Hand Parsing and Gesture Recognition with a Commodity Depth Camera

摘要: Hand pose tracking and gesture recognition are useful techniques in human–computer interaction (HCI) scenarios, while previous work this field suffers from the lack of discriminative features to differentiate track hand parts. In chapter, we present a robust parsing scheme obtain high-level representation raw depth image. A novel distance-adaptive feature selection method is proposed generate more depth-context for parsing. The random decision forest adopted per-pixel labeling, it combined with temporal prior form an ensemble classifiers enhanced performance. To enforce spatial smoothness remove misclassified isolated regions, further build superpixel Markov field, which capable handle labeling error at variable scales. demonstrate effectiveness our method, have compared benchmark methods. results show produces 17.2 % higher accuracy on synthesized datasets single-frame tests real-world sequences against complex poses. addition, develop algorithm results. experiments achieves good performance state-of-the-art