Performance analysis of the adaptive line enhancer for sinusoidal signals in broad-band noise

The authors extend earlier the optimum value of the decorrelation parameter Δ in the adaptive line enhancer (ALE) to the case of two sinusoidals. They suggested a technique for estimating this optimum choice of Δ in the finite-data case. They studied the performance of the ALE as a spectral estimator. In particular, they studied the asymptotic bias in the frequency estimates produced by the ALE with the near-optimum and the usual choices of Δ. In the finite-data case, they used computer simulations to study the variance of the frequency estimates in relation to the Cramer-Rao bound. The analysis and the simulations show that the frequency estimates obtained with the near-optimum choice of Δ are much superior with respect to bias, variance, and the threshold SNR than the estimates produced with the usual choice, Δ=1