On signal detection using support vector machines

Abstract

The detection type problems represent a special case of nonlinear mapping. This fact makes the use of neural networks attractive for signal detection problems. In order to obtain good generalization excessive tuning is needed. Also, most of the neural network learning theories does not make use of the optimal hyperplane concept. In this paper, we consider optimal hyperplane signal detection with support vector machines (SVMs), for detecting a known signal corrupted by noise. Experimental results illustrate the detection performances in various cases. The practical implementation and the robustness of SVMs are also considered.