[en] This work present a novel approach to track a
specific speaker among multiple using the Minimum Variance Distortionless Response (MVDR) beamforming and fuzzy logic ruled based classification for speaker recognition. The Sound sources localization is performed with an improve delay and sum beamforming (DSB) computation methodology. Our proposed
hybrid algorithm computes first the Generalized Cross Correlation (GCC) to create a reduced search spectrum for the DSB algorithm. This methodology reduces by more than 70 % the DSB localization computation burden. Moreover for high frequencies Sound sources beamforming, the DSB will be
preferred to the MVDR for logic and power consumption reduction.
Disciplines :
Electrical & electronics engineering Library & information sciences
Author, co-author :
Ibala, Christian
Astapov, Sergei
Bettens, Frédéric ; Université de Mons > Faculté Polytechnique > Information, Signal et Intelligence artificielle
Escobar Juzga, Fernando ; Université de Mons > Faculté Polytechnique > Electronique et Microélectronique
Chang, Xin ; Université de Mons > Faculté Polytechnique > Electronique et Microélectronique
Valderrama, Carlos ; Université de Mons > Faculté Polytechnique > Electronique et Microélectronique
Riid, Andri
Language :
English
Title :
Combining Multiple Sound Sources Localization Hybrid Algorithm and Fuzzy Rule Based Classification for Real-time Speaker Tracking Application
Publication date :
31 March 2013
Journal title :
International Journal of Microelectronics and Computer Science
ISSN :
2080-8755
Publisher :
Lodz University Technology, Lódz, Poland
Volume :
4
Issue :
1
Pages :
12-25
Peer reviewed :
Peer Reviewed verified by ORBi
Research unit :
F105 - Information, Signal et Intelligence artificielle F109 - Electronique et Microélectronique