Channel / Handset mismatch evaluation in a biometrie speaker verification using shifted delta cepstral features

Abstract

This paper examines the application of Shifted Delta Cepstral (SDC) features in biometric speaker verification and evaluates its robustness to channel/handset mismatch due by telephone handset variability. SDC features were reported to produce superior performance to delta features in cepstral feature based Language Identification systems. The result of the experiment reflects superior performance of SDC features regarding to delta features in biometric speaker verification using speech samples from Ahumada Spanish database. © Springer-Verlag Berlin Heidelberg 2007.