Advanced Signal Identification

Two types of neural network architectures are proposed for the integration of the acousto-optical (AO) system with the standard electronic warfare (EW) discriminants, rather than just one, as suggested by the topic narrative. A hierarchical neural network is proposed for use in the first stage of clustering. This network performs per-pulse classification into attribute clusters, using the EW receiver standard outputs and the AO system frequency bin outputs from the photodetectors. It is further proposed that the hierarchical neural network be enhanced with another ORINCON innovation in which novelty detection is used to identify never-seen-before classes. Multiple pulses form attribute clusters out of the clustering hierarchical neural net, which we propose to feed to second ORINCON-developed net called a fuzzy min-max neural network, which performs the final emitter identification and classification. This proposal also introduces a new concept for 2-D AO spectral analysis in which pulse frequency and PRF are determined simultaneously. Such a new AO device would replace the standard discriminants and could work in conjunction with the existing 1-D AO spectral analyzer.