Robust Human Pose Quality Assessment Using Optimal Sub-Pattern Assignment

Quantitative assessment of human pose quality is becoming more and more important in various real-world applications. This paper presents a novel optimal sub-pattern assignment based human pose assessment (OSPA-HPA) algorithm for automatically quantifying how well people perform poses. We model the human pose as a collection of finite sets of features. Then, the pose quality is measured using the dissimilarities between the collections representing the reference pose and the actual one. Inheriting the advantages of the OSPA distance, the proposed OSPA-HPA algorithm can naturally deal with the problems of feature loss and feature ambiguity. The effectiveness of the proposed algorithm was validated using three dance training scenarios.