Classification and novelty detection using linear models and a class dependent-elliptical basis function neural network

Neural networks are ideally suited for solving detection and classification problems. In the real world however often the number and labels of classes may be unknown. This is particularly true for the detection and classification of collected electromagnetic (EM) signals when new transmitter types are first turned on. Approaches that determine when new, never-seen-before, novel events are present in the system need to be considered. Described here is the development and application of a class-dependent elliptical basis function (CD-EBF) neural net developed to solve novelty detection problems. The neural net uses parameters derived from a linear model fit to the data as input. Application of the system to passive detection and classification of EM signals is shown.