Application Of Gesture Recognition Based On Deep Learning In Finger Rehabilitation Training

Gestures have the advantages of being natural and universal in human communication.Limited by a variety of factors,the number of patients with hand dysfunction in today’s society is increasing.In this context,the use of artificial intelligence technology for hand function rehabilitation has become the next hot topic in current research.This article reviews the current research status of rehabilitation treatment for patients with hand dysfunction at home and abroad,and analyzes its development direction.From the perspective of rehabilitation medicine,this paper expounds the physiological structure characteristics of the hand,the characteristics of the rehabilitation training of the hand,and the requirements of the rehabilitation training of the hand.A gesture recognition algorithm based on the fusion of YCb Cr color space and convolutional neural network is proposed.At the same time,in order to realize the application of gesture recognition in finger rehabilitation training,a set of vision-based gesture rehabilitation training system is designed.The specific research content is as follows:(1)the basic concepts of deep learning are introduced,including some typical neural networks and their models,as well as the structure and principle process of each neural network model.The evaluation methods and specific scoring indicators of deep learning models are given.The current mainstream deep learning object detection algorithms are compared and analyzed to provide theoretical support for the selection of basic deep detection models.(2)The advantages and disadvantages of commonly used gesture datasets are analyzed,and it is found that different datasets and the completeness of the datasets have a great influence on the training results of the detection model.At the same time,to avoid data problems that lead to inefficient model detection,we use unsupervised and supervised data augmentation methods to enhance the original dataset.This data augmentation method not only increases the number of datasets,but also increases the diversity of target data.(3)Aiming at the problem of low recognition rate of existing gesture recognition methods in complex environments,this paper proposes a gesture recognition algorithm based on the combination of YCBCR color space and convolutional neural network.Experimental results show that compared with other algorithms listed in this paper,the algorithm has the best convergence effect and the best comprehensive performance of the model,and the algorithm can effectively remove the influence of complex background,greatly improve the accuracy,and can accurately identify static gestures in complex environments.(4)in order to realize the application of gesture recognition in rehabilitation system,a computer vision gesture recognition system based on deep learning was proposed.The data is collected by the camera,the prediction is made by loading the deep learning model,and the rehabilitation system is constructed by the display interface technology.The rehabilitation training program is interesting,efficient and convenient,and the treatment method makes up for the defects of the traditional rehabilitation treatment.