A comparative study on wearables and single-camera video for upper-limb out-of-thelab activity recognition with different deep learning architectures
This work addresses the need for feasible patient tracking in real-world settings by evaluating simpler acquisition methods, but it is incremental as it compares existing technologies without introducing new paradigms.
The study compared wearables and single-camera video for recognizing upper-limb activities outside the lab using deep learning, finding that both methods can achieve high accuracy, with wearables reaching up to 95% and cameras up to 92% in specific scenarios.
The use of a wide range of computer vision solutions, and more recently high-end Inertial Measurement Units (IMU) have become increasingly popular for assessing human physical activity in clinical and research settings. Nevertheless, to increase the feasibility of patient tracking in out-of-the-lab settings, it is necessary to use a reduced number of devices for movement acquisition. Promising solutions in this context are IMU-based wearables and single camera systems. Additionally, the development of machine learning systems able to recognize and digest clinically relevant data in-the-wild is needed, and therefore determining the ideal input to those is crucial.