WellBe: An Intelligent Elderly Care and Well-Being Monitoring System Using Deep Learning

Deep Learning refers to a subset of Machine Learning that utilizes deep neural networks to simulate the complex decision-making process of the human brain. In recent years, DL has made remarkable progress in addressing multiple unique challenges faced by most elderly individuals. While there have been significant contributions to independent systems catered for the elderly, there is the existence of an overall integrated AI platform that unifies the use of individual modules for an optimized solution. This paper introduces “WellBe,” an innovative AI system that encompasses an alarm-based fall detection component, a facial emotion recognition (FER) system and a built-in personalized chatbot all under a user-friendly website. The novelty of the work extends beyond the aforementioned components, owing to the implementation of an optimized FER system able to handle wrinkles and changes in skin texture, thereby tailored exclusively for the elderly demographic. The system leverages a custom feature extraction module, fusing facial and wrinkle features, in addition to extending a convolutional neural network (CNN) architecture for accurate emotion classification, achieving promising performance with precision, recall, and F1-scores of 0.87, 0.91, and 0.89 for the “Happy” (one of the predicted emotions) class, respectively. The fall detection component, on the other hand, utilizes a pre-trained YOLOv7 Pose model for the real-time identification of falls, sitting at a performance accuracy of 95.3%. Finally, the personalized chatbot is implemented using a rule-based approach enhanced with the Jaccard similarity algorithm, to facilitate prompt assistance and a sense of companionship to the end user. Overall, the integration of these components into a cohesive website offers a comprehensive solution for enhancing the quality of life for elderly individuals living alone, addressing loneliness, safety, and overall well-being.