The Analysis Of Colored Observation Noise Filtering

In many aspects such as the dynamic navigation and positioning, target tracking and recognition, wireless communication and so on, filtering is the most effective and most commonly used algorithm. It has several characteristics as high calculation efficiency, small memory, simple and convenient, so it is suitable for applications in dynamic data processing. But generally the filtering study is under ideal conditions of Gauss and the linear, practical application often can not meet these conditions. So, study of the filtering algorithm under non Gauss and nonlinear conditions has important theoretical significance and practical value.Under the conditions of that the observation noise is colored noise, and combining the simulated data and the measured data, this article researches on the theory and algorithm of three filter methods. Three methods are standard Kalman Filter under linear conditions, unscented Kalman Filter under nonlinear conditions and Particle Filter under nonlinear conditions. Main structure and contents of this article are as follows:1. The paper summarizes and classifies the existing variety of colored noise filtering algorithms, and briefly describes the characteristics of various filtering algorithms. The advantages and disadvantages of these theories and algorithms are also analyzed.2. Assuming the conditions under a linear system and colored measurement noise on the basis of ordinary AR model, new observation information which is referenced by standard Kalman filter algorithm is used to suppress noise. Then, verified the feasibility of the algorithm and model by calculating the simulated and measured data. At last, the result shows that the algorithm can reduce the impact of colored noise.3. Assuming the conditions under a nonlinear system and colored measurement noise on the basis of ordinary AR model, this paper calculated the nonlinear posterior mean state by UT transformation based on the increased measurement information and minimum variance estimation. Then, come the recursive formula of unscented Kalman filter which can deal with the colored measurement noises. The feasibility of the algorithm is verified by simulated and measured data count through the strong nonlinear system.4. Assuming the conditions under a nonlinear system and colored measurement noise on the basis of ordinary AR model, the Particle filter can reduce the impact of colored measurement noise based on principles of Particle filter while its importance weights be analyzed and processed. The feasibility of the algorithm is verified by simulated and measured data count through the nonlinear system.