Research On Human Attitude Tracking Based On MEMS Sensors

The human attitude tracking technology based on the micro inertial attitude measurement is a kind of wearable dynamic tracking technology which is different form these tracking technology based on optical, acoustic or electromagnetic theory. Comparing with the traditional attitude tracking technology, the dynamic tracking technology has many advantages, such as high real-time performance, not subject to regional restriction, easy to install, etc, and becomes one of the spot of research at home and abroad in recent years. As micro inertial sensor having smaller size and lower cost, its accuracy and anti interference ability is limited. The developments of multi-sensor data fusion theory and method have laid a solid foundation for human attitude tracking system based on micro inertial attitude measurement.This paper use an attitude measurement unit made of MEMS accelerometer, MEMS gyroscope and magnetometer to realize the human attitude tracking measurement. The main contents include two parts: MEMS sensor error analysis and correction,the human body posture tracking algorithm improvement and demonstration of the results.The study of the micro inertial sensors and the micro inertial attitude tracking system is introduced. Then, the inertial attitude methods are analyzed and compared. For quaternion without singularity, the quaternion updating algorithm is elaborated in detail.This paper analyzes the error sources of accelerometer and magnetometer. According to the characteristics of error sources, the static error model based on ellipsoid constraints is established. The ant colony optimization algorithm is used to get the model parameters and complete the error calibration experiment. Compared with the least-square method, the ant colony algorithm can better correct sensor error model.On the basis of inertial sensor theory, the reference coordinates and vector coordinate for human body posture are put forward. The characteristics of accelerometer, magnetometer and gyroscope are analyzed in the process of attitude tracking and update. Complementary filter algorithm and Kalman filter algorithm are used to analyze the attitude data to get the optimal attitude estimation. On the basis of human skeleton model the human arm as the research object, the geometric constraints are analyzed in the process of the human arm movement, and the constraints are introduced to reduce the error caused by muscle deformation.Through the interface demonstration system real-time display the arm gesture tracking results and the quantitative analysis of attitude data. The experimental result shows that the attitude solution algorithm and human constraints model designed has a higher accuracy.