Monocular perception of biological motion in Johansson displays

Computer perception of biological motion is key to developing convenient an d powerful human-computer interfaces. Algorithms have been developed for tr acking the body; however, initialization is done by hand. We propose a meth od for detecting a moving human body and for labeling its parts automatical ly in scenes that include extraneous motions and occlusion. We assume a Joh ansson display, i.e., that a number of moving features, some representing t he unoccluded body joints and some belonging to the background, are supplie d in the scene. Our method is based on maximizing the joint probability den sity function (PDF) of the position and velocity of the body parts. The PDF is estimated from training data. Dynamic programming is used for calculati ng efficiently the best global labeling on an approximation of the PDF. Det ection is performed by hypothesis testing on the best labeling found, The c omputational cost is on the order of N-4 where N is the number of features detected. We explore the performance of our method with experiments carried on a variety of periodic and nonperiodic body motions viewed monocularly f or a total of approximately 30,000 frames. The algorithm is demonstrated to be accurate and efficient. (C) 2001 Academic Press.