Motion Capture Sensor-Based Pose Estimation for Dance Canonical Movements

Abstract

Abstract Motion capture technology, a developing technique for the quantification of human movement, is progressively revealing its unique significance within the realm of dance. This study introduces a motion capture approach using Kinect sensors to estimate normative dance postures. The sensors capture depth data, allowing for real-time, precise recording of a dancer’s posture. We propose a new method of similarity matching between feature planes to enhance the analysis of human movement postures. Evaluation trials have shown that this method, which focuses on feature plane similarity matching, yields more accurate assessments of movement complexity than traditional 3D model matching techniques. The feature plane matching-based similarity technique achieved the highest pairwise ranking precision, at 80.25%, using the Urban Dance Motion Quality Evaluation dataset. Additionally, it recorded the highest average bilateral ranking accuracy of 78.68% on the BEST dataset. This method has been proven to enhance the stability and efficiency of human posture analysis through its application of feature plane matching.