一种基于肌声信号的穿戴式助力系统运动意图检测算法  

A Mechanomyogram(MMG)-based Motion Intention Recognition Algorithm for Wearable Exosuit

作　　者：董为[1] 石永军 林玮琪 DONG Wei;SHI Yongjun;LIN Weiqi(State Key Laboratory of Robotics and System,Harbin Institute of Technology,Harbin 150001,China)

机构地区：[1]哈尔滨工业大学机器人技术与系统国家重点实验室,哈尔滨150001

出　　处：《载人航天》2023年第1期43-51,共9页Manned Spaceflight

摘　　要：针对未来深空探测活动中航天员在多种复杂任务环境下的运动助力需求,提出一种面向航天员穿戴式助力系统的运动意图检测算法。以航天员的关节力矩作为运动意图的表征,利用希尔伯特-黄变换对特定肌肉发出的肌声信号进行滤波处理,以消除由肢体运动导致的伪迹噪声和由传感器引入的高频噪声,并参照肌肉的发力原理对滤波后的肌声信号进行特征值提取,通过机器学习的方法建立肌声信号与关节力矩间的映射关系,使助力系统能够及时准确地识别出航天员的运动意图并实施助力。最后募集了3名志愿者进行了150000组样本数据关节力矩辨识实验,实验结果表明:所提出算法的决定系数可达0.9532,能够有效辨识航天员的运动意图。A motion intention recognition algorithm for a wearable exosuit system was proposed in this paper to meet the motion assistance requirements of astronauts in various complex mission environments in future deep space exploration activities.To estimate the astronaut joint torque,a representation of motion intention,mechanomyogram(MMG)signals was employed,and then the Hilbert-Huang transform was used for eliminating the low-frequency noise caused by limb movement and the high-frequency noise caused by sensors.After MMG signal eigenvalue extraction,the mapping relationship between MMG signal and joint torque was established by machine learning method.Finally,an experiment based on multiple volunteers was performed to verify the proposed method.The experimental results showed that the proposed algorithm could effectively identify the astronaut’s motion intention with R2 of 0.9532.

关 键 词：运动意图辨识 肌声信号 信号处理 机器学习 

分 类 号：V57[航空宇航科学与技术—航空宇航推进理论与工程]