MEDIAN LEAST MEAN SQUARES LATTICE (MLMSL) ADAPTIVE FILTERS

An efficient algorithm for updating the gradient adaptive lattice (GAL ) filter, termed the median least mean squares lattice (MLMSL) adaptiv e filter, is presented. The update in the proposed algorithm is achiev ed by employing the sample median of the gradient estimates at each st age of the lattice instead of the instantaneous value as in convention al GAL. Two structures are considered: the first is the normal lattice while the second incorporates an additional FIR configuration. It is shown that the MLMSL performs similar to the GAL in the absence of imp ulsive interference, and outperforms the GAL in the presence of such i nterference. Simulations are provided to demonstrate the efficiency of the proposed MLMSL adaptive filters.