Microgesture detection for remote interaction with mobile devices

The rise of smart rings enables for ubiquitous control of computers that are wearable or mobile. We developed a ring interface using a 9 DOF IMU for detecting microgestures that can be executed while performing another task that involve hands, e.g. riding a bicycle. For the gesture classification we implemented 4 classifiers that run on the Android operating system without the need of clutch events. In a user study, we compared the success of 4 classifiers in a cycling scenario. We found that Random Forest (RF) works better for microgesture detection on Android than Dynamic Time Warping (DTW), K-Nearest-Neighbor (KNN), and than a Threshold (TH)-based approach as it has the best detection rate while it runs in real-time on Android. This work shell encourages other researchers to develop further mobile applications for using remote microgesture control in encumbered contexts.