A prediction-correlation target discrimination algorithm using target complex natural resonances is investigated. The targets natural resonances are extracted from experimentally measured frequency domain scattering data. Scattering data were obtained from scale model targets measured in the ElectroScience Laboratory compact radar range facility. The targets used are models of three military ground vehicles. Signal processing required for the extraction of poles is described. The principles of the prediction-correlation algorithm are discussed and the algorithm is demonstrated for both noiseless data and data with additive white noise. Classification results are plotted for the targets at aspect angles of 0, 10, 20, 30, 60, and 90 degrees from nose on.