Affiliation:
1. Central University of Finance and Economy
Abstract
In the process of speaker recognition, specific algorithm is adopted to classify differentspeakers. In this paper, Multiclass Core Vector Machine (MCVM) is used to the solve speakerrecognition problem. At first, CVM transform quadratic programming of traditional SVM into theMinimum Enclosing Ball (MEB) problem, which significantly reduces the complexity ofcomputation and then, defining an SVM with vector valued output. At last, Experimental results showthat the algorithm is feasible and effective for speaker recognition.
Publisher
Trans Tech Publications, Ltd.
Reference6 articles.
1. W.M. Campbell, D. E. Sturim, and D. A. Reynolds: Supportvector machines using GMM supervectors for speaker verification, IEEE Signal Processing Letters, Vol. 13(2006), pp.308-311.
2. W.M. Campbell: A Covariance kernel for SVM language recognition, Int. Conf. Acoust. Speech and Signal Process 2008, pp.4141-4144.
3. Ivo.W. Tsang, Andras Kocsor, James T. Kwok: Simpler core vector machines with enclosing balls, Proceedings of the Twenty-Fourth International Conference on Machine Learning (2007).
4. Ivor.W. Tsang, James T Kwok, Pak-Ming Cheung: Core vector machines: Fast SVM training on very large datasets, Journal of Machine Learning Research 2005, 6, pp.363-392.
5. F. Aiolli, A. Sperduti: Multiclass Classication with Multi-Prototype Support Vector Machines, Journal of Machine Learning Research (2005), 6, pp.817-850.