Research On Dynamic Gesture Recognition Based On Machine Vision

Dynamic hand gesture recognition based on machine vision is a way of human-computer interaction,which is commonly used in virtual reality,smart home and other fields.The improvement and innovation of the dynamic hand gesture recognition algorithm based on machine vision has been the leading scientific problem and key technology in the research of robot intelligence.At present,the dynamic hand gesture recognition based on machine vision has some problems,such as inaccurate gesture detection,poor tracking effect,poor recognition effect and so on.The key and difficult point is the detection and tracking of dynamic hand gestures and the recognition of the trajectory.In this paper,to achieve better dynamic hand gesture recognition,the dynamic hand gesture recognition based on 2D camera is studied,and the algorithm of gesture detection and tracking is improved,and a new trajectory recognition method is presented.The improved algorithm is integrated into the experimental platform of dynamic hand gesture recognition.It has theoretical and practical significance.In the aspect of dynamic hand gesture detection and tracking,the color particle filter tracking algorithm is improved in two aspects.Aiming at the problem of the influence of light changes on hand tracking,the calculation method of color histogram is improved.In this paper,the skin color ellipse model is established in YCrCb color space,and the tracking algorithm is combined with the model.The improved algorithm not only inherits the advantages of the original algorithm in the target tracking,but also increases the success rate of hand tracking.Aiming at the problem of low accuracy of hand gesture tracking algorithm,the results of hand gesture detection is used to regulate the range of particle propagation.The method of particle state transfer is improved,and the accuracy and rapidity are improved effectively.In the aspect of dynamic hand gesture recognition,a new method is proposed to deal with the problems of the current recognition algorithms,such as a large number of training samples,too long training time or the need to select the appropriate features.The new method is based on dynamic hand gesture trajectory matching.The basic geometric and trigonometric functions are used to complete the definition and recognition of hand gestures,but no need to select features.The training samples and training time can be reduced to meet the requirements of real-time identification.Finally,a dynamic hand gesture recognition system is built.The system designs a graphical interface and wireless communication module,which enhance the versatility of the system and make the system easy to expand and develop.Through the test,the system realizes the recognition of the dynamic hand gesture,the accuracy is high,the effect is stable,and the recognition is real time.