Kernel Based Clustering for Multiclass Data

In this paper we address the issues in construction of discrete hidden Markov models (HMMs) in the feature space of Mercer kernels. The kernel space HMMs are suitable for complex pattern recognition tasks that involve varying length patterns as in speech recognition. The main issues addressed are related to clustering in the kernel feature space for large data sets consisting of the data of multiple classes. Convergence of kernel based clustering method [1] is slow when the size of the data set is large. We consider an approach in which the multiclass data set is partitioned into subsets, and clustering for the data in each subset is done separately. We consider two methods for partitioning the multiclass data. In the all-class-data method, the partitioning is done in such a way that each subset contains a part of the total data set of each class. In the class-wise-data method, a subset contains the total data of only one class. We study the performance of the two methods on kernel based clustering used to build discrete HMMs in the kernel feature space for recognition of spoken utterances of letters in E-set of English alphabet.