Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Collaborative Automatic Modulation Classification via Deep Edge Inference for Hierarchical Cognitive Radio Networks

Chaowei He, Peihao Dong, Fuhui Zhou, Qihui Wu

TL;DR

This work targets modulation classification in hierarchical cognitive radio networks by reducing data transmission and privacy risks through edge-assisted learning. It introduces a collaborative AMC (C-AMC) framework that partitions computation: an edge device runs SSCNet to produce a compact spectral semantic embedding, while an edge server runs MCNet (Bi-LSTM with multi-head attention) to classify modulation from the noisy embedding, with transmissions modeled as $\mathbf{y}=\mathbf{x}_s+\mathbf{w}$. Key contributions include the lightweight SSCNet design with a compression rate $r=\frac{2L}{N}$ and the MCNet architecture that robustly extracts features from semantic embeddings, achieving high accuracy on RadioML2016.10A while greatly reducing model size and FLOPs. The framework demonstrates strong anti-noise performance and practical potential for low-latency, privacy-aware cognitive radio systems, enabling scalable edge–server co-inference for AMC. \(\)In particular, with $r=16$, the system attains about $93\%$ accuracy and outperforms baselines, while maintaining substantially lower complexity, making it suitable for real-time wireless monitoring and security applications.

Abstract

In hierarchical cognitive radio networks, edge or cloud servers utilize the data collected by edge devices for modulation classification, which, however, is faced with problems of the transmission overhead, data privacy, and computation load. In this article, an edge learning (EL) based framework jointly mobilizing the edge device and the edge server for intelligent co-inference is proposed to realize the collaborative automatic modulation classification (C-AMC) between them. A spectrum semantic compression neural network (SSCNet) with the lightweight structure is designed for the edge device to compress the collected raw data into a compact semantic message that is then sent to the edge server via the wireless channel. On the edge server side, a modulation classification neural network (MCNet) combining bidirectional long short-term memory (Bi-LSTM) and multi-head attention layers is elaborated to determine the modulation type from the noisy semantic message. By leveraging the computation resources of both the edge device and the edge server, high transmission overhead and risks of data privacy leakage are avoided. The simulation results verify the effectiveness of the proposed C-AMC framework, significantly reducing the model size and computational complexity.

Collaborative Automatic Modulation Classification via Deep Edge Inference for Hierarchical Cognitive Radio Networks

TL;DR

This work targets modulation classification in hierarchical cognitive radio networks by reducing data transmission and privacy risks through edge-assisted learning. It introduces a collaborative AMC (C-AMC) framework that partitions computation: an edge device runs SSCNet to produce a compact spectral semantic embedding, while an edge server runs MCNet (Bi-LSTM with multi-head attention) to classify modulation from the noisy embedding, with transmissions modeled as . Key contributions include the lightweight SSCNet design with a compression rate and the MCNet architecture that robustly extracts features from semantic embeddings, achieving high accuracy on RadioML2016.10A while greatly reducing model size and FLOPs. The framework demonstrates strong anti-noise performance and practical potential for low-latency, privacy-aware cognitive radio systems, enabling scalable edge–server co-inference for AMC. \(\)In particular, with , the system attains about accuracy and outperforms baselines, while maintaining substantially lower complexity, making it suitable for real-time wireless monitoring and security applications.

Abstract

In hierarchical cognitive radio networks, edge or cloud servers utilize the data collected by edge devices for modulation classification, which, however, is faced with problems of the transmission overhead, data privacy, and computation load. In this article, an edge learning (EL) based framework jointly mobilizing the edge device and the edge server for intelligent co-inference is proposed to realize the collaborative automatic modulation classification (C-AMC) between them. A spectrum semantic compression neural network (SSCNet) with the lightweight structure is designed for the edge device to compress the collected raw data into a compact semantic message that is then sent to the edge server via the wireless channel. On the edge server side, a modulation classification neural network (MCNet) combining bidirectional long short-term memory (Bi-LSTM) and multi-head attention layers is elaborated to determine the modulation type from the noisy semantic message. By leveraging the computation resources of both the edge device and the edge server, high transmission overhead and risks of data privacy leakage are avoided. The simulation results verify the effectiveness of the proposed C-AMC framework, significantly reducing the model size and computational complexity.
Paper Structure (11 sections, 11 equations, 7 figures, 1 table)

This paper contains 11 sections, 11 equations, 7 figures, 1 table.

Table of Contents

  1. Introduction
  2. System Model
  3. EL-Based C-AMC Framework
  4. Overview of Framework
  5. SSCNet
  6. MCNet
  7. Simulation Results
  8. Dataset
  9. Simulation Settings
  10. The Performance of the C-AMC
  11. Conclusion

Figures (7)

  • Figure 1: Hierarchical CR network model.
  • Figure 2: EL-based C-AMC framework.
  • Figure 3: Multi-Head Attention.
  • Figure 4: Performance of C-AMC,LSTMNet-DC,SSCNet-DC and SplitAMC on RadioML2016.10A dataset.
  • Figure 5: Performance of C-AMC on RadioML2016.10A dataset under different compression rate $r$.
  • ...and 2 more figures