Cognitive Health Assessment of Decentralized Smart home Activities using Federated Learning

The Internet of Things (IoT) and smart homes provide privacy-preserving environments for the healthcare sector to manage the care of individuals with cognitive impairment or disability. These homes, equipped with various sensors, can assist in assessing cognitive health by collecting data on daily activities. As cognitive health deteriorates over time, it can often go unnoticed until it is too late. In the literature, various machine learning and deep learning techniques have been applied to assess daily tasks and differentiate between individuals with competent and impaired cognitive abilities. However, this may compromise the privacy of those living in smart homes. This paper presents a federated learning approach based on deep neural networks to address this concern. The deep neural network model is trained on a cognitive health dataset and implemented on two clients, with a server used to receive updates from both. The results are evaluated in two rounds to reduce overfitting. The experiment demonstrates the effectiveness of the proposed approach, achieving more than 99.2% accuracy while maintaining data privacy.