人体步态分析的多传感器数据融合研究

发布时间：2018-03-25 07:06

  本文选题：多传感器数据融合　切入点：人体步态分析　出处：《大连理工大学》2016年博士论文

【摘要】：人的步行过程包含着丰富的运动学、动力学和生理学信息,对这些信息进行特征提取和模式分析就可以获得人体步态的描述。通过科学的手段测量人的步态具有重要的现实意义,比如可以用于临床医学中的康复评定、下肢辅具设计和下肢机器人的开发等。本文提出了基于自行开发的可穿戴步态分析系统的步态分析方法。该系统可以采集下肢运动过程中的速度和方向变化信息。通过多层次的传感器数据融合算法进行姿态解算,步行时相划分以及误差校正,可以精确地计算各项步态参数。本文在总结国内外已有研究内容的基础上提出和解决了以下问题：如何减少姿态解算误差,增加步态时相检测精度；如何全面地评价双足对称性、步态稳定性；如何消除传感器绑定位置误差以及多个传感器的对准误差；以及零速度更新算法的适用性问题。主要研究内容如下：1.本研究使用微机电传感器并结合人体传感器网络研制了可穿戴步态分析平台,在传感器的灵敏度标定,非正交误差补偿以及粗大误差筛除等方面做了相关的研究。多数研究者忽视了传感器绑定位置对步态分析的影响,本文提出了一种基于轴线对准(Boresighting)的传感器初始对准方法,可以很好地消除初始绑定误差对于参数估算的影响,也可以用于下肢运动分析的多传感器初始对准。2.研究了基于扩展卡尔曼数字滤波器的融合算法减少系统噪声和传感器偏差对步态测量的影响；在站立期使用零速度更新算法校正速度和位置积分误差；针对传感器数据波动导致的步态时相误检测的问题,使用多阈值方法进行约束,提高步态时相检测精度,为精确计算各项步态参数奠定基础。3.目前大多数的步态分析研究缺少对于双足运动协同性的分析。本文对于放置在双侧足部的两个传感器模块进行传感器数据融合,研究了人体步行过程中双足的协调与交替运动规律。并创新性地使用动态时间规整(Dynamic time warping)算法融合双足数据,对于双足步态对称性、步态稳定性与一致性提供了一种新的评价方法。4.关于在步态分析领域广泛使用的零速度更新算法(ZUPT)的适用性问题,本文提出了一种扩展零速度更新算法(E-ZUPT),将三个传感器模块分别放置在单侧下肢的不同部位,使用迪纳维特-哈坦伯格法建立下肢运动学模型,建立约束,在检测到完全站立相之后融合下肢运动信号,通过求解欧拉-拉格朗日方程来解决位置误差最小化问题,从而使零速度更新算法的适用范围扩展到足部之外的其他下肢部位。
[Abstract]:The human walking process contains a wealth of kinematics, dynamics, and physiological information. The description of human gait can be obtained by feature extraction and pattern analysis of this information. It is of great practical significance to measure human gait by scientific means, for example, it can be used for rehabilitation evaluation in clinical medicine. The design of lower extremity accessory and the development of lower limb robot, etc. In this paper, a gait analysis method based on self-developed wearable gait analysis system is proposed. The system can collect the information of velocity and direction change during the movement of lower extremity. Through multi-level sensor data fusion algorithm for attitude resolution, Walking phase division and error correction can accurately calculate the gait parameters. In this paper, the following problems are put forward and solved: how to reduce the error of attitude resolution, based on the summarization of existing research contents at home and abroad. How to improve the accuracy of gait phase detection, how to evaluate the biped symmetry and gait stability, how to eliminate the sensor binding position error and the alignment error of multiple sensors; The main contents of this study are as follows: 1.The wearable gait analysis platform is developed by using micro electromechanical sensor and human body sensor network, and the sensitivity of sensor is calibrated. Many researchers have neglected the influence of sensor binding position on gait analysis. In this paper, an initial alignment method based on axis alignment is proposed. The effect of initial binding error on parameter estimation is well eliminated. The fusion algorithm based on extended Kalman filter (EKF) is used to reduce the influence of system noise and sensor deviation on gait measurement. In the standing period, the zero-velocity updating algorithm is used to correct the velocity and position integral error, and to solve the problem of gait phase error detection caused by the fluctuation of sensor data, the multi-threshold method is used to restrain the gait phase detection accuracy, so as to improve the gait phase detection accuracy. In order to calculate the gait parameters accurately. 3. Most of the gait analysis research lacks the analysis of bipedal motion cooperation. In this paper, two sensor modules placed in the bilateral foot are fused to the sensor data. In this paper, we study the law of coordination and alternating motion of two feet during human walking, and creatively use dynamic time warping algorithm to fuse biped data for biped gait symmetry. A new evaluation method for gait stability and consistency is provided. (4) on the applicability of Zero-Velocity updating algorithm (ZUPT), which is widely used in gait analysis, In this paper, an extended zero-velocity updating algorithm, E-ZUPTT, is proposed. The three sensor modules are placed in different parts of the lower extremity on one side, and the kinematics model of the lower extremity is established by using the Dinawitt-Hartenberg method, and the constraints are established. After detecting the fully standing phase, the lower limb motion signal is fused, and the position error minimization problem is solved by solving the Euler-Lagrangian equation, so that the applicable range of the zero velocity updating algorithm is extended to other lower extremity parts outside the foot.
【学位授予单位】：大连理工大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP202;TP212
，

本文编号：1662045