FOURTH ANNUAL
ASAP '96 WORKSHOP


Detection Performance with Combined Adaptive Filters

Steven T. Smith
MIT Lincoln Laboratory
244 Wood Street
Lexington, MA 02173-9108
tel: (617) 981-3106
email: stsmith@ll.mit.edu

Abstract Adaptive filtering is used by radar systems to detect targets in the presence of jamming and clutter interference whose precise signal characteristics are unknown a priori. The adaptive radar's detection performance depends upon the number of samples used to estimate the interference, as well as factors associated with nonadaptive detectors such as the target's signal strength, steering vector mismatch, and the methods of combining multiple pulses or coherent pulse intervals, clustering, and tracking. The radar's ultimate detection performance can be predicted by making assumptions about the statistics of the filter output. Whereas the Gaussian assumption is preponderant for nonadaptive radars, this assumption certainly does not hold for adaptive radars whose use of data to construct adaptive filters significantly alters the signal's statistics. The complicated statistics of adaptive filter outputs frustrates a closed-form analysis of their combination using all but the simplest integration methods. Therefore, approximation methods are appropriate for this problem. This presentation addresses the problem of determining detection performance when the outputs of several adaptive filters are combined. New results showing the moments of the adaptive matched filter (AMF) and generalized likelihood ratio test (GLRT) are given, which contain the effects of finite sample support, steering vector mismatch, and fluctuating and nonfluctuating targets. These results are used to analyze the detection performance of certain space-time adaptive processing (STAP) algorithms utilizing a variety of post-STAP detection schemes. For example, the performance of post-STAP binary integration (m-out-of-n detector) is compared to the performance of post-STAP square-law integration. All methods are compared to the performance achieved using a nonadaptive matched filter.


 


LL Logo Disclaimer

Direct comments and questions to: webmaster@ll.mit.edu

MIT Lincoln Laboratory. All rights reserved.