Sub-Classification of Radar Modulation via Subspace Clustering
Shuai Xu, Lutao Liu · 2024
Existing work on radar modulation classification tends to address only the type of modulation, which is insufficient to meet the needs of war radars. The unbalanced distribution of modulation types of war signals weakens the usefulness of type identification, mainly since war radars are monopolized by a small number of dominant modulations, and traditional interclass label is losing their ability to distinguish between targets. In this paper, we present the concept of radar modulation type sub-classification. Precisely, we classify multiple subtypes for each modulation type according to mode and then cluster the time-frequency representations (TFRs) of these subtypes using a subspace clustering method. Experimental results confirm that the unsupervised classification method represented by the deep subspace clustering network can realize subclassification and achieves 99% subclassification accuracy at more than 8 dB.