Falls are serious and costly for elderly people. The Centers for DiseaseControl and Prevention of the US reports that millions of older people, 65 andolder, fall each year at least once. Serious injuries such as; hip fractures,broken bones or head injury, are caused by 20% of the falls. The time it takesto respond and treat a fallen person is crucial. With this paper we present anew , non-invasive system for fallen people detection. Our approach uses onlystereo camera data for passively sensing the environment. The key novelty is ahuman fall detector which uses a CNN based human pose estimator in combinationwith stereo data to reconstruct the human pose in 3D and estimate the groundplane in 3D. Furthermore, our system consists of a reasoning module whichformulates a number of measures to reason whether a person is fallen. We havetested our approach in different scenarios covering most activities elderlypeople might encounter living at home. Based on our extensive evaluations, oursystems shows high accuracy and almost no miss-classification. To reproduce ourresults, the implementation is publicly available to the scientific community.
展开▼
