Empowering Elderly Care: Innovative Fall Detection with OpenPose and YOLO

In the elderly population, the physiological and psychological functions undergo a decline, rendering them susceptible to various risks. Families seek solutions to assist elderly individuals in times of distress. A devised system employs a webcam and utilizes the You Only Look Once (YOLO) algorithm to detect elderly movements and positions within the household. YOLO employs a single neural network, segmenting the image into regions and predicting bounding boxes and probabilities to classify objects. The highest probability bounding box becomes the object separator. The camera is positioned in a room, capturing a 3m x 3m area, and inference is made. The project compares performance between the OpenPose and non-OpenPose systems. The optimal model under OpenPose demonstrates exceptional results, boasting 100% precision, recall, F 1 score, and mAP, with an accuracy of 100%. This success is achieved with a ratio of 70%:30%, a batch size of 64, and a learning rate of 0.01.