An Outlier Robust Finite Impulse Response Filter With Maximum Correntropy

Non-Gaussian noise is common in industrial applications, and it a severe challenge to existing state estimators. In this paper, novel robust maximum correntropy finite impulse response (MCFIR) filter proposed deal with the estimation problem linear state-space system corrupted by outliers. The operates as memory form, thus obtains superior immunity statistics process uncertainties than Kalman-like filters. Gaussian adopted generate new cost function, which improves robustness outlier interference. We derive an unbiased MCFIR that ignores propose improvement bias-constrained achieve better estimate accuracy. To improve filtering performance degradation caused improper kernel size, adaptive size algorithm further proposed, adjusts bandwidth within specific range adaptively achieves significant filter. An illustrative example based on moving target tracking presented evaluate of filter, simulation results confirmed obtained outliers