The idea seems simple: build a Web camera that will identify military veterans who may be struggling with depression, anxiety, or post-traumatic stress by reading their faces.
DARPA is creating a computer program that analyzes voice tones and 68 points on the face that indicate happiness or tension, says Louis-Philippe Morency, research assistant professor at the University of Southern California, in Los Angeles.
"Clinicians have a really hard time detecting these things. This takes what is subjective and makes it objective,” he says.
In the future, Dr. Morency envisions the software, known as Detection and Computational Analysis of Psychological Signals (DCAPS), being used to train young adults diagnosed with autism to better read nonverbal cues so that they can prepare for, say, job interviews. It could also make reaching out for help easier, Morency hopes, because it turns out that test patients report feeling more comfortable speaking with an avatar than they do with a real human being.
“It may be less fear of being judged,” he says. “They can show more sadness, or be more themselves.”
