Saturday, July 01, 2006

Computerized Voice Stress Analysis

The subject of Computerized Voice Stress Analysis (CVSA) has been brought up to me several times in the last month or so. Here is an article that sums up the issues in a non-technical way fairly well.

For the record, I believe that the prevailing opinion in the community is that CVSA should be viewed as an interrogation aid and not a lie-detector. Detecting if someone has micro-tremors in their speech pattern is not the same as detecting deception. Police departments should understand that it may be relatively inexpensive and simple to administer but it is no more reliable than the bit of folk lore about cops using a copy machine and letting the suspect believe that it was a lie detector.

1 comment:

a.a said...

“The research done by the Department of Defense Polygraph Institute shows no more than 50 percent [validity] for the voice stress analyzer,” he said. “Now, the polygraph’s reliability is up to 95 percent.”

I seriously doubt these kind of statistics. How would they be able to establish 95% reliability? In any study of this phenomenon the test subject has to consciously 'lie' and tell the 'truth', the effect would be different from a 'free liar' or a 'free' truth-sayer. I'm doubtful there will be a decent peer reviewed research paper to back that 'statistic'.