中文摘要：

为了实现人与机器人之间最自然、最直观的交互，设计了一种基于Kinect手部姿势与挥动速度识别的方法。该方法利用Kinect for Windows SDK中的API追踪并绘制上肢骨架信息定义识别5种遥控机器人的手势，并通过无线接口遥控机器人运动。Kinect传感器固定和移动两种测试环境验证手势识别方法的准确率。测试结果表明，传感器固定不动时，识别性能几乎不受光照、背景的干扰，接近100％；当Kinect缓慢移动时，识别率略有下降；当Kinect移动速度达到0．4m／s时，识别性能显著下降约25％。这种基于Kinect手部姿势与挥动速度识别的方法为移动机器人提供了一种自然便捷的交互方式。

英文摘要：

In order to implement more natural interaction between human and robot, an approach based on hand gesture and waving recognition with Microsoft Kinect was developed by means of tracking and mapping the upper limb skeleton information with the Windows SDK API for Kinect. Five gestures were defined to direct the movement of mobile robot through a wireless interface. This approach was validated by a mobile robot. The results indicate that this approach performs well with almost 100% accuracy when the Kinect sensor is stationary, and the recogni- tion rate declines slightly when the Kinect moves slowly. The recognition performance is significantly decreased by about 25% when the moving speed reaches at 0.4 m/s. Such method based on hand posture and waving speed rec- ognition provides a natural and convenient way for human-computer interaction.