Markerless human body pose estimation from consumer depth cameras for simulator

In recent years, many studies have shown that horse riding exercises have positive effects on promoting both physical and psychological health. To maximize the effects, the correct posture is essential when riding a horse. Therefore, the purpose of this study is to present an algorithm for estimating a human pose from depth data while riding a horse simulator. This estimated information can be used for analyzing the riders posture. The proposed rider pose estimation algorithm is divided into four steps: (1) head detection, (2) body part segmentation, (3) joint position prediction, and (4) updating the joint positions. Each step is dependent on the previous step being completed successfully. We compared the experiment results between our joint prediction algorithm and ground truth data to show the performance of the proposed methodology.