NN-Based detector for known targets in coherent weibull clutter
IDEAL'06 Proceedings of the 7th international conference on Intelligent Data Engineering and Automated Learning
| Hi-index | 0.00 |
Obtaining analytical expressions for coherent detection of known signals in Weibul-distributed clutter and white Gaussian noise has been a hard task since the last decades. In fact, nowadays, these expressions have not been found yet. This problem lead us to use suboptimum solutions to solve this problem. Optimum approximations can be done by using Multilayer Perceptrons (MLPs) trained in a supervised way to minimize the mean square error. So, MLP-based detectors are constructed and compared with one of the suboptimum detectors commonly used to solve the detection problem under study. First, a study of the dimensionality of the MLP is done for typical values of the target and clutter conditions. And finally, a deep study is done according to the variations of the most important parameters of the target and clutter signals. The last study let us to be conscious about the importance of the selection of the parameters to design both detectors. Moreover, the difference of performances between each other and the superiority of the MLP-based detector against the suboptimum solution is emphasized.