Recognizing Running Movement Changes with Quaternions on a Sports Watch

Abstract

Sports watches are popular amongst runners but limited in terms of sensor locations (i.e., one location at the wrist). Thus, apps on the watch cannot directly sense movement changes in arbitrary body locations. This, in turn, severely limits training support services, contextual awareness and seamless interaction mechanisms. Our approach addresses this gap and connects the watch with two strategically placed inertial measurement units (IMUs) via bluetooth low energy (BLE). Our prototypical app for Wear OS receives orientation (quaternion) data and matches the sensors to the arm or leg segments using a flexible and simple procedure. We collected data from eight runners and used support vector machines (SVMs) to recognize different movements. Findings from our evaluation with different parameters indicate feasible recognition and low false-positive rates for two to three different movements, for both placements. Our approach can thus help to improve applications that support training and thus contributes to developing motion capture for personal use; it also enables movement-based interaction while running.