wrote on Sep 30^th, 2003 at 12:45am:

:D I worked in the polygraph/voice analysis field in the early 70's and performed double blind analyses of tape recordings of interrogations in both English and in a non-Romance language to ascertain truth or deceit with a prototype voice stress analyzer. The research work was paid for, in part, by several federal agencies. The analyses that I performed resulted in an assessment of 83% accuracy, based on post-interrogation determination of the truthfulness of taped responses. I felt good about the 83%, it was comparable to polygraph and a whole lot more efficient, but it wasn't good enough to call this technology "lie detection", which is why we continued to call it voice stress analysis. I read the LVA brochure - very interesting and a quantum leap in the technology. I'd love to know the specifics of their "18 voice parameters", since I was pretty well versed in this stuff once. I find their claims to be a great stretch and am very curious how they achive this. They don't have to worry about patent infringement, I have my own and am not interested in theirs.

Double blind eh?  Where is the study published? I would be most interested in understanding it (especially being an EE).

-Marty