A tutorial on text-independent speaker verification
EURASIP Journal on Applied Signal Processing
| Hi-index | 0.00 |
Speaker verification domain is a research field constantly performing upgrades that are quite difficult to measure. It was demonstrated that system conditions have major influence in the expected results. Database's size is the main concern of this paper. We performed a set of SV experiments over the large SpeechDat II Spanish corpus that shows excellent results. The impact of adaptation, number of mixtures, and segment duration are studied. As our baseline system applies standard parameters in each stage, it can be easily contrasted against other configurations.