institution :
Dept. of Comput. Sci., Queen´s Univ., Belfast
fDate :
2/3/1994 12:00:00 AM
Abstract :
A hidden Markov model (HMM) with first-order dependent observation densities is presented to account for the statistical dependence between successive frames. A modified Viterbi algorithm is described to optimise jointly the state sequence and dependence relation for the model parameter estimation as well as likelihood calculation. Preliminary experiments show that this approach achieves better performance than the standard multivariate Gaussian HMM
Keywords :
hidden Markov models; parameter estimation; probability; speech recognition; HMM; dependence relation; first-order dependent observation densities; hidden Markov model; interframe dependent model; likelihood calculation; model parameter estimation; modified Viterbi algorithm; speech recognition; state sequence; statistical dependence;
Journal_Title :
Electronics Letters
DOI :
10.1049/el:19940134
