As someone working in that area, I have to disagree. Sure, speaker recognition can work very well under certain circumstances, but there is no unique signature. (that would imply that you could successfully discriminate between any two speakers in the world, irrespective of any other factors)