Detection of narrowband signals in noise using adaptive linear prediction filters

This paper evaluates the detection performance of an application of adaptive linear prediction filtering known as the adaptive line enhancer (ALE) for the signal known except for phase problem. ROC curves are derived for an L-weight ALE followed by a K-point DFT under the conditions that K < L and that K and L are much less than N, the total number of data points processed. ROC curves are also obtained for detection based on an L-point DFT of the L adaptive filter weights. It is shown that both ALE implementations give detection performances roughly comparable to N/L incoherent averages of an L-point DFT, but differences between the two ALE devices are noted. ALE detection performance is shown to be relatively insensitive to the choice of adaptive filter time constant. It is also shown that ALE detection performance is relatively insensitive to the value of K for K < L.