Class-based Gaussian selection for efficient decoding in PTM HMMs

A new Gaussian selection (GS) method is presented for fast decoding in phonetic tied-mixture (PTM) hidden Markov models (HMMs). For efficient likelihood computation, a constraint is imposed on the context-dependent weights as well as the number of Gaussians. Experimental results demonstrate the superiority of the proposed method over conventional GS methods.

[1]  Mark J. F. Gales,et al.  State-based Gaussian selection in large vocabulary continuous speech recognition using HMMs , 1999, IEEE Trans. Speech Audio Process..

[2]  Enrico Bocchieri,et al.  Vector quantization for the efficient computation of continuous density likelihoods , 1993, 1993 IEEE International Conference on Acoustics, Speech, and Signal Processing.