Research on posture recognition based on Kalman filter and quaternion

The accuracy of quaternion method in strap-down inertial navigation used for posture recognition is subjected to sensor data drift. Analysis shows that quaternion is a state variable in discrete time linear constant system. A method based on Kalman filter to correct the error is proposed. Recursive calculations on state variables can effectively reduce system error. Experiments show that this method significantly increases the accuracy of posture recognition.