Inferring radar mode changes from elementary pulse features using Fuzzy ARTMAP classification

Abstract

A method for radar mode inference using Fuzzy ARTMAP classification is presented. In this method elementary radar parameters, Pulse Width (PW) and Pulse Repetition Interval (PRI) originating from a radar operating in a certain mode is input to a Fuzzy ARTMAP classifier. Radar parameters were simulated at different signal-to-noise ratios (SNRs) to train and evaluate the Fuzzy ARTMAP classifier without prior knowledge of radar operating modes. Thus Fuzzy ARTMAP classification is used in the analysis of radar mode behaviour. Training resulted in map field weights with high code compression and broad generalization of the input space. The choice of ARTa categories accurately correlated with the current radar mode input data presented to the classifier. It resulted in a 1.8% error in category choice (radar mode) at worst. Classifier training may be done on data with low SNR as the broad generalization during training will accommodate high SNR data without compromising accuracy during evaluation. Knowledge about the amount of radar modes and mode transition can also be gained by an initial training and evaluation (analysis) process to assign pseudo modes to a particular radar. The resultant modes can then be included into a Fuzzy ARTMAP classifier by increasing the dimension of the predicted output, B to classify both radar class and operating mode.