基于机器学习的人体虚拟惯性测量组件构建方法

发布时间：2018-10-29 20:25

【摘要】：利用人体特征辅助行人导航与外骨骼机器人控制是近年来导航与机器人领域中的热点研究方向。针对惯性测量组件足部安装方式在过载较高时无法实现有效测量的问题,研究了一种基于机器学习的人体虚拟惯性测量组件构建方法。该方法以同步采集安装于足部与下肢其他部位的惯性测量组件的输出作为数据样本,通过遗传算法改进的误差反向传播(GA-BP)神经网络实现虚拟惯性测量组件的构建。为进一步改善训练效果,采用基于步态相位检测方法对训练样本进行筛选。基于Anybody与MATLAB的联合仿真结果表明,本文所研究的方法可实现采用安装于髋关节附近位置的惯性测量组件数据,有效模拟足部位置的惯性测量组件数据。该方法对未经训练的步态也有一定的适应性。本文所研究的方法可进一步应用于行人精确定位与外骨骼机器人控制等领域。
[Abstract]:The application of human features in pedestrian navigation and exoskeleton robot control is a hot research direction in the field of navigation and robot in recent years. Aiming at the problem that the foot mounting mode of the inertial measurement module can not realize the effective measurement when the overload is high, a method of constructing the virtual inertial measurement module of human body based on machine learning is studied in this paper. In this method, the output of the inertial measurement module installed in the foot and other parts of the lower limb is taken as the data sample, and the virtual inertial measurement module is constructed by the improved error back Propagation (GA-BP) neural network based on genetic algorithm. In order to improve the training effect, the gait phase detection method was used to screen the training samples. The simulation results based on Anybody and MATLAB show that the method studied in this paper can realize the inertial measurement module data installed near the hip joint and simulate the foot position inertial measurement module data effectively. This method also has certain adaptability to the untrained gait. The method studied in this paper can be applied to the precise location of pedestrians and the control of exoskeleton robot.
【作者单位】：南京师范大学电气与自动化工程学院;东南大学仪器科学与工程学院;
【基金】：国家自然科学基金(61304227,61503077,61601228) 江苏省自然科学基金(BK20141453,BK20161021)
【分类号】：TN96;TP18

【参考文献】