Telemonitoring of Daily Activities Based on Multi-Sensors Data Fusion
Mouazma Batool, Ahmad Jalal
Abstract
Multi-sensor data fusion has played a crucial role in a wide range of daily life activities including smart homes, sports, healthcare, and so on. This work has presented a telemonitoring approach for the recognition of daily life activities by analyzing multi-sensor data including depth, inertial, and RGB sensors. Combining multiple sensors enhances the accuracy of daily life activities, as each sensor captures the unique detail of human activities. The RGB sensors provide information about object appearance and visual cues. The depth sensors disseminate information on 3D spatial and geometric. The inertial sensors impart knowledge about the angle and orientation details of each motion. The prescribed methodology is based on multi-sensor data preprocessing, feature extraction, data fusion, data optimization, and classification. The prescribed method has achieved an optimal accuracy of 98.32% and 98.03% on Berkeley MHAD and UTD-MHAD standard datasets. The proposed model has outperformed the classification results over cutting-edge models. The prescribed methodology could be an opportunity for an IoT-based solution, to optimize the activities of daily living through seamless integration. The prescribed methodology can be further used in developing intelligent systems to detect human behavior and improve interactive responsiveness between humans and technology.