人体运动捕捉和检测技术广泛应用于影视创作、电子游戏、动作分析、体育科研训练、康复医疗、虚拟现实、人机交互和机器人全自主控制等领域，具有十分广阔的应用前景。MEMS惯性传感器是微电子机械系统（Micro-Electro- Mechanical Systems，MEMS）传感器，具有小型化、低功耗、低成本和抗干扰能力强等优点。MEMS惯性传感器的出现和发展，促进了基于MEMS可穿戴式惯性传感器的人体运动捕捉和检测方法的发展。基于MEMS可穿戴式惯性传感器的人体运动捕捉和检测方法的基本原理是用户在身体各部位穿戴惯性测量单元，利用MEMS惯性传感器测量加速度和角速度等人体运动信息，通过解算得到人体运动的姿态角变化，进而利用模式识别方法对人体运动状态和姿态角进行识别分类，达到人体运动捕捉和检测的目的。相比于其他人体运动捕捉和检测方法，该方法具有轻便简洁、成本低廉、穿戴方便和实时性强等优点。但由于MEMS陀螺仪较大的漂移误差和运动加速度等因素的影响，姿态角估计计算精度不高，导致人体运动状态识别率低。基于此，本文开展基于MEMS可穿戴式惯性传感器的人体运动捕捉和检测方法的研究工作，以人体手臂运动状态测量和识别为目标，重点对高精度姿态角解算和模式识别的特征提取进行研究，以期实现高精度的人体运动状态检测和识别。
首先，论文确定了基于Zigbee无线传输的系统结构方案，在此基础上完成了系统方案总体设计。整个系统由基于MEMS传感器的运动检测模块、Zigbee无线网络数据通信模块以及上位机运动识别算法模块组成。MEMS传感器运动检测模块负责测量手臂运动过程中的角速度、加速度和磁场信息，通过Zigbee无线网络将测量数据上传到PC机，在PC机上基于Labview平台完成姿态角解算和运动识别。
姿态角的测量和解算是该方法的关键之一，论文对此进行了深入的分析和研究。阐述了惯性系统和姿态参考系统姿态角解算的优缺点；为了提高姿态角估计计算精度，论文利用卡尔曼滤波数据融合方法，把惯性系统和姿态参考系统有效地结合起来，并根据载体运动加速度的大小，适时调整卡尔曼滤波器的量测噪声方差的大小，以此减弱卡尔曼滤波过程中运动加速度对姿态角解算精度的影响，达到提高姿态角估计计算精度的目的。仿真分析和实验研究都表明，论文提出的基于预测调整观测噪声方差的卡尔曼滤波算法，比传统方法能获得更高的姿态角估计计算精度。在此基础上，论文分析了手臂运动方式，建立了手部运动轨迹解算模型，并具体对画横线、竖线、斜线和封闭线等四种典型手臂运动方式进行了分析，解算得到相应手腕处的运动轨迹。实验结果验证了该人体手臂运动状态解算模型的正确性。此外，论文还建立了轨迹计算误差模型，分析了姿态角对轨迹计算精度的影响情况。
运动状态的识别是另一个要解决的关键问题。传统的典型方法是提取传感器测量信号的时域特征作为模式识别算法的输入，但时域特征容易受到人体、环境等因素影响，导致特征值之间的差异不明显，模式识别的识别率不高。论文把手部运动轨迹特征值和传感器测量信号的时域特征值相结合，使用BP神经网络和支持向量机对画横线、竖线、斜线和封闭线等四种典型手臂运动状态进行了识别研究，识别率较传统方法得到明显提高，分别达到97.14%和100%。
在完成了MEMS传感器测量单元和算法研究的基础上，建立了基于Zigbee的无线传输测试系统。设计了具有数据接收、姿态角计算、轨迹计算、特征值提取、BP神经网络识别、显示和数据保存等功能的Labview软件，对整个系统功能进行了测试。测试结果表明， Zigbee网络无线数据传输正常，上位机数据检测、处理程序运行稳定，能实现四种手臂运动状态的正确识别。
 

Human motion capture and detection technology is widely used in film and television creation, video games, action analysis, sports research and training, rehabilitation, virtual reality, human-computer interaction, robot autonomy control and other fields. MEMS inertial sensor is a kind of Micro-Electro Mechanical Systems (MEMS) sensors, with miniaturization, low power consumption, low cost and anti-interference ability. The emergence and development of MEMS inertial sensors have promoted the development of human motion capture and detection methods based on MEMS wearable inertial sensors. For human body motion capture and detection method based on MEMS sensors, the MEMS inertial sensor are used to measure the acceleration and angular velocity information of human motion, and then the attitude angles of human motion are calculated. Further the pattern recognition method is used to identify the human body movements. Compared with other methods of human motion capture and detection, it has the advantages of simplicity, low cost, easy access to wear and nice performance of real-time. However, due to the large drift errors of MEMS inertial sensors, the accuracy of attitude angles is not often high, which further leads to a low recognition rate. Therefore, this paper carried out the research work of human body motion capture and detection method based on MEMS wearable inertial sensors, and aimed at the measuring and recognizing of arm movements. It focuses on a high accurate attitude angles calculation and an effective feature extraction of pattern recognition, to achieve a more accurate detection and identification of human body movements.
First of all, the paper determined the system structure scheme based on Zigbee wireless transmission. On this basis, the overall design of the system scheme was established. The whole system consists of the motion detection module based on MEMS sensors, the module of Zigbee wireless network for data communication and the module of host computer motion identification algorithm. The module of motion detection based on MEMS sensors was designed to measure the angular velocity, acceleration and magnetic field information when arm moving. The Zigbee wireless network then uploaded the measured data to the PC. The attitude angles were calculated and identified based on the Labview platform on the PC.
The measurement and calculation of attitude angles is one of the key of this method, and was deeply analyzed and researched in the paper. In respect of the attitudes calculation, the advantages and disadvantages of inertial system and the attitude reference system are compared and analyzed. In order to improve the accuracy of attitudes estimation, the paper used the Kalman filter data fusion method to combine the inertial system and the attitude reference system together effectively. And the value of the noise variance of the Kalman filter was adjusted according to the measured acceleration, which could weaken the influence of the motion acceleration on the attitude angles estimation, and then improve the accuracy of the attitudes estimation. Both the simulation analysis and the experimental study showed that the proposed Kalman filter algorithm with the capability to predict and adjust the observed noise variance could realize more accurate attitudes estimation than that with the traditional methods. On this basis, the paper further analyzed the arm movement model, and established the hand movement trajectory model. The four kinds of typical arm movement models were analyzed, and the trajectories of the wrist were obtained. The corresponding experiment verified the calculation model of arm movement. The paper also established the error model of trajectory calculation, and analyzed the influence of attitudes angle on the accuracy of trajectory calculation.
The identification of the state of movement is another key issue to be addressed. The traditional method is to extract the time domain acteristics of the signal measured from sensor as the input of the pattern recognition algorithm. The time domain feature is vulnerable to the human body and the environment, resulting in the difference between the eigenvalues unobvious, and thus makes the recognition rate of pattern recognition is not high. The features extracted from the hand trajectories were combined with the time domain features extracted from the measurement signals so that the recognition rates of the BP neural network and the SVM algorithm for four types of typical arm movements are increased and reach up to 97.14% and 100%, respectively.
Based on the preceding researches, a wireless transmission test system based on Zigbee technique was established. The corresponding Labview software was designed, which has the functions of receiving, displaying and saving of data, calculation of attitude angles, trajectories feature extraction, and identification of BP neural network. And the whole system functions were tested. The test results showed that wireless transmission based on Zigbee network technique was normal, and the host computer programs for data detection and processing were running stably, and realized the identification of four kinds of typical arm movements correctly.
 

[ ] 张海鹏．基于无线通信技术的人体运动捕捉系统研究与设计[D]．西安工程大学，硕士学位论文，2011：1-1

[ ] Klein A, De Assis G A. A Markeless Augmented Reality Tracking for Enhancing the User Interaction during Virtual Rehabilitation[C]// Xv Symposium on Virtual and Augmented Reality. IEEE Computer Society, 2013:117-124.

[ ] Zhou H, Hu H. Human motion tracking for rehabilitation—A survey[J]. Biomedical Signal Processing & Control, 2008, 3（1）:1-18.

[ ] Altun K, Barshan B, Tunçel O. Comparative study on classifying human activities with miniature inertial and magnetic sensors[J]. Pattern Recognition, 2010, 43（10）:3605-3620.

[ ] 陈姝．基于视频的人体运动跟踪与重构方法研究[D]．中南大学，博士学位论文，2008：3-7

[ ] 容志能．惯性运动捕捉系统中传感数据的传输与处理[D]．浙江大学，硕士学位论文，2012：7-15

[ ] Klette R, Tee G. Understanding Human Motion: A Historic Review. Technical Report 192, Communication and Information Technology Research group, University of Auckland, 2007.

[ ] Roetenberg D, Slycke P. A portable magnetic position and orientation tracker[J]. Sensors and Actuators A: Physical, 2007, 135（2）: 426-432.

[ ] King K, Yoon S W, Najafi K, etal. Wireless MEMS inertial sensor system for golf swing dynamics[J]. Sensors and Actuators A, 2008, 141: 619–630.

[ ] Seamana A, McPheea J. Comparison of Optical and Inertial Tracking of Full Golf Swings[J]. Procedia Engineering, 2012, 34: 461 – 466.

[ ] Yuan Q, Chen I M, Shang P L. SLAC: 3D localization of human based on kinetic human movement capture[C]// IEEE International Conference on Robotics and Automation. IEEE, 2011:848-853.

[ ] Roetenberg D., Baten, C.T.M., Veltink, P.H. Estimating body segment orientation by applying inertial and magnetic sensing near ferromagnetic materials[J]. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2007, 15（3）: 469–471.

[ ] Favre J, Jolles, B M., Aissaoui, R., etal. Ambulatory measurement of 3D knee joint angle[J]. Journal of Biomechanics, 2008, 41: 1029–1035.

[ ] Favre J, Crevoisier X., Jolles B M, et al. Evaluation of a mixed approach combining stationary and wearable systems to monitor gait over long distance[J]. Journal of Biomechanics, 2010, 43: 2196–2202.

[ ] Scapellato S, Cavallo F, Martelloni C, et al. In-use calibration of body mounted gyroscopes for applications in gait analysis[J]. Sensors and Actuators A:Physical, 2005, 123（3）: 418–422.

[ ] Chardonnens J, Favre J, Cuendet F. A system to measure the kinematics during the entire ski jump sequence using inertial sensors.[J]. Journal of Biomechanics, 2013, 46（1）:56-62.

[ ] Nguyen K D, Luo Z Q, Song H Y, et al. A wearable sensing system for tracking and monitoring of functional arm movement[J]. IEEE/ASME Transactions on Mechatronics, 2011, 16（2）: 213–220.

[ ] 路永乐, 张欣, 龚爽,等. 基于MEMS惯性传感器的人体多运动模式识别[J]. 中国惯性技术学报, 2016, 24（5）:589-594.

[ ] 刘震,王雪梅,倪文波. 基于MEMS传感器的高精度姿态角测量研究[J]. 中国测试,2017, 43 (2):6-12.

[ ] 陈旭光, 杨平, 陈意. MEMS陀螺仪零位误差分析与处理[J]. 传感技术学报, 2012, 25（5）:628-632.

[ ] 张秋阳. 无人机姿态测算及其误差补偿研究[D]. 中南大学, 2011.

[ ] 卢志才, 高敏, 贾春宁. 基于地磁测量的火箭弹滚转角解算误差补偿方法研究[J]. 中国测试, 2014, 40（2）:9-12.

[ ] Einsmann C, Quirk M, Muzal B, et al. Modeling a wearable full-body motion capture system[C]// IEEE International Symposium on Wearable Computers, 2005. Proceedings. IEEE, 2005:144-151.

[ ] FOERSTER F, SMEJA M, FAHRENBERG J. Detection of posture and motion by accelerometry: a validation study in ambulatory monitoring [J]. Computers in Human Behavior, 1999, 15（5）:571-583.

[ ] Mantyjarvi J, Himberg J, Seppanen T. Recognizing human motion with multiple acceleration sensors[C]// IEEE International Conference on Systems, Man, and Cybernetics. IEEE, 2001:747-752 vol.2.

[ ] Slyper R, Hodgins J K. Action capture with accelerometers[C]// Eurographics/acm SIGGRAPH Symposium on Computer Animation, SCA 2008, Dublin, Ireland. DBLP, 2008:193-199.

[ ] V. Lympourides, D. K. Arvind, M. Parker. Fully wireless,full body 3-d motion capture for improvisational performances[C].In Proceedings of the Workshop on Whole Body Interaction.Boston,USA:ACM,2009:221-233.

[ ] Sakaguchi T, Kanamori T, Katayose H, et al. Human motion capture by integrating gyroscopes and accelerometers[C]// Ieee/sice/rsj International Conference on Multisensor Fusion and Integration for Intelligent Systems. IEEE, 1996:470-475.

[ ] Bachmann E R, Mcghee R B, Yun X, et al. Inertial and magnetic posture tracking for ing humans into networked virtual environments[C]// VRST '01 Proceedings of the ACM symposium on Virtual reality software and technology. 2001:9-16.

[ ] Schwarz L A, Bigdelou A, Navab N. Learning Gestures for Customizable Human- Computer Interaction in the Operating Room[M].Medical Image Computing and Computer-Assisted Intervention – MICCAI 2011. Springer Berlin Heidelberg, 2011:129-136.

[ ] Dowling A V, Favre J, Andriacchi T P. Inertial sensor-based feedback can reduce key risk metrics for anterior cruciate ligament injury during jump landings.[J]. American Journal of Sports Medicine, 2012, 40（5）:1075-1083.

[ ] Lim C K, Luo Z, Chen I M, et al. Wearable wireless sensing system for capturing human arm motion[J]. Sensors & Actuators, 2011, 166（1）:125-132.

[ ] Innalabs,Inc.2009.3Dsuit惯性运动捕捉系统.http://www.3dsuit.com

[ ] Xsens,Inc.2009.Moven惯性动作捕捉系统. http://www.xsens.com/en/general/mvn

[ ] Brahem M B, Ménélas B A J, Otis J D. Use of a 3DOF Accelerometer for Foot Tracking and Gesture Recognition in Mobile HCI[J]. Procedia Computer Science, 2013, 19:453-460.

[ ] Tao G, Sun S, Huang S, et al. Human modeling and real-time motion reconstruction for micro-sensor motion capture[C]// IEEE International Conference on Virtual Environments, Human-Computer Interfaces and Measurement Systems Proceedings. IEEE, 2011:1-5.

[ ] 吴健康. 人体运动捕获三维再现系统:, CN 201431466 Y[P]. 2010.

[ ] LiG, Wu Z,MengX,et al. Modeling of human body for animation by micro-sensor motion capture[C].Knowledge Acquisition and Modeling, 2009. KAM'09. Second International Symposium on. IEEE, 2009,3: 98-101.

[ ] 李启雷，金文光,耿卫东.基于无线惯性传感器的人体动作捕获方法浙江大学学报:工学版,2012,46（002）: 280-285.

[ ] 王科俊,陈玮.一种可穿戴式人体运动捕捉系统的设计与实现[J].黑龙江大学工程学报,2013,4（2）: 97-102.

[ ] Tao Y. Ben W. Accelerometer-based Chinses Traffic Police[J].Chinese Joumal of Electronics,2010.9（2）

[ ] 张丽娜．基于加速度传感器的步态特征身份识别[D].沈阳工业大学，2014

[ ] 徐昺．基于MEMS传感器与Zigbee网络的人体动作捕捉系统的设计与实现[D]，电子科技大学，2013

[ ] R.S. Bucy, K.D. Senne. Digital synthesis of non-linear filters [J]. Automatica, 1970, 7（3）:287-298.

[ ] Julier S, Uhlmann J, Durrantwhyte H F. A new method for the nonlinear transformation of means and covariances in filters and estimators[J]. IEEE Transactions on Automatic Control, 2000, 45（3）:477-482.

[ ] Baritzhack I Y, Goshenmeskin D. A unified approach to inertial navigation system error modeling[C]// AIAA Guidance, Navigation and Control Conference. AIAA Guidance, Navigation and Control Conference, 1990:472-480.

[ ] 霍元正. MEMS陀螺仪随机漂移误差补偿技术的研究[D]. 东南大学, 2015.

[ ] Zhu R, Zhou Z. A Small Low-Cost Hybrid Orientation System and Its Error Analysis[J]. 2009, 9（3）:223-230.

[ ] 杨晓东, 王炜. 地磁导航原理[M]. 北京：国防工业出版社, 2009.

[ ] 秦伟伟, 郑志强, 刘刚,等. 基于小波分析与LSSVM的陀螺仪随机漂移建模[J]. 中国惯性技术学报, 2008, 16(6):721-724.

[ ] 杨新勇, 黄圣国. 智能磁航向传感器的研制及误差补偿算法分析[J]. 北京航空航天大学学报, 2004, 30(3):244-248.

[ ] 陈瑜, 张铁民, 彭孝东. 一种磁阻式电子罗盘测试和标定方法研究[J]. 传感技术学报, 2014, 27（5）:622-626.

[ ] 秦永元. 惯性导航[M]. （第二版） . 北京：科学出版社, 2014.

[ ] 孙丽, 秦永元. 捷联惯导系统姿态算法比较[J]. 中国惯性技术学报, 2006, 14(3):6-10.

[ ] 陈哲. 捷联惯导系统原理[M]. 北京：宇航出版社, 1986.

[ ] 米刚, 田增山, 金悦,等. 基于MIMU和磁力计的姿态更新算法研究[J]. 传感技术学报, 2015（1）:43-48.

[ ] 朱荣, 周兆英. 基于MEMS的姿态测量系统[J]. 测控技术, 2002, 21(10):6-8.

[ ] Vasconcelos J F, Cardeira B, Silvestre C, et al. Discrete-Time Complementary Filters for Attitude and Position Estimation: Design, Analysis and Experimental Validation.[J]. IEEE Transactions on Control Systems Technology, 2011, 19（1）:181-198.

[ ] Brigante C M N, Abbate N, Basile A, et al. Towards Miniaturization of a MEMS-Based Wearable Motion Capture System[J]. IEEE Transactions on Industrial Electronics, 2011, 58（8）:3234-3241.

[ ] Gao C, Zhao G, Lu J. In-flight misalignment attitude estimation for UAV based on GPS and UKF/MPF filter[C]// Guidance, Navigation and Control Conference. IEEE, 2014:452-457.

[ ] 杨勇. 手臂康复机器人系统研究[D]. 哈尔滨工程大学, 2009.

[ ] 丛爽. 面向MATLAB工具箱的神经网络理论与应用-第3版[M]. 合肥：中国科学技术大学出版社,2009.

[ ] 朱彦平, 陈善超, 严良文,等. BP神经网络手势动作识别在脑卒中患者手臂训练中的应用[J]. 机械制造, 2015, 53(10):18-20.

[ ] Vapnik V N. The nature of statistical learning theory[J]. IEEE Transactions on Neural Networks, 1995, 8（6）:1564-1564.

[ ] 洪涛, 黄志奇, 杨畅. 涡轮泵实时故障检测的快速支持向量机算法[J]. 仪器仪表学报, 2012, 33（8）:1786-1792.