基于Kinect手势识别的研究与应用

摘要

For the problem of dynamic gestures recognition,first acquiring the depth images of gestures by Kinect sensors,and combining the color image with the depth image to segmenting the gestures.Then extracting the gesture feature as input of hidden markov model,tuning the parameters on the hidden markov model in the experimental process to improve the recognition rate of dynamic gesture.Finally using the gesture recognition method in the 3D earth simulation system,you can collect informations of human bones by the sensor and control the platform real-time and accurately by gestures.And it has high accuracy and the recognition rate,and improving the people's interaction experience greatly.%针对动态手势识别的问题,首先利用Kinect传感器获取手势的深度图像,结合彩色图像和深度图像对手势进行分割.然后将提取的手势特征作为隐马尔科夫模型的输入,在实验过程中提出对隐马尔科夫模型参数进行调优,进而提高动态手势识别率.最后将优化后的手势识别的方法运用于三维地球仿真系统中,利用传感器获取的人体骨骼信息,实现了手势对平台进行实时控制,并具有较高的准确率和识别率,大大提高了人们的交互体验.