The Research Of Action Recognition Based On Kinect And Application Of Semaphore Traning

Semaphore is a kind of body language which is used in industrail and traffic magagement, the purpose of motion sensing technology is to make computer understands and expalin body language.It also have widespread applications in body language teaching.The new generation of motion sensing technology has overcame the limitations of traditional input device using KINECT, Wii and other equipment (such as data glove).From the view of teaching,the new generation of motion sensing technology makes human-computer interaction better for human,and provides new way for semaphore training.In order to lead the motion sensing techonogy to training,this paper tried to construct a training system to achieve the purpose of semaphore training.Based on the semaphore training,this topic compared several current-popular motion sensing technology.The system contains two parts.First part is training module,second is action recognition module.In the training module,users could chose the semaphore action to train;in the action recognition module users could use the templet to recognize the actions inputed from KINECT.The main part of this article describes how to extra and recognize the feature of body laguage,and tried to improve pretreatment method and action recognition algorithm of skeleton data.In the extra part,this article extracted the three-dimensional coordinates of arthrosis、normallized and weighted the data.In the recognition part,this article applied the Dynamic Time Warping to the three-dimensional coordinate,and modified it to nested Dynamic Time Warping.The result of this experiment shows that,the improved algorithm has significantly improves the accuracy.On the basics of extra part and recognition part,this article added teaching module and voice recognition module.User can use action and voice to paly tutorial video、watch tutorial pictures、interact with computer in real-time through teaching module and voice moduleAll of the above studies have obtained sufficient experimental data to verify its feasibility.This thesis gives a complete scheme of system design and code.Based on the result,this topic leads the motion sensing techonogy to semaphore training,and expands the diversity of it. To the further it could promote the field of talent training, and assist the industry standards for the improvement and progress.