Efficient Speaker Recognition based on Multi-class Twin Support Vector Machines and GMMs

This paper proposes a new approach for text-independent speaker recognition using twin support vector machines (TWSVMs) and feature extraction based on Gaussian mixture models (GMMs). Because of the perfect discriminability and the ability of managing large scale dataset, the proposed approach performs better than the traditional support vector machines (SVMs) on Ahumada Biometric Database and Gaudi Biometric Database.