Fusing IoT Wearable and Object Motion Sensors for Enhanced Activity Recognition in Smart Homes.

With a growing older population and a shortage of aged care staff, there is a pressing need for technology solutions to support the independent living of older adults in their own homes, facilitating aging in place. Internet of Things (IoT)-based technologies present a promising solution through the use of distributed sensors to facilitate daily activity monitoring and deliver timely alerts to caregivers and clinicians. This study investigates the integration of wearable and object-based IoT motion sensors to accurately recognize human activities that are essential for assessing the independent living ability of older adults. Real-world data was collected from 17 key daily activities in a home setting, using 30 wearable and object motion sensors. We then implemented several machine learning algorithms to classify the activities based on the sensor data, focusing on enhanced accuracy and system scalability.Our findings demonstrate that the Random Forest (RF) classification algorithm can achieve an accuracy of 90.63% in recognizing the activities using 11 wearable sensors. When combined with 19 object sensors, the accuracy improves to 97.88%. To enhance scalability and cost-effectiveness, we optimized the sensor configuration, reducing the number of required sensors by 57% (from 30 to 13) without compromising activity recognition performance. The proposed approach offers a robust mechanism to monitor daily activities that can be used to provide real-time alerts to caregivers and clinicians. This capability supports timely assistance and interventions, ultimately promoting the well-being and independence of older adults in their own homes.