Automated Fall Detection From a Camera Using Support Vector Machine

Falls and fall-related fractures of a person is a major health problem and this issue is increasing day-by-day, especially for elderly who live alone. In this paper, we propose a camera based, novel, real-time automated fall detection framework for indoor environments. We process the frames on-the-fly for real time processing. We first apply background subtraction for detecting moving personin the indoor environment. We then extract relevant geometric features to classify fall from other daily activities of a person. Support vector machine (SVM) is applied to distinguish fall and other activities of a person. We have done experiments on publicly available dataset which is UR Fall Dataset. Our experiments demonstrate that our proposed framework has achieved detection rates compared to the state-of-art methods.