OWA Aggregation Operator in Robust Filtering

T. Pander (Silesian Univ. of Techn., Poland), J. Wróbel (ITAM Zabrze, Poland)

In this paper, an overview of the robust filtering problem is provided and the ordered weighted aggregation (OWA) operator is described. This operator is applied as a weighted average of ordered signal samples in a sliding window to produce output values of a filter. The appropriate selection of weights allows to efficiently suppress the impulsive noise. This process is similar to the moving average smoothing however this one is very sensitive to outliers. Several weighting window functions have been tested due to the ability to minimize impact of outliers. Introducing the non-linear sorting process and attributing to the sorted samples appropriate values of weights this leads to improvement of filtering in an impulsive environment. Obtained results show the effectiveness of the proposed robust filtering method in comparison to the reference filters.