A Deep Learning Based Human Fall Detection Solution
Hamid Reza Tohidypour, Anahita Shojaei-Hashemi, Panos Nasiopoulos, Mahsa T. Pourazad · 2022
Automatic human fall detection is a challenging task in healthcare. Using deep learning models in conjunction with cameras have been proven to be more accurate and efficient in detecting falls in comparison to wearable technology and other more intrusive alternatives. In this paper, we propose a long-term recurrent convolutional network (LRCN) architecture for human fall detection, which consists of a custom convolutional neural network specifically designed to address the challenges of this task and the limited amount of available data, followed by a long short-term memory (LSTM) neural network which decodes sequential information to determine if there has been a fall. Performance evaluations show that our LRCN architecture outperforms the existing state-of-the-art method by 11.20%.