Abstract

ras) moving? but, rather What is happening? Unfortunately, this new labeling problem is not as welldefined as the previously addressed questions of geometry. Bobick [6] considers the range of motion interpretation problems and proposes a taxonomy of approaches. At the top and intermediate levelsaction and activity, respectively are situations in which knowledge other than the immediate motion is required to generate the appropriate label. The most primitive level, however, is movementa motion whose execution is consistent and easily characterized by a definite space-time trajectory in some feature space. Such consistency of execution implies that for a given viewing condition there is consistency of appearance. Put simply, movements can be described by their appearance. This paper presents a novel, appearance-based approach to the recognition of human movement. Our work stands in contrast to many recent efforts to recover the full threedimensional reconstruction of th

Citations