Research on Motion Capture and Phase Segmentation Based on Wireless Body Sensor Networks in Competitive Equestrian

This paper presents an equestrian motion phase segmentation framework with wearable body sensor network (BSN). In our method, the wearable inertial sensor nodes were arranged on the riders' body segments and horseback. An extend Kalman filter is proposed for multi-sensor data fusion whose accuracy is validated by optical motion capture system, and the 3-D posture of rider in equestrian is reconstructed by constructing a three-dimensional human kinematics model. In addition, a deep network TCN-Bi-LSTM is proposed for equestrian phase segmentation under different equestrian gaits, and five baseline classifiers are chosen for comparison to validate the robustness of TCN-Bi-LSTM. Furthermore, the impact of different sensor numbers on the accuracy of phase segmentation is also studied to determine the optimal sensor numbers. Detection performance is evaluated using metrics of accuracy, specificity, recall, and F1 score, and the performance indicators after final convergence remain stable at 98.7%, 98.8%, 98.7% and 98.7% for walking, 97.8%, 97.8%, 97.9% and 97.8% for rising trot, 97.6%, 97.4%, 97.5%, 97.4% for sitting trot, 97.9%, 97.9%, 97.9% and 97.9% for canter, respectively. We believe that by exploring the application of BSN in this curious field of equestrian competition, it is very meaningful to expand the application field of BSN.