Modulation Consistency-based Contrastive Learning for Self-Supervised Automatic Modulation Classification

2026-05-12

AI Summary

This work addresses the challenges in automatic modulation classification posed by the reliance on large amounts of labeled data and the suboptimal representations learned by existing self-supervised methods due to task-agnostic pretraining. To this end, the paper proposes Mod-CL, a novel framework that, for the first time, incorporates modulation consistency as a task-aware structural prior into self-supervised learning. By constructing positive pairs from different time segments of the same signal and designing a contrastive objective that avoids intra-instance supervision conflicts, the model effectively captures shared modulation features while suppressing irrelevant variations induced by channel effects and noise. Experiments on the RadioML dataset demonstrate that Mod-CL significantly outperforms current baselines, achieving substantial gains in linear probing accuracy—particularly under low-label regimes.

Abstract

Deep learning-based AMC methods have achieved remarkable performance, but their practical deployment remains constrained by the high cost of labeled data. Although self-supervised learning (SSL) reduces the reliance on labels, existing SSL-based AMC methods often rely on task-agnostic pretext objectives misaligned with modulation classification, leading to representations entangled with nuisance factors such as symbol, channel, and noise. In this paper, we identify intra-instance modulation consistency as a task-aware structural prior, whereby different temporal segments of the same signal may differ in waveform while preserving the same modulation type, thus providing a principled cue for task-aligned self-supervision. Based on this prior, we propose Mod-CL, a Modulation consistency-based Contrastive Learning framework that constructs positive pairs from different temporal segments of the same signal instance, to encourage the model to learn shared modulation information while suppressing nuisance variations. We further develop a contrastive objective tailored to Mod-CL, which jointly exploits temporal segmentation and data augmentation to pull together views sharing the same modulation semantics while avoiding supervisory conflicts within each signal instance. Extensive experiments on RadioML datasets show that Mod-CL consistently outperforms strong baselines, especially in low-label regimes, achieving substantial improvements in linear probing accuracy.

Problem

• Automatic Modulation Classification
• Self-Supervised Learning
• Modulation Consistency
• Contrastive Learning
• Nuisance Factors

Innovation

• Modulation Consistency
• Contrastive Learning
• Self-Supervised Learning
• Automatic Modulation Classification
• Task-Aware Representation