基于卷积神经网络的教育机器人智能AR教学系统与人机交互  

Convolutional neural network-based intelligent AR teaching system for educational robots and human-computer interaction

作　　者：陈静漪[1] 肖娜 CHEN Jingyi;XIAO Na(Hohai University,Nanjing 211100,China)

机构地区：[1]河海大学,南京211100

出　　处：《自动化与仪器仪表》2024年第11期88-91,95,共5页Automation & Instrumentation

基　　金：江苏省习近平新时代中国特色社会主义思想研究中心一般项目《江苏省义务教育优质均衡发展的路径优化与创新研究》(23ZXZB027);江苏省教育科学“十四五”规划重点资助项目《江苏县域内义务教育优质均衡发展路径与策略研究》(B/2022/01/24)。

摘　　要：对基于卷积神经网络的教育机器人智能AR教学系统与人机交互进行研究,提出一种基于改进HRNet网络的手部姿态估计方法提高机器人智能AR教学系统对用户手势进行识别的速度与准确率。首先,在常用的人机交互流程的基础上,对提高机器人系统指令识别正确率方法的整体框架进行设计,然后根据HRNet网络的缺点进行改进,即采用Ghost模块对HRNet网络结构中的传统卷积进行代替,解决了HRNet网络计算量大、运算复杂的问题;同时在HRNet网络残差结构中融入ECA-Net模块,进一步增强了网络模型对手部姿态特征信息的学习能力。实验结果显示:提出的基于改进HRNet网络的手部姿态估计方法具有可行性、有效性,且能够快速、精准地实现特征信息的提取,完成手部姿态估计任务,为教育机器人智能AR教学系统提供了更为高效的人机交互技术。In this paper,the intelligent AR teaching system of educational robots based on convolutional neural networks and human-computer interaction are studied,and a hand pose estimation method based on improved HRNet network is proposed to improve the speed and accuracy of the intelligent AR teaching system of robots to recognize user gestures.Firstly,on the basis of the commonly used human-computer interaction process,the overall framework of the method for improving the accuracy of instruction recognition of robot system in this paper is designed,and then improved according to the shortcomings of HRNet network,that is,the Ghost module is used to replace the traditional convolution in the structure of HRNet network,which solves the problem of large computational cost of HRNet network.At the same time,the ECA-Net module is integrated into the residual structure of HRNet network,which further enhances the learning ability of the network model to learn the hand pose feature information.The experimental results show that the hand pose estimation method based on the improved HRNet network proposed in this paper is feasible and effective,and can quickly and accurately extract feature information and complete the hand pose estimation task,which provides a more efficient human-computer interaction technology for the intelligent AR teaching system of educational robots.

关 键 词：教育机器人 智能AR教学系统 人机交互 HRNet网络 手部姿态估计 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]