A Non-Intrusive Deep Learning Based Fall Detection Scheme Using Video Cameras
Mahsa T. Pourazad, Anahita Shojaei-Hashemi, Panos Nasiopoulos, Maryam Azimi, Michelle Mak, Jennifer Grace, Doojin Jung, Taran Bains · 2020
Video-based wellness monitoring is relatively new and attractive field in eHealth, as it enables personalized healthcare with improved accuracy. In this paper we propose a non-intrusive deep learning based real-time scheme to detect falls using video cameras. Our end-to-end model is composed of Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN) modules, which automatically learn discriminative features from training data. We have trained and tested our system on a video dataset we have collected, and have achieved very promising results.