Unobtrusive estimation of in-stroke boat rotation in rowing using wearable sensors

Abstract

The rotational motion of a rowing boat during single strokes has significant impact on the boat velocity and overall rowing performance. However, a method for automatic in-stroke field quantification remains challenging. In this work, we propose a robust stroke segmentation algorithm in combination with a 3D-rotation estimation during segmented strokes. Our method is designed to process unobtrusively obtained inertial sensor data of one sensor device attached to rowing boats. A template-based matching algorithm is implemented to detect all strokes in the collected sensor data. The segmented strokes are then analyzed for the corresponding in-stroke rotation. The evaluation of the stroke segmentation was performed with professional race and amateur training data. The resulting precision was 99.8 % for professional and 97.2 % for amateur data. The in-stroke rotation angle calculation was validated with amateur training data of four boat classes. The results were compared to corresponding measurements from the literature.