基于YOLO算法的手势识别  

Gesture Recognition Based on YOLO Algorithm

作　　者：王粉花[1,2,3] 黄超 赵波 张强 WANG Fen-hua;HUANG Chao;ZHAO Bo;ZHANG Qiang(School of Automation and Electrical Engineering, University of Science and Technology Beijing, Beijing 100083, China;The Institute of Artificial Intelligence, University of Science and Technology Beijing, Beijing 100083, China;Beijing Engineering Research Center of Industrial Spectrum Imaging, Beijing 100083, China)

机构地区：[1]北京科技大学自动化学院,北京100083 [2]北京科技大学人工智能研究院,北京100083 [3]北京市工业波谱成像工程中心,北京100083

出　　处：《北京理工大学学报》2020年第8期873-879,共7页Transactions of Beijing Institute of Technology

基　　金：国家重点研发计划重点专项资助项目(2017YFB140010101);北京科技大学中央高校基本科研业务费专项资金资助(FRFBD19002A)。

摘　　要：研究YOLO算法在手势识别中的应用,提升在近肤色和光线明暗不一的背景下检测的速度和精度.YOLO算法是端到端的检测方法,通过卷积神经网络自动提取目标的特征,可以大幅度提高运算速度.鉴于YOLO算法在目标检测任务中的优良表现,将YOLO算法应用到手势识别问题中.通过对YOLO系列算法的研究对比表明,YOLO算法在手势识别中具有良好表现.同时,在YOLOv3算法的快速版本YOLOv3-tiny的基础上提出了YOLOv3-tiny-T算法.YOLOv3-tiny-T在包含5种手势的UST数据集上,平均精度均值为92.24%,较YOLOv3-tiny获得了5%左右的提升.The application of YOLO(you only look once)algorithm in gesture recognition was studied to improve the speed and accuracy of detection under the background near the skin color,light and shade.Based on the end-to-end detection function,the YOLO algorithm could be arranged to improve operation speed greatly by automatically extracting target feature from convolution neural networks.Considering the excellent performance in target detection process,YOLO algorithm was applied to gesture recognition.Comparing with other application results with YOLO series algorithm,this application result of YOLO algorithm shows better performance in gesture recognition.At the same time,based on a YOLOv3-tiny algorithm,the fast version of YOLOv3 algorithm,a YOLOv3-tiny-T algorithm was proposed.The YOLOv3-tiny-T algorithm can achieve a mean average precision of 92.24%on the UST dataset with five gestures,increasing about 5%combined with YOLOv3-tiny.

关 键 词：手势识别 YOLO算法 YOLOv3-tiny-T算法 平均精度均值 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]