Abstract

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