.png){kind=icon}
Return
Radar Phrase Extraction via Structure Informed Conditional Random Fields
.png){kind=icon}
PRE
AI
DOI：10.1109/taes.2026.3722774.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The extraction of radar phrases from multifunction radar (MFR) signals presents a significant challenge in noncooperative radar signal processing. Traditional probabilistic graphical models (PGMs) often lack the necessary structural priors, limiting their accuracy and robustness when processing imperfect radar word sequences. To address this limitation, we propose the structure-informed conditional random field, which integrates radar phrase structure knowledge as prior distributions into a data-driven PGM. To efficiently implement this, we design a neuro-symbolic architecture combining a BiLSTM-GRU neuro-component for feature extraction with a lightweight, training-free symbolic detector for structure identification. Instead of conducting full physical-link simulations, we focus on comprehensive symbolic-level evaluations under diverse imperfect scenarios. Both theoretical analysis and experimental results demonstrate that incorporating structural priors effectively mitigates the impact of sequence imperfections, providing a robust algorithmic reference for noncooperative radar semantic understanding.
Keywords:
Multifunction radar (MFR)
probability graphical model
radar phrase extraction
signal processing
Journal
IF:
5.7
Papers:
780
Citations:
2.4W
Organization
Cited Papers
No cited papers available