EKF with Measurement Noise Estimation Based on Wavelet Transform and Application for Target Tracking

The measurement noise variance in the process of EKF is prone to bring error accumulation and can lead to filter divergence. Aiming at this kind of shortcoming, in this paper we build model of target motion observed on a single measurement point in a two-dimensional plane firstly. Secondly, we compare two methods, the variance estimation based on the signal-to-noise separation of wavelet transform and EKF algorithm based on noise variance estimation, applying in target tracking. Then, we adopt the wavelet transform to distinguish noise from the measurement signal real-timely. And the median variance estimator is used to estimate the measurement noise, which can improve the precision in EKF of target tracking by combining with EKF. Finally, the method of Monte Carlo simulation is used to prove its effectiveness and practicality.