26 August 1998 Weak transient signal detection in non-Gaussian noise using RBF and recurrent networks
Author Affiliations +
In most of the developed transient signal detection algorithms, the background noise is usually assumed to be Gaussianly distributed to simplify the derivation of generalized likelihood ratio test (GLRT). However, in many real-world applications like target detection in ultra wideband SAR images, the distribution of dominant interference usually have long tails, and can not be characterized simply by Gaussian distribution. The performance of the linear GLRT detector, which is optimal under the Gaussian background noise assumption, would be badly degraded.In this paper, we take locally optimum approach to develop weak transient signal detection utilizing Laguerre recurrent networks for subspace projection, as well as radial basis function networks for tracking non-Gaussian noise statistics. Experiments show that the proposed weak transient signal detectors perform better than GLRT in the impulsive background noise environment.
© (1998) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
LiKang Yen, LiKang Yen, Jose C. Principe, Jose C. Principe, } "Weak transient signal detection in non-Gaussian noise using RBF and recurrent networks", Proc. SPIE 3395, Radar Sensor Technology III, (26 August 1998); doi: 10.1117/12.319452; https://doi.org/10.1117/12.319452

Back to Top