Application Research Of Network Gesture Recognition Technology Based On Object Detection

With the rapid development of computer technology and the application of artificial intelligence,people’s demand for technology to change their lives is increasing day by day,especially in the field of intelligent furniture.Human-computer interaction has become an indispensable part of people’s life.With characteristics of simple gesture recognition among them,the implementation is convenient,strong combination become the focus of the research field of human-computer interaction,the traditional gesture recognition method has low accuracy and poor robustness,real-time character,with the development of deep learning in recent years at home and abroad,for the research and application of gesture recognition algorithm has brought the new direction.This study based on the deep study on the gesture recognition performance and application of in-depth research,fully based on in-depth study of gesture recognition research,can greatly improve the accuracy of the gesture recognition,the application of development of gesture recognition under different scenarios,improving the use value,depth study of gesture recognition in the use of smart home and development has an important significance.The main research contents of this thesis are as follows:(1)Research on gesture recognition using deep learning target detection algorithm.The target detection network is studied,and a total of 2250 gesture images in gesture data set under multiple environments are established.After data expansion,the performance of gesture recognition model in self-built gesture data set of target detection networks Faster R-CNN,SSD and YOLOv5 is compared.By analyzing and comparing the results,It is found that YOLOv5 can meet both high accuracy and high speed,but it still has low recognition accuracy for distant gestures and dense gestures.There is room for improvement in model performance,so the YOLOv5 model algorithm needs to be improved.(2)Improve the YOLOv5 algorithm according to the points to be improved.For the existing YOLOv5 algorithm,in order to improve its performance,the attention mechanism SElayer is introduced to improve the model performance.In order to improve the detection performance of small targets,Kmeans clustering is used to find appropriate prior boxes,and another detection head is added to improve the detection performance of small targets.In order to improve the detection performance of dense targets,Introduce VFloss in VFnet to replace the loss function in YOLO.YOLOv5 is improved through experimental analysis,and its performance is improved,and the accuracy of gesture recognition for small targets and dense targets is improved,and the real-time gesture recognition can be well met.(3)Deploy the improved YOLOv5 algorithm on mobile terminal and realize the application of gesture recognition.The practical application of the improved YOLOv5 algorithm is carried out,and the improved YOLOv5 algorithm is deployed to the mobile terminal Jeston Nano B01.The switch command is programmed for Arduino,and the Arduino controlled light switch is connected through serial port communication,so as to realize the control of switch through gesture recognition.The experimental results show that the improved YOLOv5 can recognize gestures well on the mobile terminal,meet the real-time requirements of the mobile terminal,and realize the control of gestures on the light switch.