Hand Gestrue Recognition Algorithm Based On CNN

In recent years,gesture has become a very active subject of computer vision research with its natural and fast features.It can be applied in many fields,such as allowing aphasic and deaf patients to communicate with others,somatosensory games,remote control and so on.Therefore,gesture recognition has attracted a lot of attention and research.The existing gesture recognition algorithms have many problems,such as large amount of computation,long time consumption and low recognition rate under background interference and illumination changes.Deep Learning is good in many aspects,such as visual recognition,speech recognition,and natural language processing.In different types of deep learning architectures,Convolutional Neural Network(CNN)is the most suitable for processing image data.This paper focuses on study CNN-based gesture recognition algorithms.The main work and achievements of this paper are as follows:1.This paper briefly introduces the traditional hand gesture recognition methods,template matching method and hidden Markov model method.It introduces the method of hand gesture recognition based on CNN,and outlines related theories involved in the process of hand gesture recognition.2.In the study of the hand gesture recognition algorithm without detection stage,the Hand-CNN gesture recognition algorithm is proposed in this paper,which is mainly used for the gesture in the image with a large proportion of the target and background.The Hand-CNN algorithm mainly adjusts the AlexNet model network's structure and uses feature fusion method to improve the accuracy of gesture recognition.Experimental results show that the accuracy of the Hand-CNN algorithm on the American Sign Language(ASL)data set is 98.2%.Compared with the original AlexNet model,the recognition rate is increased by 3.2%.3.When studying the gesture recognition algorithm whose recognition is after detection,a gesture recognition algorithm based on Faster RCNN is proposed in this paper.The algorithm is mainly used for the gestures in images with smaller target and background ratio.The key parameters in the Faster RCNN are modified to detect and identify the gestures simultaneously.Then DisturbIoU algorithm is proposed to avoid the over-fitting problem of the training modeland further improve the recognition accuracy.The results of gesture recognition experiments on the NTU and VIVA data sets show that the algorithm can effectively avoid the over-fitting problem of the training model,and obtain a better performance in both accuracy and robustness compared with other existing algorithms.4.The contrastive experiments show that the gesture recognition method used in this paper is much better and more advanced than the traditional methods.