ATVS-UAM NIST SRE 2010 SYSTEM DESCRIPTION

Javier Gonzalez-Dominguez, Ignacio Lopez-Moreno, Javier Franco-Pedroso, Daniel Ramos, Doroteo T. Toledano, Joaquin Gonzalez-Rodriguez

Abstract: This paper describes the system submitted by ATVS-UAM to the 2010 edition of NIST Speaker Recognition Evaluation (SRE). Instead of focusing on multiple, complex and heavy systems, our submission is based on a fast, light and efficient single system. Sample development results with English SRE08 data are 0.53% EER in tel-tel male data (optimistic since all English speakers in SRE08 are included in the total variability matrices), going up to 3.5% (tel-tel) and 5.1% EER (tel-mic) in pessimistic cross-validation experiments (25% of test speakers totally excluded from development data in each cros-validation set). These results are achieved with an extremely light system in computational resources, running 77 times faster than real time.

Index Terms: speaker recognition, speaker recognition evaluation, factor analysis.

Full Paper