提出一种将两个人的互动分解为主动行为和被动行为的新方法,以实现更有效的行为识别,并创建了一个包含6种复杂人体行为的新数据集 K3HI,其中包括踢、指、拳击、推、交换物品和握手。对每一种主动行为提取出其关节、面和速度3个特征,并对这3个特征序列进行分析从而判断其行为。通过采用自己的数据集,利用连续的隐马尔可夫模型(HMMs)对主动行为识别方法和传统两人互动行为识别方法进行对比分析。结果表明,本文识别方法相比于传统方法更准确,并且可缩短样本训练时间,体现了其综合优势。%A novel approach is proposed to decompose two-person interaction into a positive action and a negative Action to im-prove the efficiency of behavior recognition.A new dataset with six types of complex human interactions (i.e.,named K3HI)is created,including kicking,pointing,punching,pushing,exchanging an object,and shaking hands.Three types of features are extracted and analyzed for each positive action:joint,plane,and velocity features to recognize the action.We used continuous Hidden Markov Models (HMMs)to evaluate the positive action-based interaction recognition method and the traditional two-per-son interaction recognition approach.Experimental results show that the proposed recognition technique is more accurate than the traditional method,shortens the sample training time,and therefore achieves comprehensive superiority.
展开▼