From Blob Metrics to Posture Classification to Activity Profiling

The development of unobtrusive monitoring systems is important to obtain informative cues of human postures and behaviours for the next generation pervasive home care environment. To this end, this paper applies a set of computationally efficient vision techniques to classify human postures, and consequently, to analyze human behaviours such as fall detection. The method starts with the extraction of human silhouettes, then blob metrics using multiple appearance representations, and finally activity profiling based on frame-by-frame posture classification. A large number of experimental results have demonstrated its validity regardless of its simplicity