Study On The Key Technologies For Inertial Human Motion Sensor

As one of the most critical equipment for motion capture, human motion sensor has beena great concern of many experts and researchers in recent years. The human motion sensor isdivided into five categories from the working principle for now: mechanical electrical,electromagnetic, acoustic, optical and inertial type, their data accuracy, number of movingobjects they can capture, real-time, capture environment site, ease of use and themanufacturing costs are present on their respective merits. With the development of MEMS(Micro Electro-Mechanical System) technology, the advantages of inertial human motionsensor are more prominent. Therefore, this paper studies on the key technologies of inertialhuman motion sensor in human motion capture system.Firstly, the structure and composition of the motion capture system are discussed for thecurrent mainstream inertial human motion capture system. Terminal nodes, routing nodes,host nodes and Zigbee protocol stack WBAN (Wireless footer area network) conducted adetailed analysis and built a set of inertial human motion sensor hardware platform. Theplatform uses inertial measurement unit consisting of three-axis MEMS accelerometer,three-axis gyroscope and three-axis magnetometer to measure the movement of the sensorparameters, and transmission data through WBAN, which is the basis for this research work.Then, in order to get accurate raw output signals of each inertial device, the workingprinciples of MEMS accelerometer, gyroscope, magnetometer are introduced. The medianfilter processing is used because the raw data of the accelerometer contain impulse noiseinterference characteristics, the mean filter processing is used because the raw data of thegyroscope contain frequency noise, the Butterworth filter processing is used because the rawdata of the magnetometer contain more randon noise. The experiments indicate that thepreprocessing significantly reduced the interference and lay the foundation for furtherprocessing.Thirdly, to improve the accuracy and reliability of MEMS sensor, the sensor needs to becalibrated. The error models of MEMS accelerometer, gyroscope and magnetometer aredesigned. Static six-potion calibration method is carried out to determine the error modelcoefficients for the accelerometer static output characteristics. Dynamic multi-rate calibrationmethod is carried out to determine the error model coefficients for the gyroscope dynamicoutput characteristics. Free rotation calibration method is carried out to determine the errormodel coefficients for the magnetometer output data characteristics. The experiments show that the calibration improves the accuracy of the devices.Fourthly, to capture the human trajectory in three-dimensional space, attitude algorithmis needed to get the attitude from these inertial data. The filters attitude algorithm commonlyuse are complementary filter and Kalman filter, the methods attitude algorithm commonly useare Euler angle method and quaternion method. Therefore, four attitude solutions aredesigned: Euler Angle based complementary filter, Quaternion based complementary filter,Euler Angle based Kalman Filter, and Quaternion based Kalman Filter. The experimentsindicate that the Quaternion based Kalman Filter has the best performance.Finally, a set of attitude reference system using the total station, attitude referenceplatform and geological compass is designed to verify the inertial human motion sensoraccuracy. In accordance with this, the arbitrary rotation experiment, singular point experimentand magnetic mutation experiment output results were compared to verify the attitudealgorithm. In addition, we visualize the sensor output data based on OpenGL, the armmovements of experimenter are restituted vividly and have a good human-computerinteraction effects, once again prove the feasibility of the system.