A discriminative GMM technique using product of likelihood gaussians

Abstract

In classification tasks, the error rate is proportional to the commonality among classes. Conventional GMM-based modeling techniques fail to capture the unique features of a class. Classification accuracy can be improved if the modeling technique is able to capture the unique features of each class. For any two models and their corresponding training data, the log-likelihoods may be assumed to be normally distributed; under such an assumption, the amount of overlap between the Gaussian likelihoods may be attributed to the commonality(number of common features) between the classes. This paper proposes a technique to improve the performance of a classifier by capturing the unique features while modeling and filtering common feature vectors during classification. Experiments were conducted on speaker identification task, using speech data of 40 female speakers from NTIMIT corpus, and on a language identification task, using speech data of 2 languages(English and French) from OGI MLTS corpus. The performance of the proposed system is compared with that of conventional GMM technique and significant improvement is noted.