Introduction of NIST Speech Processing Evaluations: Speaker Recognition and Language Recognition

Abstract:
We have seen an increase of demand of speaker and language recognition technology. The National Institute of Standards and Technology (NIST) has coordinated evaluations of speaker and language recognition over several years. The purposes of speaker and language recognition are to find out the identity information of the speaker and identify what kind of spoken language from speech signal, respectively. They are popular on security and communication applications. The state-of-the-art speaker and language recognition used signal processing and statistical modeling techniques. They are typically formulated as a hypothesis test problem, including three major components: front-end analysis, statistical modeling and verification decision. I will show a brief overview of speaker and language recognition in this talk. I will share the experience of NIST speech processing evaluations.