Robust Asymmetric Safety Kalman Filter for MIMO Radar Resisting Temporary Outliers

Abstract

This study presents a novel concept of robust Kalman filtering, especially dealing with outliers when detecting human hand intrusion using multiple-input–multiple-output radar. Outliers that severely deviate from the normal value for a certain period have not been sufficiently discussed from the viewpoint of safety and can seriously hamper the performance of the Kalman filter. Therefore, two types of outliers were intensively discussed: the temporary outlier that originated in misdetection and the additive outlier that is generated by heavy-tailed error distribution noise. Moreover, an asymmetric outlier processing strategy was proposed depending on positive and negative outliers. The result shows that the proposed method can resist the additive and temporary outliers.