ISCA Archive SLTU 2012
ISCA Archive SLTU 2012

Subspace-GMM acoustic models for under-resourced languages: feasibility study

Xueru Zhang, Kris Demuynck, Dirk Van Compernolle, Hugo Van hamme

Acoustic model parameter estimation is hampered by a lack of data. To reduce the number of parameters to be estimated, we propose sub-GMM modelling, which constrains the acoustic models to a lowdimensional manifold embedded in the space of Gaussian mixture weights. The manifold model is obtained through non-negative matrix factorization with sparsity constraints. Our preliminary monolingual experiments show that the proposed model is as efficient as clustering the distributions to a smaller set, while it opens perspectives for a new parameter tying technique. In the example, the number of parameters to be estimated per distribution is reduced more than an order of magnitude.

Index Terms: under-resourced languages, manifold, sparsity, non-negative matrix factorization, substructure


Cite as: Zhang, X., Demuynck, K., Compernolle, D.V., Van hamme, H. (2012) Subspace-GMM acoustic models for under-resourced languages: feasibility study. Proc. 3rd Workshop on Spoken Language Technologies for Under-Resourced Languages (SLTU 2012), 1-4

@inproceedings{zhang12_sltu,
  author={Xueru Zhang and Kris Demuynck and Dirk Van Compernolle and Hugo {Van hamme}},
  title={{Subspace-GMM acoustic models for under-resourced languages: feasibility study}},
  year=2012,
  booktitle={Proc. 3rd Workshop on Spoken Language Technologies for Under-Resourced Languages (SLTU 2012)},
  pages={1--4}
}