Speaker-Independent Vowel Recognition in Persian Speech

In this paper we discuss the applicability of the kernel-based feature extraction for speaker-independent vowels recognition, focusing on non-linear dimension reduction methods. The Increasing of feature space dimension lead us to improve accuracy of vowels recognition system but we lost realtime system. So, using dimension reduction algorithms, help us to improved accuracy and we study the applicability of this idea to build a quasi-realtime system in Persian speech. In vowels recognition and other similar applications that need a mapping technique that introduces representation of low-dimensional features with enhanced discriminatory power and a proper classifier, able to classify those complex features. In this short paper, we combine nonlinear kernel based mapping of data with support vector machine (SVM) classifier to improve efficiency of system. The proposed here method is compared, in terms of classification accuracy, to other commonly used vowels recognition methods on FarsDat database.