Matching feature distributions for the Robust Speaker Verification
Daniel J. Mashao · Open University of Cape Town (University of Cape Town) · 2004
In this work we improve the performance of a speaker verification system by matching the feature vector distributions obtained when training and testing the system. In particular, we perform experiments using speech that has been degraded by telephone transmission. Speaker Verification experiments are performed on the NIST 2000 database. Significant improvements, above the baseline, are reported.