Research On Key Issues Of Gesture Recognition For Artificial Intelligence

With the rapid development of computer vision and artificial intelligence,traditional human-computer interaction methods such as keyboard and mouse have become more and more unable to meet human requirements.The gesture recognition that is flexible,intuitive,and can express a variety of information has gradually entered the human vision due to its advantages of easy access and easy operation,and has become a hotspot of new humancomputer interaction methods.At present,there are still some key problems to be solved in gesture recognition based on monocular camera,such as: inaccurate gesture detection,less ambiguous gesture definition,large algorithm calculation and long recognition time,the key and difficult point of solving this kind of problem lies in the effective detection of skin color and the choice of recognition algorithm.Based on the research on gesture recognition of 2D camera,this paper improves the skin color detection method in view of the problems of background interference and inaccuracy of gesture segmentation in gesture detection.At the same time,static gestures and dynamic gestures are innovative in feature extraction,a new feature extraction scheme is proposed,and compared with commonly used machine learning algorithms,the advantages and disadvantages of the support vector machine algorithm are analyzed,on the basis of which improvements are made,and a new gesture recognition method is proposed.Through the experimental results,it is proved that this method can effectively implement gesture recognition,the average recognition rate of more than 98%,can be used as a gesture recognition method,the main work content is as follows.First,in terms of gesture segmentation,this paper compares the commonly used segmentation methods,analyzes the advantages and disadvantages of skin color detection method based on YCb Cr color space,and proposes a de-background gesture detection technology based on YCb Cr color space.By using the maximum connected area,reduce the impact of possible skin-type objects and other skin tone parts in the background when the gesture is divided,extract the gesture to be recognized,and then consider that the face area may be larger than the gesture or form the same connected area as the gesture,to interfere with the results,the use of face position detection,face cover and removal,to realize the gesture detection of removing the background.Second,in the aspect of gesture extraction,this paper mainly studies the extraction of static gesture shapes and dynamic motion trajectories,because they are two separate parts of each other,which will lead to ambiguous gesture feature extraction.This paper proposes a method that can extract both static shapes and dynamic trajectories,and solves the problem of ambiguous gesture feature extraction.Third,in the current study of gesture recognition algorithm,static recognition and dynamic recognition are often identified separately by different algorithms.In this paper,the dynamic trajectory is transformed into a static picture,solves the problem of non-uniform static and dynamic gesture recognition methods,and realizes the common recognition of static and dynamic gestures.Finally,it is proposed to use the principal components analysis(PCA)and the support vector machine(SVM)to realize gesture recognition,so as to avoid the dimensional disaster that may occur when recognizing,and to solve the problem of large calculation,long recognition time and low recognition rate of commonalgorithmic algorithms.Last,in order to verify the feasibility of the algorithm,a gesture control system is designed,combined with the hardware platform,and the control of gesture stoue to hardware is realized through information transmission.After repeated verification,it can achieve the control effect well,and prove the effectiveness of the designed system.