Research On Robust Kalman Algorithm And Its Applications

Kalman filter is widely used in numerous engineering applications, ranging from radartechnology, computer vision, and to other engineering projects related to signal processing.Meanwhile, it is an important concept in control theory and control system of the project.Kalman filter is the optimal linear estimator in the sense of least-square error. It uses aninverse process, not only to handle the estimation problem of a smooth process, but also toensure the minimum mean square error under conditions of multidimensional andnon-stationary random process.In this dissertation, I firstly introduce the basic theory of Kalman filter, and thenconsidering the shortcomings and limitations of the tradition Kalman filter, a Kalman filterwith on-line parameter adjustment was put forward, the simulation result of target trackinginstance proves the validity of the algorithm. The problem of abnormal event detection intarget tracking process is ever-present, such as sensor failures, measurement outliers, or evenintentional jamming; these will ensure the generation of sparse noise. There exist the sparsenoises in the process of speech enhancement. In this dissertation, we develop a robustKalman as our filter algorithm for non-stationary temporal sparse noise estimation, theproposed method with the solution of convex optimization to solve the calculation of thetradition Kalman and robust Kalman, so as to get the better estimation of sparse noise.In order to make the robust Kalman more effectively in engineering application, wedevelop a numerical iterative algorithm called Separable Surrogate Function (SSF) method toupdate the robust Kalman algorithm. Compared with traditional Kalman filter, the simulationresults showed that robust Kalman filter is superiority and effective in two applications whichare target tracking and speech enhancement.