Harnessing Deep Learning for Activity Recognition in Seniors' Daily Routines with Wearable Sensors

Wearable sensors are increasingly popular for monitoring the daily activities of older individuals, sparking significant interest in human activity recognition (HAR). Accurate and timely identification of these activities is crucial for providing personalized care and timely interventions to maintain the well-being of elderly individuals. However, existing HAR methods often need support with complex activity patterns, limited labeled data, and the need for real-time performance. To address these challenges, this study proposes a novel deep-learning approach called ResBiGRU for precisely identifying daily activities in older adults using wearable sensors. The ResBiGRU architecture combines residual connections with BiGRU networks to capture short-term and long-term dependencies in sensor data effectively. We also propose a data augmentation strategy to improve generalization performance. Evaluation of the HAR70+ dataset, consisting of activities performed by 18 elderly individuals, demonstrates superior performance of our ResBiGRU model with an accuracy of 97.39% and an F1-score of 97.74%. Furthermore, our model supports real-time inference, making it suitable for practical applications. Implementing the ResBiGRU approach could significantly enhance monitoring and care for older adults by accurately identifying their daily activities. This research advances HAR methods tailored for elderly individuals, opening new avenues for personalized healthcare and assisted living technologies.