Human Motion Capture Technology Based On Wsn And Inertial Sensor

The currently typical inertial motion capture system is composed of MEMSgyroscopes, MEMS accelerometer and magnetometer, and different sensor mutuallycomplement with each other. When the carrier is in a different motion state, theestimation algorithm will give the optical estimation of the carrier orientation based onthe output of sensors. For the low cost of the sensor, the sensor output is disturbed byvarious errors, including measurement error and environmental disturbances. On theother hand, the currently orientation estimation algorithm is based on the PC, and themodel is complexity so that the computation burden is obviously beyond the capability ofthe embedded system, especially the matrix operations. Therefore, the thesis is focus onthe sensor static correction and the improvement of the orientation estimation algorithm.According to the error model of accelerometer and magnetometer, the thesisintroduces the static calibration based on the PSO algorithm to improve estimationaccuracy of static calibration. The new algorithm not only avoids the complexmathematical models, but also simplifies the sampling operation of the original data.Experimental results show that the static calibration based on the PSO is accurate, easy tooperate, and provides a more accurate data for the orientation estimation algorithm.For the different characteristics of sensors composed of inertial motion capturesystem, the thesis introduces a Two-steps Kalman filter, which treats accelerometer andmagnetometer separately. The nine dimensional matrix operations are also reduced tothree dimensional. The algorithm improves and simplifies the current orientationestimation algorithm, increases the computing speed, so that the orientation estimationalgorithms can be implemented in embedded systems. In order to improve the estimationaccuracy of the filter, an adaptive adjustment method based on fuzzy logic is proposed.For the different motion state, there is a corresponding adjustment factor, so theadjustment of the filter is converted to the motion state recognition. The adaptiveadjustment method ensures the estimation accuracy under different motion states andreduces the computational burden. In order to test the performance of the adaptive Two-steps Kalman filter in differentenvironments, five sets of experiments were designed，including dynamic, static, tappingvibration and electromagnetic disturbance. Experimental results show that, the filter canguarantee orientation estimation accuracy in different motion state, has better faulttolerance for external disturbance.