for National Geographic News
As debate about the Bush Administration's domestic surveillance activities reverberates across Capitol Hill, the next generation of surveillance tools is already under development (related photos: high-tech surveillance).
The eavesdroppers of the future may well have microphones and processing chips in place of ears and brains.
Advanced software programs augment human analysts' efforts to stop terrorist attacks and identify criminal deceptions, security researchers say.
The aim of such programs is to mine critical intelligence from written communications, audio intercepts, or videotapesor all three media simultaneously.
Efficient, automated interpretation of raw data would spare human analysts from trying to pinpoint clues amid haystacks of words, sounds, and images.
"In the past it's all been done by individuals poring over transcripts," said psychologist James Pennebaker of the University of Texas at Austin.
Unlike computers, "humans have limits to vigilance, limits to coalescing enormous amounts of information," said information-systems expert Tom Meservy of the University of Arizona in Tucson.
While wiretaps can record scores of conversations simultaneously, "having an agent listen to them all is impractical," Meservy said.
"But if you could focus in on a subset of [each] conversation that would help focus our resources."
Moreover, he said, a machine's objectivity could prevent "biases and prejudices" from leading investigators to inaccurate conclusions about a suspect.
Already computers can usually determine someone's sex and other characteristics (age, home region, and so on) by analyzing a voice recording, said researcher Venkata R. Gadde of SRI International, a Menlo Park, California-based independent research institute.
With greater than 98 percent accuracy, he said, automated analysis of sound wave frequencies can sort audio clips into those spoken by a child and those spoken by an adultas shown in experiments by Gadde several years ago.
SOURCES AND RELATED WEB SITES