| Title: | Radio signals recognition with unsupervised deep learning : a survey |
|---|
| Authors: | ID Milosheski, Ljupcho, Institut "Jožef Stefan" (Author)
ID Bertalanič, Blaž, Institut "Jožef Stefan" (Author)
ID Fortuna, Carolina, Institut "Jožef Stefan" (Author)
ID Mohorčič, Mihael, Institut "Jožef Stefan" (Author) |
|---|
| Files: |  URL - Source URL, visit https://ieeexplore.ieee.org/document/11311988
 PDF - Presentation file, download (3,84 MB)
MD5: A1AC2C36A0A3C52CC9D42472D35093A7
|
|---|
| Language: | English |
|---|
| Typology: | 1.02 - Review Article |
|---|
| Organization: |  IJS - Jožef Stefan Institute
|
|---|
| Abstract: | Optimization of wireless network parameters relies on the awareness of a dynamically changing radio environment, which depends on the presence of active devices characterized by various radio access technologies (RATs), modulation schemes, and overall spectrum usage patterns, and can be determined by advanced radio signal recognition methods. While various supervised machine learning (ML) models have been explored for signal recognition, their actual deployment has been limited so far due to challenges in acquiring labeled datasets. The emergence of Open Radio Access Network (O-RAN) architectures and open experimental testbed setups has enabled access to large-scale, unlabeled data through standardized interfaces, paving the way for unsupervised deep learning methods. These methods, unlike supervised approaches, require minimal labeled data and have shown promising results in domains such as computer vision and time-series processing. However, their application in wireless communications remains relatively unexplored. This survey aims to provide a comprehensive overview of unsupervised deep learning techniques for addressing key challenges for signal recognition in wireless communications, including automatic modulation classification (AMC), signal sensing, specific emitter identification (SEI), and anomaly detection. Specifically, we examine state-of-the-art approaches such as deep clustering, contrastive learning, autoencoder-based reconstruction, and generative models. Additionally, we discuss available open datasets and identify research opportunities to advance this field, leveraging the substantial successes of self-supervised learning in computer vision and natural language processing. By organizing the survey into two key complementary perspectives—wireless communication challenges and unsupervised deep learning solutions—this work provides a roadmap for researchers and practitioners seeking to develop innovative, data-efficient models for the next generation of AI-native wireless networks. |
|---|
| Keywords: | anomaly detection, automatic modulation classification, radio signal recognition, signal sensing |
|---|
| Publication status: | Published |
|---|
| Publication version: | Version of Record |
|---|
| Submitted for review: | 13.11.2025 |
|---|
| Article acceptance date: | 15.12.2025 |
|---|
| Publication date: | 31.12.2025 |
|---|
| Publisher: | IEEE |
|---|
| Year of publishing: | 2025 |
|---|
| Number of pages: | str. 217769-217798 |
|---|
| Numbering: | Vol. 13 |
|---|
| Source: | ZDA |
|---|
| PID: | 20.500.12556/DiRROS-24984  |
|---|
| UDC: | 004.8 |
|---|
| ISSN on article: | 2169-3536 |
|---|
| DOI: | 10.1109/ACCESS.2025.3647686 |
|---|
| COBISS.SI-ID: | 263441667 |
|---|
| Copyright: | © 2025 The Authors. |
|---|
| Note: | Nasl. z nasl. zaslona;
Soavtorji: Blaž Bertalanič, Carolina Fortuna, Mihael Mohorčič;
Opis vira z dne 5. 1. 2026;
|
|---|
| Publication date in DiRROS: | 07.01.2026 |
|---|
| Views: | 65 |
|---|
| Downloads: | 34 |
|---|
| Metadata: |    |
|---|
|
:
|
Copy citation |
|---|
| | | | Share: |  |
|---|
Hover the mouse pointer over a document title to show the abstract or click
on the title to get all document metadata. |
