Existing voice identification systems rely on a person's voiceprint alone before granting security clearance. A typed password can also be used, but this must be easy to remember and so may also easy to guess.
The new system goes further by combining the two. It combines a spoken password and the voiceprint of the speaker to generate a cryptographic key. This is then used to encrypt data automatically.
The link for this article located at newscientist.com is no longer available.