Abstract
Based on the matrix representation, a new method is proposed to calculate the partial derivatives of the likelihood probability of an HMM with respect to its parameters. Experimental results are presented and a comparison is made with other methods.
References
L. Rabiner, B. H. Juang, Fundamentals of Speech Recognition, Englewood Cliffs, Prentice Hall, 1993, 362–364.
S. J. Young, Competitive training: A connectionist approach to the discriminative training of hidden Markov models, IEE Proc.-I, 138(1991)1, 61–68.