IMU-based underwater sensing system for swimming stroke classification and motion analysis

Swimming stroke classification and underwater motion analysis are important in swimming training. In this paper, we propose an IMU-based wearable sensing system for recognizing swimming strokes and motion analysis, focusing on lower-limb movements. The system measures 12 channels of posture signals from the shank, the thigh, and the foot of two legs. Three competitive swimmers were recruited in experiments. With a stroke-dependent quadratic discriminant analysis classifier and selected time-domain features, the proposed system can achieve a satisfactory classification accuracy of 98.63%±1.9%, 99.04%±0.91%, 99.10%±1.43%, 97.24%±1.71% for butterfly stroke, breaststroke, backstroke, front crawl, respectively. Besides, we carry out kinematics analysis of breaststroke. Preliminary results show that the IMU-based sensing system can be used for both swimming stroke classification and motion analysis.