Recognizing Disguised Faces

点击翻译按钮获取中文摘要

Observers are amazingly proficient at identifying human faces under different viewing conditions and across time. However, such variations are not intentional attempts to change the appearance of an individual. With funds from NSF, Dr. Michael J. Tarr and Yi Cheng are investigating how disguised faces are recognized. That is, are we equally good at recognizing someone when they try to hide their identity? And if not, can we train observers to do better? The investigators are addressing these questions by exploring the perceptual mechanisms used in recognizing faces that have been disguised. They use an innovative combination of computer graphics, computational modeling, human psychophysics, and neuroimaging to study the underlying cognitive and neural mechanisms used by human observers to achieve the invariant recognition of individual faces. This project will also identify the conditions under which we are prone to fail in recognizing a familiar face, because of how it has been disguised. Given such information, the goal is to develop training protocols that improve observers' ability to identify those individuals who are attempting to mask their identity. The intellectual merits of this research arise from improved understanding of human face recognition abilities, particularly in a rarely studied domain. Broader impacts of this project include the creation of a standardized "face database" available to the scientific community, consisting of multiple races for use in both behavioral and computational studies. Broader impacts also include increased knowledge of how to train observers (and possibly computational algorithms) to detect disguised individuals. This research will also involve several Brown undergraduates interested in pursuing scientific careers.