One-Dimensional Frequency-Domain Features for Aircraft Recognition from Radar Range Profiles

Abstract

To extract effective one-dimensional frequency-domain features from high-resolution radar range profiles, the differential power spectrum (DPS) and the product spectrum, which were originally proposed for the speech signal processing, are introduced to the radar target recognition community. Through differentiating the power spectrum with respect to frequency, we obtained the DPS, which is translation invariant. The DPS can preserve the spectral information contained in the range profiles. The product spectrum is defined as the product of the power spectrum and the group delay function. Thus, it can combine the information contained in the magnitude spectrum and phase spectrum of the range profiles and then carry more details about the shape of the aircrafts. In the classification phase, an optimal choice can be determined by implementing six different training algorithms of multilayered feed-forward neural network. The range profiles were measured by using the two-dimensional backscatters distribution data of four different scaled aircraft models. Simulations were demonstrated to evaluate the classification performance with the DPS and the product spectrum-based features. The simulation results have shown that both DPS and product spectrum-based features are effective for the automatic target recognition (ATR) of aircrafts.