Mce-Based Training of Subspace Distribution Clustering HMM

XiaoBing Li, LiRong Dai, RenHua Wang

For resource-limited platforms, Subspace Distribution Clustering Hidden Markov Model (SDCHMM) is better than Continuous Density Hidden Markov Model (CDHMM) for its smaller storage and lower computations while maintaining a decent recognition performance. But the normal SDCHMM obtaining method doesn’t ensure the optimality in classifier design. In order to obtain an optimal classifier, a new SDCHMM training algorithm that adjusts the parameters of SDCHMM according to Minimum Classification Error (MCE) criterion is proposed in this paper. Our experimental results on TiDigits and RM tasks show the MCE-based SDCHMM training algorithm provides 15-80% Word Error Rate Reduction (WERR) compared with the normal SDCHMM that is converted from CDHMM.

Cite as: Li, X., Dai, L., Wang, R. (2004) Mce-Based Training of Subspace Distribution Clustering HMM. Proc. International Symposium on Chinese Spoken Language Processing, 113-116

@inproceedings{li04_iscslp,
  author={XiaoBing Li and LiRong Dai and RenHua Wang},
  title={{Mce-Based Training of Subspace Distribution Clustering HMM}},
  year=2004,
  booktitle={Proc. International Symposium on Chinese Spoken Language Processing},
  pages={113--116}
}