Sixth International Conference on Spoken Language Processing (ICSLP 2000)

Beijing, China
October 16-20, 2000

Data-Dependent Kernels in Svm Classification of Speech Patterns

Nathan Smith (1), Mahesan Niranjan (2)

(1) Cambridge University Engineering Dept., UK
(2) Dept of Computer Science, Sheffield University, UK

Support Vector Machines (SVMs) have recently proved to be powerful pattern classification tools with a strong connection to statistical learning theory. One of the hurdles to using SVMs in speech recognition, and a crucial aspect of SVM design in general, is the choice of the kernel function for non-separable data, and the setting of its parameters. This is often based on experience or a potentially costly search. This paper gives some experimental justification for the Fisher kernels proposed in [1]; kernels are obtained and their extra regularisation and use of labelled and unlabelled data discussed. Fisher kernels are derived from generative probability models of the data, and are a first step to implementing kernels for variable length sequences.

Reference

  1. T. Jaakkola and D. Haussler. Exploiting generative models in discriminative classifiers. In M.S. Kearns, S.A. Solla, and D.A. Cohn, editors, Advances in Neural Information Processing Systems 11. MIT Press, 1999.


Full Paper

Bibliographic reference.  Smith, Nathan / Niranjan, Mahesan (2000): "Data-dependent kernels in svm classification of speech patterns", In ICSLP-2000, vol.1, 297-300.