DHCT: A Distributed Hierarchical Network for Multifunction Radar Work Mode Recognition

Multi-Function Radar work mode recognition is crucial for electronic warfare but challenged by signal agility, in complete observations, and complex electromagnetic environments. To address these challenges, this paper proposes a Distributed Hierarchical Conformer-Transformer (DHCT) model within a ro bust distributed architecture. DHCT operates through a two-stage process in which independent Conformer encoders initially extract robust temporal features per receiver, followed by a Transformer module that fuses these features using cross-receiver attention to capture inter-receiver relationships. Its attention pooling mechanism adaptively weights fused features, outperforming static pooling, while auxiliary and contrastive supervision during training further enhance discriminative power and overall robustness. Experimen tal results demonstrate DHCT's superior accuracy and resilience over single-station methods and alternative fusion strategies under various challenging non-ideal conditions, particularly against high spurious pulse rates and in challenging mixed scenarios where conventional methods show limitations. A notable secondary contribution is the demonstration that the core architecture achieves high efficacy even in a single-receiver configuration.