| Stroke is one of the three major killers that seriously endanger human health,and the proportion of strokes among Chinese has reached as high as 39.3%.The postural hemiplegia caused by stroke will greatly reduce the quality of life of patients,especially the symptoms of finger stiffness directly weaken the functions of the fingers such as grasping and stretching,so it is necessary to correct the finger shape of stroke patients in time.On its basis,the intelligent recognition of finger gestures can realize barrier-free communication.For this reason,the research goal of this article is to study the finger shape correction of stroke patients based on the principles of ergonomics.Based on the recognition of gestures,the intelligent recognition of postures will pave the way for the development of intelligent rehabilitation equipment.The thesis studied the movement of flexible correction of the finger shape of stroke patients,and found that there are two problems that affect the rehabilitation effect:1)the types of rehabilitation gestures;2)the precise reset of the rehabilitation fingers.Aiming at these two problems,the thesis designs a flexible rehabilitation glove based on a three-capsule segmented driver,which can realize 3 kinds of rehabilitation gestures,and at the same time,realize the accurate reset of the rehabilitation finger through the torsion spring reset device.Through multiple sets of performance experiments:1)When the bending angles are 30°,50°,70° and 90°,the driving air pressure required by the three-capsule segmented bending actuator is smaller than the bending angle of the composite soft glove:20%,44%,57%and 64%;2)The end of the driver can produce a maximum force of 3.2N,which is sufficient to meet the requirements of daily rehabilitation;3)The model is verified by experimental research on the bending performance and reset performance of the glove The effectiveness of the structure;4)Tested the auxiliary function of rehabilitation gloves used for hand rehabilitation exercises of stroke patients.The core problem of gesture recognition is how to improve the accuracy of gesture recognition.Aiming at this problem,a gesture recognition visual detection and recognition algorithm based on Double-Channel DC Net is proposed to realize simple communication.Firstly,the original gesture image is preprocessed,denoised and edge detected to get the hand edge image.Secondly,the gesture image and the hand edge image are respectively selected as the two input channels of CNN.Each channel contains the same number of convolutional layers and the same parameters,but each channel has a separate weight.Finally,perform feature fusion in the fully connected layer,and use the softmax classifier to classify the output results.Experiments have shown that the dual-channel DC-Net algorithm can effectively improve the recognition rate of rehabilitation gestures,the training accuracy rate reaches 99.6%,and the loss value is reduced to 0.06,which enhances the generalization ability of CNN.On the basis of the research on the intelligent recognition of gestures,the application research of gesture recognition visual detection is launched to form a set of gesture recognition visual detection application system,which can realize the collection of gesture data,the training of deep learning network and the testing function of deep learning network.The data collection module is used for training and test data collection;the deep learning training module is used for modification of training parameters and the visualization of training results;the deep learning test module is used for testing the performance of the network.In the future,on the basis of this research,combined with the intelligent recognition of posture,we will develop in-depth intelligent interaction methods,expand more rehabilitation training of limbs,and form a complete set of intelligent rehabilitation equipment for stroke patients... |
PDF Full Text Request