Adaptive classification of radar pulses with improved fuzzy ARTMAP

Radar emitter identification is an indispensable part of electronic intelligence (ELINT). Due to its ability to assign a new class label to unfamiliar classes and continue learning during testing while at the same time holding the information obtained during training, fuzzy ARTMAP is one of the methods that has been considered for this problem up to now. In this paper, fuzzy ARTMAP is improved in order to identify radar emitters directly from radar Pulse Description Words (PDWs). The first improvement is the use of a one-step similarity check mechanism instead of the two-layer similarity check mechanism of conventional fuzzy ARTMAP in order to decrease the complexity. The second one is that vigilance parameter is set according to the current environment during an extra vigilance-validation stage within training. The results prove that fuzzy ARTMAP is improved with the addition of these two improvements in terms of complexity and classification accuracy.