Fall detection using body geometry and human pose estimation in video sequences

According to the World Health Organization, falling is a significant health problem that causes thousands of deaths every year. Fall detection and fall prediction tasks enable accurate medical assistance vulnerable populations whenever required, allowing local authorities predict daily care resources reduce damages accordingly. We present in this paper, approach explores human body geometry available at different frames video sequence. Especially, pose estimation, angle distance between vector formed by head-centroid identified facial image center hip body, aligned with horizontal axis hip, are employed construct new distinctive features. A two-class Support Vector Machine (SVM) classifier Temporal Convolution Network (TCN) trained on newly constructed feature images. At same time, Long-Short-Term Memory (LSTM) network calculated sequences classify non-fall activities. perform experiments Le2i FD dataset UR dataset, where we also propose cross-dataset evaluation. The results demonstrate effectiveness efficiency developed approach.