基于微型惯性传感器腿部康复动作捕捉系统研究

发布时间：2018-04-15 04:13

本文选题：MEMS惯性传感器 + 椭圆拟合　；参考：《哈尔滨工程大学》2014年硕士论文

【摘要】：基于微型惯性传感器腿部康复动作捕捉系统,其设计的意义在于使患者在康复训练时能够实时监护其运动状况,并通过再现本体运动来判断训练姿势是否合理。是动作捕捉技术在康复医疗领域中的一种应用。文中论述了现有的动作捕捉系统的优缺点,综合MEMS惯性元件特点以及惯导系统的姿态解算方法,在此基础上以腿部康复训练时的运动采集为任务,对惯性式动作捕捉系统系统设计、捕捉环境的要求、使用成本进行分析,得到采用基于MEMS惯性元件的动作捕捉方式来获取患者腿部姿态是一种较优越方法。研究工作主要内容如下:(1)进行了系统的结构设计,完成节点捕捉层,节点汇聚层和节点接收层体系结构分类,以CAN总线为基础,实现了节点捕捉层和节点汇聚层的网络有线连接。以Zigbee无线通信为基础,实现了节点汇聚层和节点接收层的无线连接。实验表明,这种网络结构可以有效完成各层间的通信。(2)以MPU9150集成惯性元件为核心结合STM32F103T8U6和CAN总线收发器VP230完成运动单元惯性数据捕捉节点的设计;基于上述所用芯片,结合CC2530F256无线SoC单片机,完成汇聚节点设计;以CC2531F256无线SoC单片机为核心完成接收节点设计。实验表明,系统可以完成运动信号的数据采集,处理和传输。(3)针对MPU9150惯性传感器的数据采集特点,设计FIR低通滤波器对传感器采集到的9轴数据进行滤波处理,实现对高频噪声的抑制。并针对MPU9150内置磁力计的灵敏度高,易受外界磁场干扰的问题,使用椭圆拟合的方法对磁力计进行误差补偿。实验表明,滤波及补偿后的数据在平滑性和稳定性方面具有提升。(4)以四元数法对惯性传感器9轴数据进行姿态解算,利用3轴加速度数据和3轴磁感应强度数据,实现姿态初始解算。基于3轴角速度数据,设计卡尔曼滤波器更新四元数,并结合基于高斯牛顿法,融合加速度数据和磁力计数据估计得到的四元数,设计了互补滤波器,对捕捉节点的姿态实时更新。实验表明,初始姿态的解算和实时姿态更新可以快速准确得到捕捉节点姿态角。系统在上位机虚拟现实环境中进行单捕捉节点和多捕捉节点的实验,实验表明,设计的系统能够对腿部的运动实时进行捕捉,数据传输稳定,采集并经过解算的数值可用于虚拟人体模型的连续驱动,满足康复动作训练的使用要求。
[Abstract]:Based on the micro inertial sensor leg rehabilitation motion capture system, the significance of its design is to enable patients to monitor their movement status in real time during rehabilitation training, and to judge whether the training posture is reasonable or not by reproducing self-motion.It is an application of motion capture technology in rehabilitation medical field.This paper discusses the advantages and disadvantages of the existing motion capture system, synthesizes the characteristics of the MEMS inertial element and the attitude calculation method of the inertial navigation system. On this basis, the task of motion acquisition during leg rehabilitation training is taken as the task.The design of the inertial motion capture system, the requirement of capturing environment and the cost analysis are analyzed. It is concluded that it is a better method to obtain the leg posture of the patient by using the motion capture method based on the MEMS inertial element.The main contents of the research are as follows: (1) the structure of the system is designed, and the architecture classification of node capture layer, node convergence layer and node receiving layer is completed, which is based on CAN bus.The network wired connection between node capture layer and node convergence layer is realized.Based on Zigbee wireless communication, wireless connection between node convergence layer and node receiving layer is realized.The experimental results show that this network structure can effectively complete the communication between the layers. (2) based on the MPU9150 integrated inertial element as the core, STM32F103T8U6 and CAN bus transceiver VP230 are used to complete the design of the inertia data capture node of the motion unit.Combined with CC2530F256 wireless SoC single chip computer, the design of convergent node is completed, and the receiving node design is completed with CC2531F256 wireless SoC single chip microcomputer as the core.Experiments show that the system can complete the data acquisition, processing and transmission of motion signals. (3) according to the data acquisition characteristics of MPU9150 inertial sensors, the FIR low-pass filter is designed to filter and process the 9-axis data collected by the sensor.The suppression of high frequency noise is realized.Aiming at the problem that the MPU9150 built-in magnetometer is sensitive and easily disturbed by external magnetic field, the elliptical fitting method is used to compensate the error of the magnetometer.Experimental results show that the smoothness and stability of the filtered and compensated data are improved. (4) Quaternion method is used to calculate the attitude of 9 axis data of inertial sensor, and 3 axis acceleration data and 3 axis magnetic induction intensity data are used.The initial attitude solution is realized.Based on the 3-axis angular velocity data, the Kalman filter is designed to update the quaternion, and the complementary filter is designed based on Gao Si Newton method, which combines the acceleration data and the magnetometer data to estimate the quaternion.Update the attitude of the capture node in real time.The experimental results show that the attitude angle of the captured node can be obtained quickly and accurately by the initial attitude calculation and real-time attitude updating.The experiment of single capture node and multi-capture node in the virtual reality environment of upper computer shows that the designed system can capture the leg motion in real time and the data transmission is stable.The values collected and solved can be used to drive the virtual human model continuously and meet the requirements of rehabilitation training.
【学位授予单位】：哈尔滨工程大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：R496;TP212;TN92

【参考文献】