A real-time data processing method based on twice weighted improved Kalman filter algorithm

Abstract

The data obtained by sensor is strong randomness and deviates from Gaussian distribution. The Extended Kalman Filter (EKF) is cost-effective and easy to cause large errors while processing data. At the same time, the accuracy of the data prediction is not high enough. Therefore, an improved filter based on the original EKF is proposed to improve the prediction accuracy and noise reduction effect in this paper, which combines the predicted value and the measured value of the EKF twice to find the root mean square to obtain a more accurate predicted value. The noise reduction effect of EKF and TWEKF is compared by using MATLAB. And the root means square error (RMSE), signal-to-noise ratio (SNR), smoothness (S) and other quantitative judgment indexes prove that TWEKF has better noise reduction effect than EKF.