A hybrid GMM/SVM approach to speaker identification
- 13 November 2002
- conference paper
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
Abstract
This paper proposes a classification scheme that incorporatesstatistical models and support vector machines.A hybrid system which appropriately combines the advantagesof both the generative and discriminant modelparadigms is described and experimentally evaluated on atext-independent speaker recognition task in matched andmismatched training and test conditions. Our results provethat the combination is beneficial in terms of performanceand practical in terms of computation. We report...Keywords
This publication has 10 references indexed in Scilit:
- HTIMIT and LLHDB: speech corpora for the study of handset transducer effectsPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2002
- Speaker recognition in two-speaker data: recent results from Dragon SystemsPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2002
- The lincoln speaker recognition system: NIST eval2000Published by International Speech Communication Association ,2000
- Probabilities for SV MachinesPublished by MIT Press ,2000
- Natural Regularization from Generative ModelsPublished by MIT Press ,2000
- Classification by pairwise couplingThe Annals of Statistics, 1998
- The Nature of Statistical Learning TheoryPublished by Springer Science and Business Media LLC ,1995
- Solving Multiclass Learning Problems via Error-Correcting Output CodesJournal of Artificial Intelligence Research, 1995
- Maximum a posteriori estimation for multivariate Gaussian mixture observations of Markov chainsIEEE Transactions on Speech and Audio Processing, 1994
- A training algorithm for optimal margin classifiersPublished by Association for Computing Machinery (ACM) ,1992