A Smartphone-Based Adaptive Recognition and Real-Time Monitoring System for Human Activities

• Published in: IEEE transactions on human-machine systems Vol. 50; no. 5; pp. 414 - 423
• Main Authors: Qi, Wen, Su, Hang, Aliverti, Andrea
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.10.2020; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accelerometers; Adaptive systems; Algorithms; Classification algorithms; Classifiers; Cluster analysis; Clustering; Data compression; Decision trees; deep learning (DL); Distance learning; Feature extraction; Gyroscopes; hierarchical classification (HC); Human activity recognition; human activity recognition (HAR); Human performance; Machine learning; Monitoring; Monitoring systems; Moving object recognition; Real time; Real-time systems; Sensors; Smartphones; Statistical models; Telemedicine
• ISSN: 2168-2291; 2168-2305
• DOI: 10.1109/THMS.2020.2984181

Human activity recognition (HAR) using smartphones provides significant healthcare guidance for telemedicine and long-term treatment. Machine learning and deep learning (DL) techniques are widely utilized for the scientific study of the statistical models of human behaviors. However, the performance of existing HAR platforms is limited by complex physical activity. In this article, we proposed an adaptive recognition and real-time monitoring system for human activities (Ada-HAR), which is expected to identify more human motions in dynamic situations. The Ada-HAR framework introduces an unsupervised online learning algorithm that is independent of the number of class constraints. Furthermore, the adopted hierarchical clustering and classification algorithms label and classify 12 activities (five dynamics, six statics, and a series of transitions) autonomously. Finally, practical experiments have been performed to validate the effectiveness and robustness of the proposed algorithms. Compared with the methods mentioned in the literature, the results show that the DL-based classifier obtains a higher recognition rate (<inline-formula><tex-math notation="LaTeX">\text{95.15}\%</tex-math></inline-formula>, waist, and <inline-formula><tex-math notation="LaTeX">\text{92.20}\%</tex-math></inline-formula>, pocket). The decision-tree-based classifier is the fastest method for modal evolution. Finally, the Ada-HAR system can monitor human activity in real time, regardless of the direction of the smartphone.