Multiplier-free NLMS for adaptive IIR filtering

Using geometric series expansion, a multiplier-free version of the NLMS is proposed for adaptive IIR filtering. Inputs rather than the coefficients are quantized to the next power of 2. Hence, single quantization is done per data sample. A different approach in deriving the NLMS is also presented. Simulation results in system identification illustrate the usefulness of the proposed algorithm, even for long durations (poles near the unit circle) and non-exact modeling