Title :
A Novel Data Description Kernel Based on One-Class SVM for Speaker Verification
Author :
Yufeng Shen ; Yingchun Yang
Author_Institution :
Coll. of Comput. Sci. & Technol., Zhejiang Univ., Hangzhou, China
Abstract :
In this paper we develop a novel data description kernel based on one-class SVM (OCSVM-DD kernel) used for text-independent SVM speaker verification. The basic idea of the new kernel is to combine the data description model OCSVM with SVM discriminant classifier. Utterances are firstly mapped to the normal vector of the separating hyperplane in OCSVM model. Then a SVM classifier with linear kernel is applied on those mapped vectors. Experiments results on NIST 2001 SRE database show that the performance of our new kernel is superior to generalized linear discriminative sequence (GLDS) kernel and comparative with UBM-MAP-GMM method.
Keywords :
speaker recognition; support vector machines; NIST 2001 SRE database; UBM-MAP-GMM method; data description kernel; generalized linear discriminative sequence kernel; one-class SVM discriminant classifier; text-independent SVM speaker verification; Computer science; Databases; Educational institutions; Hidden Markov models; Kernel; NIST; Polynomials; Support vector machine classification; Support vector machines; Training data; Kernel One-Class SVM; SVM; Speaker verification;
Conference_Titel :
Acoustics, Speech and Signal Processing, 2007. ICASSP 2007. IEEE International Conference on
Conference_Location :
Honolulu, HI
Print_ISBN :
1-4244-0727-3
DOI :
10.1109/ICASSP.2007.366279