Kharvey: I could come back to this, since it basically comes down to your FBI symposium members' surety versus the opinion of another professional voice recognition specialist. They want absolute surety, because they have to make cases before judges to get warrants before they break into houses to plant bugs or hack computers or put GPS trackers on or in things. It becomes a liability if they're wrong, and of course we should always go for better tools. But that still doesn't mean that comparing dissimilar samples doesn't come up with viable data or verifiable outcomes. It just doesn't. You say he did it wrong, but you're wrong in the use of the word, implying less than optimal is difinitively useless or totally incorrect. That's your opinion, not even the opinion of the FBI symposium members you cite who stop at stating what they would rather have for satisfaction.
Now I'm not going to do this.
http://i.imgur.com/Iuzfg.jpg This is what I will not do. I'm done. But be ware. The chew toy is yours. For now.