提出一种基于分类特征提取的手部动作识别方法,该方法通过自适应的混合高斯模型构建背景模型,使用背景减除法并充分利用人体手部肤色信息分割出人体手部区域,结合手部关节、骨骼特征及肤色信息估算手部关节点位置,构建三维手部骨架模型,然后提取手部各关节角度、位置信息并利用隐马尔柯夫(HMM)模型对其所表示的动作进行训练识别.%This paper proposes a hand action recognition approach based on classification feature extraction. This method utilises adaptive Gaussian mixture model to construct background model, uses the background subtraction method and makes full use of human hands' skin colour information to segment the hand area. It combines with the hand joints, bone characteristics and skin colour information to estimate the position of hand joints and to construct three-dimensional hands' skeleton model. Then all angles of hand joint and their location information are extracted, and the Hidden Markov Model (HMM) is utilised to train the identification of actions expressed by them.
展开▼