For a practical application of biometrics authentication on cellular phones, a speaker verification system was implemented on a cellular phone and a PDA type mobile terminal. The system has the following three features: 1) a distributed system configuration for connectivity to internet services, 2) text-prompted speaker verification using connected digit patterns for robustness to imposture, and 3) incremental model update for preventing deterioration of accuracy. The system was examined in a two-month test, and effects of the incremental model update were evaluated with those data. Experimental results showed that the average equal error rate over this period was reduced from 8.6% to 4.4% by monthly update, to 2.9% by weekly update, and to 0.9% by daily update.
Cite as: Kato, T., Kawai, H. (2006) A text-prompted distributed speaker verification system implemented on a cellular phone and a mobile terminal. Proc. Interspeech 2006, paper 1896-Tue1CaP.11, doi: 10.21437/Interspeech.2006-177
@inproceedings{kato06_interspeech,
author={Tsuneo Kato and Hisashi Kawai},
title={{A text-prompted distributed speaker verification system implemented on a cellular phone and a mobile terminal}},
year=2006,
booktitle={Proc. Interspeech 2006},
pages={paper 1896-Tue1CaP.11},
doi={10.21437/Interspeech.2006-177}
}