摘要
本文提出了一种基于深度学习的手势识别系统,该系统利用YOLO捕获手部信息,利用ReXNet捕捉手部21个关键点,利用ResNet-50实现对手势进行分类。试验结果表明,YOLOv5可以准确地将手部信息提取出来,准确度达98.5%;ReXNet捕捉手部21个关键点的准确度达97.2%;ResNet-50手势识别的准确度达98.2%。表明该方法能够检测图像中手势相关信息,对于手势控制有一定的借鉴意义。
This article proposes a gesture recognition system based on deep learning, which uses YOLO to capture hand information, ReXNet to capture 21 key points of the hand, and ResNet-50 to classify gestures. The experimental results show that YOLOv5 can accurately extract hand information with an accuracy of 98.5%;the accuracy of ReXNet in capturing 21 key points of the hand reached 97.2%;the accuracy of ResNet-50 gesture recognition reached 98.2%. This method can detect gesture-related information in images and has certain reference significance for gesture control.
出处
《传感器技术与应用》
2024年第4期570-578,共9页
Journal of Sensor Technology and Application
