Frequency tracking performance of adaptive lattice filters

Adaptive linear prediction filters have been proposed for the detection of narrowband signals in nonstationary noise and for the removal of narrowband interference from spread spectrum communications signals. In the present work, analytical results and computer models for the PARCOR coefficients, signal recovery error, and output misadjustment of the stochastic gradient adaptive lattice filter are obtained for a complex linear chirp FM signal in white Gaussian noise. From these expressions an optimal adaptation rate is obtained as a function of the filter order and the chirp rate. These results indicate that short filter lengths are sufficient for tracking chirp signals.< >