Robustness with respect to the signal-to-noise ratio of MLP-based detectors in Weibull clutter

The Neyman-Pearson detector can be approximated by MultiLayer Perceptrons (MLPs) trained in a supervised way to minimize the Mean Square Error. The detection of a known target in a Weibull-distributed clutter and white Gaussian noise is considered. Because of the difficulty to obtain analytical expressions for the optimum detector under this environment, a suboptimum detector like the Target Sequence Known A Priori (TSKAP) detector is taken as reference. A study of the MLP size shows as a low complexity MLP-based detector trained with the Levenberg-Marquardt algorithm to minimize the MSE is able to obtain good performances. Low performance improvement is achieved for greater sizes than 20 hidden neurons. The MLP-based detector is better than the TSKAP one, even for very low complexity MLPs (6 inputs, 5 hidden neurons and 1 output). Moreover, it is demonstrated empirically that both detectors are robust with respect to changes in the target parameters (signal to noise ratio). So, MLP-based detectors are proposed to detect known targets in Weibull-distributed clutter plus white Gaussian noise.