Speaker recognition system with good generalization properties

Author

Dustor, Adam ; Klosowski, Piotr ; Izydorczyk, Jacek

Author_Institution

Inst. of Electron., Silesian Univ. of Technol., Gliwice, Poland

fYear

2014

fDate

14-16 April 2014

Firstpage

206

Lastpage

210

Abstract

This paper presents speaker recognition system possessing very good generalization properties. Relatively low equal error rate for speaker verification and high identification rate for identification are achieved for very short training and testing sequences. This behaviour is achieved for the kernel modification of a classic Ho-Kashyap linear classifier. Achieved results for the new approach are compared with results for the classic GMM and VQ techniques. Speech of a moderately good quality from the Polish speech corpus was used for development of recognition system.

Keywords

generalisation (artificial intelligence); learning (artificial intelligence); speaker recognition; Polish speech corpus; VQ techniques; classic GMM; classic Ho-Kashyap linear classifier; generalization properties; speaker identification; speaker recognition system; speaker verification; testing sequences; training; Kernel; Mathematical model; Speaker recognition; Speech; Speech recognition; Training; Vectors; kernel matrix; pattern recognition; speaker recognition; speaker verification;

fLanguage

English

Publisher

ieee

Conference_Titel

Multimedia Computing and Systems (ICMCS), 2014 International Conference on

Conference_Location

Marrakech

Print_ISBN

978-1-4799-3823-0

Type

conf

DOI

10.1109/ICMCS.2014.6911233

Filename

6911233