ISAR feature extraction from targets with non-rigid body motion using adaptive chirplet representation

A basic assumption in inverse synthetic aperture radar (ISAR) imaging is that the target is a rigid body during the data acquisition time. However, non-rigid body motions (associated with a target that is not a rigid body) can exist in many real targets. A chirplet-based adaptive signal representation is applied to the feature extraction of radar data from targets with non-rigid body motion. In particular, a four-parameter chirplet basis is used to account for the joint time-frequency localization of the signal after we formulated a time-varying model for the radar signal. We also applied the method to two real-world data sets. In the aircraft data, it is shown that the aircraft body image can be better recovered after the removal of the return from the spinning jet engine blades. In the walking person data, we can more accurately estimate the rate of the swinging arms by autocorrelation analysis of the arm-only data. Our results demonstrate the potential application of the proposed method in radar target identification.