A probabilistic fuzzy method for emitter identification based on genetic algorithm

This paper presents a probabilistic fuzzy method for emitter identification (EID) based on the data-driven model. The input attributes of the EID problem include the radio frequency (RF), pulse repetition interval (PRI), pulse width (PW), etc. Given a fuzzy partition of the input attributes, a method for deriving a set of probabilistic fuzzy rules from training data is presented. With the aid of genetic algorithm (GA), the fuzzy partition can be adjusted to achieve high classification accuracy and good interpretability simultaneously. Data-driven candidate of fuzzy partitions of the input space is adopted, which guarantees the interpretability of the resulting rules and enables GA to find good fuzzy partition quickly. The experimental results show high performance of the proposed method.