FAST ADAPTATION AND PERFORMANCE-CHARACTERISTICS OF FIR-WOS HYBRID FILTERS

Fast adaptive algorithms are developed for training Weighted Order Sta tistic (WOS) filters and FIR-WOS Hybrid (FWH) filters under the mean a bsolute error (MAE) criterion. These algorithms are based on the thres hold decomposition of real-valued signals introduced in this paper. Wi th this method an N-length WOS filter can be implemented by thresholdi ng the input signals at most N times independent of the accuracy used. Beside saving in computations, the proposed algorithms can be applied to process arbitrary real-valued signals directly. Performance charac teristics of FWH filters in 1-D and 2-D signal restoration are investi gated through computer simulations. We show that both in restoration o f signals containing edges and in the case of heavy tailed nonGaussian noise, considerable improvement in performance can be achieved with F WH filters over WOS filters, Ll filters, and adaptive linear filters. Two new FWH filter design strategies are found for removal of impulsiv e noise and for restoration of a square wave, respectively.