Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multi-task Learning for Radar Signal Characterisation

Zi Huang, Akila Pemasiri, Simon Denman, Clinton Fookes, Terrence Martin

TL;DR

This work addresses radar signal characterisation by formulating it as a multi-task learning problem that jointly classifies radar signal types and estimates pulse descriptors. It introduces the IQ Signal Transformer (IQST), a transformer-based backbone that operates directly on raw IQ data, and presents the RadChar synthetic dataset as a public benchmark for RSC. The authors implement a hard parameter sharing MTL framework with five task heads and a joint loss $L_{ ext{mtl}}$, combining classification and regression objectives. Experimental results show IQST-based MTL outperforms CNN baselines, particularly at low SNR, and ablations reveal the importance of transformer capacity and balanced task weighting for robust RSC performance, enabling practical utility in ISAC and electronic warfare contexts.

Abstract

Radio signal recognition is a crucial task in both civilian and military applications, as accurate and timely identification of unknown signals is an essential part of spectrum management and electronic warfare. The majority of research in this field has focused on applying deep learning for modulation classification, leaving the task of signal characterisation as an understudied area. This paper addresses this gap by presenting an approach for tackling radar signal classification and characterisation as a multi-task learning (MTL) problem. We propose the IQ Signal Transformer (IQST) among several reference architectures that allow for simultaneous optimisation of multiple regression and classification tasks. We demonstrate the performance of our proposed MTL model on a synthetic radar dataset, while also providing a first-of-its-kind benchmark for radar signal characterisation.

Multi-task Learning for Radar Signal Characterisation

TL;DR

This work addresses radar signal characterisation by formulating it as a multi-task learning problem that jointly classifies radar signal types and estimates pulse descriptors. It introduces the IQ Signal Transformer (IQST), a transformer-based backbone that operates directly on raw IQ data, and presents the RadChar synthetic dataset as a public benchmark for RSC. The authors implement a hard parameter sharing MTL framework with five task heads and a joint loss , combining classification and regression objectives. Experimental results show IQST-based MTL outperforms CNN baselines, particularly at low SNR, and ablations reveal the importance of transformer capacity and balanced task weighting for robust RSC performance, enabling practical utility in ISAC and electronic warfare contexts.

Abstract

Radio signal recognition is a crucial task in both civilian and military applications, as accurate and timely identification of unknown signals is an essential part of spectrum management and electronic warfare. The majority of research in this field has focused on applying deep learning for modulation classification, leaving the task of signal characterisation as an understudied area. This paper addresses this gap by presenting an approach for tackling radar signal classification and characterisation as a multi-task learning (MTL) problem. We propose the IQ Signal Transformer (IQST) among several reference architectures that allow for simultaneous optimisation of multiple regression and classification tasks. We demonstrate the performance of our proposed MTL model on a synthetic radar dataset, while also providing a first-of-its-kind benchmark for radar signal characterisation.
Paper Structure (11 sections, 3 equations, 3 figures, 1 table)

This paper contains 11 sections, 3 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Proposed Method
  3. Dataset Generation
  4. Multi-task Learning Framework
  5. Shared Feature Extraction Backbones
  6. Experiments
  7. Training Details
  8. MTL Model Performance
  9. Ablation Study
  10. Conclusion
  11. Acknowledgement

Figures (3)

  • Figure 1: Radar signals sampled from the RadChar dataset illustrating polyphase Barker codes at varying SNRs.
  • Figure 2: The proposed hard parameter shared MTL architecture for RSC. This model shows an IQST backbone with task-specific classification and regression heads.
  • Figure 3: Test performance of MTL models across an SNR range of -20 to 20 dB. MAE results of the regression tasks are shown in (a) to (d), and signal type classification accuracy is shown in (e).