面向投影—摄像交互环境下的多点手势识别系统

Multi-point Gesture Recognition in Projector-camera Interaction System

针对投影—摄像人机交互系统中手势指令描述困难、手势辨识度难以提高、多点手势交互实现困难的问题,将多点手势分解为2个或者2个以上的元手势的组合,它们之间满足一定的空间位置关系和时序逻辑;将经典的曲线形状描述方法Freeman链码应用于手势描述系统,采用该方法可以使手势描述满足形状不变性、唯一性和鲁棒性的特点;并结合改进的Hausdorff距离建立手势识别系统,在距离变换空间内,对交互人手势与事先建立的模板库进行匹配识别。实验结果表明,采用以上方法搭建的交互系统对手势具有较高的识别率,有很强的实用价值。

According to described gesture commands, improving gesture recognition, achieving multi-point gesture interaction are difficult in Projec tor-Camera human-computer interaction systems. The multi-point gesture combination is divided into two or more meta gesture, satisfies certain spa- tim relationships and timing logic between them. Gestures described can satisfy shape invariance, uniqueness and robustness characteristics with the Freeman Chain Code. Gesture recognition system is established combining with the Modified Hausdorff Distance, and it can match the gestures of human with pre-buih gesture template library after transformed distance space. The experimental result shows that using the gesture interaction system based on the above method can have a high recognition rate, and grate practical value.