Research And Implementation Of Attitude Measurement Algorithm For Human Motion Capture

The somatosensory interaction is one of the main forms of the new generation of human-computer interaction.In order to achieve interaction of human motion and computer,the first step is to capture the motion of human body.For the sensor data capture and motion attitude analysis in MEMS inertial human motion capture system,attitude measurement unit design,sensor error calibration,attitude estimation and information fusion algorithm are studied in detail.Firstly,the attitude measurement unit is designed to acquire the raw motion data such as acceleration,angular velocity and magnetic field component.Based on the consideration of scalability and real-time performance,a data communication mechanism based on asynchronous acquisition and time division multiplexing is designed.Aiming at the characteristic that the magnetic sensor in the attitude measurement unit is susceptible to interference from its own instrumentation error and the external environment,a calibration model based on the ellipsoidal fitting is established and the error is effectively compensated.Aiming at the problem that random drift is difficult to be directly quantized in the gyroscope,an ARMA model is established and Kalman filter is used to eliminate the random drift.Based on the comparison of the direct calculation solution of using gyroscope integration and magnetic sensor tilt compensation,two attitude estimation methods with multisensor fusion are implemented based on gradient descent algorithm and complementary filter&PI control algorithm respectively.The attitude estimation precision is tested.The test results show that the two multisensor fusion methods achieve 0.5° static precision and 2° dynamic precision.Compared with the other common Kalman filter,the two fusion methods can achieve the corresponding accuracy and the complexity is smaller,which are more suitable for low power embedded systems.