New methods for adaptive nonstationary spectrum estimation for ISAR imaging are proposed. ISAR data will be processed in the range and cross range directions to remove the effects of resonant non-point scattering and target acceleration. The essential idea is to construct a set of orthogonal windows that have little spectral leakage and use each window to create an independent spectral estimate. A data dependant, frequency dependant weighted average of these individual estimates provides the final estimate. The proposed algorithms are robust, computationally efficient, and suited to optical processing. Phase I demonstrated the utility of this approach for ISAR imaging in the nonadaptive stationary case. In Phase II and its Options, adaptive nonstationary spectral estimation algorithms will be tested and refined on air and ground ISAR data supplied by NRL and ONR. An ISAR imaging software toolkit containing conventional spectral estimation algorithms, and the algorithms described here, will be delivered to ONR for evaluation. Although the algorithms will be used to analyze radar returns, they can be used for the detection and analysis of any nonstationary signal in additive or conventional broadband nongaussian noise.
Benefits: An ISAR Imaging Toolkit (ISAR IT) containing spectral estimation used for ISAR imaging will be created for Government use. An analogous toolkit, MSET (Multi-window Spectral Estimation Toolkit) will be created for the private sector. Both toolkits are unique and can be used with the SigniScope, Signition's flagship product for signal and image processing. Signition will also use adaptive nonstationary spectral estimation algorithms to create real-time spectrograms for acoustic signal processing and for speech processing for computer speech recognition systems. Applications in MRI and NMR spectroscopy are expected.
Keywords: ISAR Spectral estimation Multi-window Slepian Digital prolate spheroidal sequences Optical processing Lapped orthogonal transform Radar