Improved noise canceller performance by means of an adaptive arrangement of IIR and FIR filters

It is shown that when the primary input of an LMS (least mean square) adaptive noise canceler contains interference derived from a white source by means of an autoregressive process, a hybrid arrangement of adaptive IIR (infinite impulse response) and FIR (finite impulse response) filters can achieve improved performance. Convergence and stability problems are avoided by using a conventional LMS FIR adaptive filter to identify the interference while an auxiliary IIR filter cancels the residual components for which the FIR filter is insufficient. Stability is analyzed and simulation results are presented. The stability of the IIR filter is shown to depend on the LMS step size, the variance of the wanted signal, and the interference due to the uncancelled tail. >