Action recognition of skiers based on inertial sensors

In this study, we present a method for identifying ski jumpers' actions based on a variety of inertial data. Multiple convolutional neural networks are used to recognize each action phase and find the nodes with high recognition accuracy. The data acquired during ski jumping is separated into action phases and then processed to give stacked inertial signal images. According to the results, it is possible to distinguish the wearing part of a speed-type wearable inertial sensor and recognize the action phase during the ski jumping competitions more correctly and effectively.