基于手部关键点检测的手势识别研究  

Research on Gesture Recognition Based on Hand Key Point Detection

作　　者：王森宝 杨晋骁 王子昂 李世尧 秦娟[1] 石艳梅[1] WANG Sen-bao;YANG Jin-xiao;WANG Zi-ang;LI Shi-yao;QIN Juan;SHI Yan-mei(Tianjin University of Technology,Tianjin 300384)

机构地区：[1]天津理工大学集成电路科学与工程学院,天津300384

出　　处：《电脑与电信》2022年第5期29-33,47,共6页Computer & Telecommunication

基　　金：天津市天津理工大学大学生创新项目,项目编号:202110060156。

摘　　要：针对当今社会手势识别应用度逐渐提高的热点现象,提出一种基于手部及背景环境等21个特征点检测的手势识别算法。首先采用数据流处理机器学习应用开发框架,对图像进行变换与渲染,定义5个手指二维向量的相关角度;然后根据经验确定手指弯直状态改变的阈值角度,通过5个手指不同弯直状态的组合表征不同的含义;同时进行鲁棒预测控制,降低了相关的外部干扰与建模误差,提高了图像识别的精准度。采用公开数据集对提出的方法进行验证,平均识别率达到95%,提高了识别的精准度,为人机交互手势识别的发展提供了新思路。In view of the hot phenomenon of increasing application of gesture recognition in today’s society, a gesture recognition algorithm based on the detection of 21 feature points such as hand and background environment is proposed. Firstly, the data flow processing machine learning application development framework is applied to transform and render the image, and the correlation angle of the five-finger two-dimensional vector is defined. Then the threshold angle of finger bending state change is determined according to experience, and the different meanings are represented by the combination of five fingers with different bending state. At the same time, robust predictive control can reduce the external interference and modeling error, and improve the accuracy of image recognition. The proposed method is verified by using public data sets, and the average recognition rate reaches 95%, which improves the accuracy of recognition and provides a new idea for the development of human-computer interaction gesture recognition.

关 键 词：手势识别 特征点检测 鲁棒预测控制 手势追踪 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术]