一种改进的DAGSVM 手势识别方法及其应用

在支持向量机多分类方法基础上，提出了一种改进的有向无环图支持向量机( Directed Acyclic Graph Support Vector Machine，DAGSVM) 手势识别方法。首先根据Kinect 采集到 的场景深度信息将前景和背景分开，分割得到手，然后提取其特征向量，利用特征向量训练多 个SVM 两分类器，采用DAG 拓扑结构构成DAGSVM 多分类器，并对其结构排序进行改进。 实验证明，与其他支持向量机多分类方法相比，改进后的DAGSVM 分类器能够达到更高的识 别率，并将这个手势识别方法用于智能轮椅的控制上，收到了良好的效果。

An Improved DAGSVM Hand Gesture Recognition Approach and Its Applications

On the basis of SVM( Support Vector Machine) multiclass classification，an improved DAGSVM ( Directed Acyclic Graph Support Vector Machine) hand gesture recognition approach is put forward． Firstly，depth information of the scene is collected by using Kinect sensor and hand region is obtained． Then feature vectors are extracted，which are used to train multiple binary SVM classifiers． DAGSVM classifier is constructed using DAG topological structure with trained binary SVM classifiers and its structure sequence is improved． Finally，the experimental results proved that the improved DAGSVM could reach higher recognition rate and can be used in the control of intelligent wheelchair successfully．