The fusion of forensic speaker verification systems
Huapeng Wang · 2011
The discriminant performance of fused forensic speaker verification systems in likelihood ratios framework was tested. Likelihood ratios were calculated using a multivariate kernel-density (MVKD) approach and the parameters used in this paper are LPC,LPCC and MFCC extracted from the Chinese vowel /a/ in the telephone recordings database, and the reliability of the fused forensic speaker verification systems was measured using the log-likelihood-ratio-cost function.