基于眼动、位姿及场景视频的人体运动方向预测方法

外骨骼机器人作为新提出的改善、提高人类生活能力的智能设备，同样需要高效智能的人机交互系统，而人机交互的第一步，则是精准预测人的行为意图。从外骨骼机器人的顶层控制角度出发，介绍了人体运动意图识别和外骨骼机器人智能交互能力的研究现状，并对人体运动方向识别进行了研究。提出了一种结合眼动信息、位姿信息及场景视频信息的多信息融合的人体运动意图识别网络架构，并进行了采集设备的穿戴实验。利用实验数据，对提出的网络方法进行了实验验证。结果表明:所提出的人体运动方向的识别系统，可以预测出人体运动过程中的运动方向。 

Human motion direction prediction method based on eye tracking,pose and scene video

Exoskeleton robots, as newly proposed smart devices to improve and enhance human life ability, require efficient and intelligent human-computer interaction systems, and the first step of human-computer interaction is to accurtely predict human behavior intention. From the perspective of top-level control of exoskeleton robots, the current states and progress of human motion intention recognition and the intelligent interaction capabilities of exoskeleton robots are reported. Then, the recognition of human motion direction is studied. A network framework of human motion intention recognition combining eye tracking information, position and posture information, and scene video information is proposed, and wearable experiments of acquisition devices are carried out. The predictive capability of the network has been proved by experiments. The results show that the proposed recognition system can predict the movement direction during human movement.