一种面向舱外活动的视觉感知启发的手势识别方法  

A Hand Gesture Recognition Method Inspired by Visual Perception for Extravehicular Activity

作　　者：王莉[1] 马钟 唐雪寒[1] 王竹平[1] 周革强[2] 何双亮[2] WANG Li;MA Zhong;TANG Xuehan;WANG Zhuping;ZHOU Geqiang;HE Shuangliang(Xi'an Micro-Electronics Technique Institute, Xi'an 710065 , China;China Astronaut Research and Training Center, Beijing 100194, China)

机构地区：[1]西安微电子技术研究所,西安710065 [2]中国航天员科研训练中心,北京100194

出　　处：《载人航天》2017年第6期805-810,共6页Manned Spaceflight

基　　金：载人航天预先研究项目(060601)

摘　　要：针对航天员出舱作业(EVA)时的人机交互需求,结合着舱外服手套后手部动作的工效学分析,提出了一种结合眼动仪和手势识别技术的航天员目标选取方案,使航天员真正达到"解放双手"的目的。该方案采用眼动仪获取航天员"感兴趣目标区域"的位置信息,利用摄像头捕获航天员手势动作,并基于稀疏滤波的特征提取方法和神经网络分类器对手势进行识别,根据预定义的目标选取指令实现交互功能。试验结果表明,手势识别算法的识别率可达到95%以上,为后期将该交互系统集成到航天服中提供了原型依据。To satisfy the needs of the human-machine interaction during the astronaut extravehicularactivity(EVA),a target selection scheme for astronauts was proposed based on the hand motion ergonomicanalysis of the EVA spacesuit gloves.The scheme combined with the eye tracker and thehand gesture recognition technology could realize the“hands free”interaction during EVA tasks.A neye tracker was used to determine the target of interest and a camera was used to capture the astronaut’s hand gesture.Then,a gesture recognition algorithm was developed based on the sparse filteringfeature extraction and the neural network classifier to realize the interaction function according tothe predefined target selection command.The experimental results showed that recognition rate wasmore than95%,which can provide a prototype basis for the future integration into the spacesuit.

关 键 词：舱外活动 手势识别 神经网络 稀疏滤波 

分 类 号：R857.3[医药卫生—航空、航天与航海医学]