Voiceprints...individual identification through speech patterns

Researchers from the University of Hertfordshire are working on speech pattern recognition. A mathematical modelling of voice pattern will be able to improve security by identifying individuals in the future, like fingerprints nowadays do.

"Once we receive our sentence from a speaker we select their voice prints, and then we select from that features which represent the voice prints. We use these special features to build a statistical model representing the individual speaker and we store them. Then at the time of testing – when somebody for example wants use his voice to access his bank over the telephone – we receive the speech data from the user, extract the features the same way and compare these against the reference model we have for that individual. And then we decide whether the user is genuine or an impostor." Dr Aladdin Ariyeeinia, University of Hertfordshire.

Useful links:

Biometrics White Paper
VoiceBiometrics