An Efficient and Explainable Multi-Task Learning Framework for Wireless Signal Recognition Tasks

Ruixiang Zhang

doi:10.1109/LWC.2025.3644312

An Efficient and Explainable Multi-Task Learning Framework for Wireless Signal Recognition Tasks

2026-01-01

0

PRE

AI

DOI：10.1109/LWC.2025.3644312

Original

Original request for help

Save

Abstract

En 中文

As two critical tasks in wireless signal recognition, Automatic Modulation Classification (AMC) and Specific Emitter Identification (SEI) are correlated since features of modulations and radio frequency fingerprint are coupled together. Existing methods typically handle these two tasks independently and suffer from lack of interpretability, leading to weak performance especially in low signal-to-noise ratio (SNR) conditions. In this letter, we propose an explainable multi-task learning framework for both AMC and SEI tasks, in order to enhance recognition performance and interpretability. Specifically, we design a multi-task learning module, which exploits the feature correlations between two tasks to achieve performance improvement compared to single-task models. Furthermore, a signal generator and a local linear module are introduced to explain the model’s decision-making process. Experiment results confirm the superiority of our framework over single-task benchmarks in both recognition accuracy and interpretability, particularly in low-SNR conditions.

Keywords:

Multi-task learning

automatic modulation classification

specific emitter identification

low-SNR

An Efficient and Explainable Multi-Task Learning Framework for Wireless Signal Recognition Tasks

Abstract

Journal

Organization