ALE Gain performance for narrowband signals in white Gaussian noise

The ALE is an adaptive digital filter whose performance in separating narrowband signals from broadband noise has been of considerable interest in recent years. This paper offers analytical evaluation of output statistics given a class of inputs consisting of finite bandwidth signals in white noise. Unlike analogous results for random-phase sinusoids, SNR gain curves for signals with damped sinusoidal autocorrelations demonstrate two important considerations in choosing ALE parameters: 1) bulk delay should be minimized in order to prevent significant signal decorrelation; and 2) filter length should be sufficiently short to avoid unnecessary misadjustment error from unused weights.