Partial Label Learning of RF Emitters with LSTMs
Richard H. Moseley · AIAA Scitech 2020 Forum · 2020
As modern military radars are becoming more agile, Radio Frequency (RF) is becoming less of a discriminator for identification. Along with RF agility, radars that are low probability of intercept (LPI) make consistent detection and measurement of discriminating modulation features more difficult. In light of these challenges, a class of Recurrent Neural Networks (RNNs) called Long Short Term Memory (LSTM) networks will be demonstrated on a real dataset to exploit the temporal features of measured RF from two ambiguous agile emitters and classify with an accuracy of 92.5%. Future areas of research and application on this topic will be discussed as well.