Federated Learning Based Elderly Fall Detection Using Edge Computing

Human falls lead to severe complications in daily life activities. In most cases, it may affect the physical and mental well-being of the person. Fall events become critical when they happen to elderly people. Most of the human falls occurring in the elderly community aggravate their physical and mental conditions. The severity of fall incidents may be mitigated by real-time detection. Many research works were done and many were in progress to detect the fall events so that immediate medical action can be taken to save their lives. Edge computing platforms are gaining popularity in the 5G and beyond 5G communication eras as they provide low latent, high-energy-efficient, and secure computing. Recent research trends in artificial intelligence and machine learning include improvising methods for the classification and detection of events from a set of inputs. Federated learning is one of the most promising machine learning algorithms. In this paper, we propose a federated learning approach for elderly fall detection using edge computing techniques. Two sensory inputs, one from a smart shoe and the other from a smartwatch, are used to collect data from the person, which is then analysed on the edge computing server. If any one of the input sets detects a fall event, it will be immediately communicated as a medical emergency. Also, images of the event can be captured using the camera placed in the room to confirm the event.