Phonetic Refraction for Speaker Recognition

This paper describes a newly realized highperformance speaker recognition system and examines methods for its improvement. Innovative experiments early this year showed that phone strings are exceptional features for speaker recognition. The original system produced equal error rates less than 11.5% on Switchboard-I audio files. Subsequent research indicates that the equal error rate can be nearly halved by improving the feature extraction and score fusion methods. Pre-processing the speech files to remove cross-talk, improved techniques for combining scores, and gender-specific phone models each reduce the error rates significantly.