Spectrum Estimation Using an Analytic Signal Representation

High resolution spectrum estimation algorithms traditionally are constrained to process a finite amount of data, assuming the data to be zero outside the analysis interval. These assumptions ultimately limit the resolution that can be achieved by these estimators. Analytic signals provide an alternate signal representation whereby long-term phase information can be incorporated into the analysis data. Analytic signal-based estimators are shown to achieve higher resolution than their real signal counterparts, due to the phase-invariance property of an analytic signal. The linear predictive estimates of stationary signals in additive white Gaussian noise, using a complex linear predictor, are shown to be more consistent than those obtained using a comparable real linear predictor.