Monday, April 16, 2012

Face Recognition Could Catch Bad Avatars

New Scientist
(04/11/12) Jacob Aron

University of Louisville researchers are developing the field of artificial biometrics, known as artimetrics, to serve as a way to authenticate and identify non-biological agents such as avatars, physical robots, and chatbots. The researchers, led by Roman Yampolskiy, have developed facial recognition techniques specifically designed for avatars. "Not all avatars are human looking, and even with those that are humanoid there is a huge diversity of color," Yampolskiy says. Therefore, the software uses a large variety of colors to improve the recognition of avatars. The researchers also are studying how to match a human face to an avatar generated from that face. Combining the color-based technique with existing facial recognition software led to the best results, suggesting that it could be possible to track users between the physical and virtual worlds. Yampolskiy also intends to create recognition algorithms for robots, forecasting that, in Japan at least, autonomous robots may become sufficiently ubiquitous to require their own identification. Meanwhile, Yampolskiy is working with his Louisville colleagues to develop methods for determining the legitimacy of chatbots. The researchers fed text written by chatbots into software originally designed to identify human authors and found that they were often able to determine the chatbot responsible.

View Full Article