The research blurbs below are of possible interest to CPaT. Blurbs are posted, most recent first! If interested in the topic, just search the web page for that item….
4/6 – Digital Shrinks Find Depressed Faces and Body Language
4/6 – Scientists Use AI Cobweb Analysis to Determine Spider Species
From ACM Tech News, April 3, 2013: Digital Shrinks Find Depressed Faces and Body Language (posted by Judy)
New Scientist (03/28/13) Niall Firth
Automatic systems that analyze gestures and facial expressions might improve the challenging task of diagnosing depression. One such system, SimSensei, is a digital avatar that interviews people to determine their state of mind using facial-recognition technology and depth-sensing cameras integrated with Microsoft Kinect to capture and analyze body language. University of Southern California researchers identified characteristic movements that signal possible depression by interviewing non-depressed volunteers and those who had been diagnosed with depression or post-traumatic stress disorder. Focusing a high-definition webcam on a subject’s face and tracking body movements with Kinect, the team noted that depressed people are more likely to fidget and drop their gaze, and smile less than average. Another automatic depression diagnosis system is underway at the University of Canberra, which is working with the Black Dog Institute to develop a machine-vision system that looks for distinctive facial expressions, slower-than-usual blinking, and specific upper-body movements. Meanwhile, University of Pittsburgh researchers are studying changes in facial expression as a person receives depression treatment. These new systems will be tested in October, when global researchers gather at the ACM Multimedia conference in Barcelona, Spain, to participate in a contest to discover the most accurate depression diagnosis system.
PhysOrg.com (03/28/13) Nancy OwanoPattern recognition can be used to determine which species of spider has spun a cobweb. The University of Las Palmas de Gran Canaria’s Carlos Travieso and colleagues have used an artificial intelligence (AI) recognition system with special software to analyze the images provided from a spider expert. The researchers applied pattern-recognition techniques to the pictures until they were confident they could reliably identify the type of web. The team was able to achieve an accuracy rate of 99.6 percent. Their method involved principal component analysis (PCA), independent component analysis, Discrete Cosine Transform, Wavelet Transform, and discriminative common vectors such as features extractors, with PCA providing the best performance. The research tool extends from a way of measuring spider species to understanding and measuring biodiversity. The team plans to use the AI recognition system on other species of animals and insects.