摘要
设计并开发了基于视觉手势识别的人机交互实验平台。该平台采用Leap Motion传感器设备跟踪并采集用户手部运动图像数据信息;接着提取用户的手部特征,包括手指关节坐标、手指长度和关节角度,然后利用长短期记忆网络根据特征数据对手势动作进行提前预测识别。在4 500个不同样本数据集上进行模型训练以及测试,结果表明手势预测实时性高、正确率可达99.31%,同时在AR-10仿人多指机械手和NAO双足人形机器人上完成了猜拳游戏人机交互实验,验证了平台的有效性。该平台通用性和综合实践性较强,可开发多种应用案例,为机器人相关实践教学提供创新实践平台,促进高素质人才培养。
A human-computer interaction experimental platform is designed and developed,which uses Leap Motion controller to collect the data of the hand and extracts the finger joint coordinates,length and angle features.Then the gesture action is predicted and recognized in advance by long-short term memory network(LSTM).Model training and test results carried out on 4 500 different sample data sets show that the gesture prediction has high real-time performance,and the accuracy can reach 99.31%.At the same time,the application experiment of the finger-guessing game is performed on the AR-10 robotic hand and NAO robot to verify the effectiveness.The platform is highly comprehensive and universal,which can develop a variety of application research and provide an innovative experimental platform for practical teaching related to robotics to promote the cultivation of high-quality talents.
作者
占宏
吴碧霄
杨辰光
ZHAN Hong;WU Bixiao;YANG Chenguang(School of Automation Science and Engineering,South China University of Technology,Guangzhou 510641,China)
出处
《实验室研究与探索》
CAS
北大核心
2022年第7期23-25,29,共4页
Research and Exploration In Laboratory
基金
国家自然科学基金项目(U20A20200)
华南理工大学2021年校级教研教改项目(C9213106)
华南理工大学第七批探索性实验项目(C9202240)。
关键词
机器人
视觉技术
人机交互
手势识别
Robots
Vision technology
Human-computer interaction
Gesture recognition
作者简介
占宏(1987-),女,湖北黄冈人,硕士,实验师,现主要从事实验教学以及机器人智能控制研究。Tel.:15902037681,E-mail:auhzhan@scut.edu.cn。
