Detection performance of the adaptive line enhancer (ALE) with post-detection integration

This paper discusses the detection performance of a data adaptive detector based on adaptive linear prediction filtering. The implementation considered utilizes the Widrow-Hoff LMS algorithm to provide continuous data adaptive estimates of the linear prediction filter coefficients. (This implementation is often denoted the Adaptive Line Enhancer (ALE).) Receiver-Operating Characteristic (ROC) curves are derived analytically for two distinct adaptive detector implementations for the case of a sinusoid of known frequency but unknown phase embedded in white Gaussian additive noise. The detection statistics and ROC curves of the ALE are obtained experimentally from extensive Monte Carlo simulations. The effect of post detection integration is evaluated analytically and experimentally. Comparative performance curves for adaptive and non-adaptive detector implementations are presented for this classical detection problem. The dependence on the frequency resolution and time constant of the adaptive filter is also discussed.