A fast convergence algorithm for adaptive FIR filters undercomputational constraint for adaptive tap-position control

摘要

This paper proposes a fast convergence algorithm for adaptive FIRnfilters with tap-position control. The proposed algorithm consists ofntwo stages: flat-delay estimation and constrained tap-position control.nIn the flat-delay estimation, the scattered coefficients are allowed tonchange their positions to achieve fast and correct flat-delaynestimation. For constrained tap-position control, special attention isnpaid to a limit in computational power imposed by the hardware. Byndividing a first-in-first-out queue into two parts, which store indexesnto inactive taps with no assigned coefficient, fast convergence isnachieved even when computation per sampling period for tap-positionncontrol is limited. Simulation results show that under the samencomputational limit as the conventional algorithm, the proposednalgorithm reduces the convergence time by as much as 60%. Thenconvergence speed remains unchanged for different computational limits.nThis algorithm is promising for echo cancellation in satellite links andnin data transmission with modems