吴天强  袁学俊  胡文庆 

（广东人工智能与先进计算研究院，广东 广州 510535）

摘要：针对目前人体动作捕捉系统采集节点数量少、动作采样精细度不足、成本高等问题，设计一套基于MEMS惯性传感器的47节点人体动作捕捉系统。该系统通过在人体关节部位佩戴传感器节点，实时采集关节部位的精细运动数据；运动数据汇聚后通过Wi-Fi发送至上位机进行数据处理与姿态再现，从而实现手势识别、碰撞检测、物体交互等功能。经实验测试，该系统能够正确采集人体姿态数据，驱动虚拟人运动，具有采集数据节点多、动作精细、实时性强等特点，可应用于动漫、虚拟现实、影视、体感游戏等领域。

关键词：惯性传感器；动作捕捉；人体姿态；手势识别；碰撞检测；物体交互

中图分类号：TP212.9          文献标志码：A           文章编号：1674-2605(2023)03-0006-06

DOI：10.3969/j.issn.1674-2605.2023.03.006

Human Motion Capture System Based on Inertial Sensor 

WU Tianqiang  YUAN Xuejun  HU Wenqing 

(Guangdong Institute of Artificial Intelligence and Advanced Computing, Guangzhou 510535, China)

Abstract: A 47 node human Motion capture system based on MEMS inertial sensors is designed to solve the problems of the current human Motion capture system, such as the small number of acquisition nodes, the insufficient precision of motion sampling, and the high cost. The system collects real-time fine motion data of joints by wearing sensor nodes on the joints of the human body; After the motion data is gathered, it is sent to the upper computer through Wi Fi for data processing and pose reproduction, so as to achieve Gesture recognition, collision detection, object interaction and other functions. The experimental test shows that the system can correctly collect human body posture data and drive the movement of Virtual humans. It has the characteristics of multiple data collection nodes, fine action, and strong real-time, and can be applied to animation, virtual reality, film and television, and somatosensory games.

Keywords: inertial sensor; motion capture; human body posture; gesture recognition; collision detection; object interaction