Electromagnetic Modulation Signal Classification Using Dual-Modal Feature Fusion CNN.

Area of Science:

• Electrical Engineering
• Computer Science
• Signal Processing

Background:

• Automatic Modulation Classification (AMC) is crucial for spectrum monitoring and detecting electromagnetic anomalies.
• Existing AMC methods often overlook the benefits of combining features from different data types and effective fusion strategies.

Purpose of the Study:

• To propose a novel Dual-Modal Feature Fusion Convolutional Neural Network (DMFF-CNN) for enhanced AMC.
• To leverage the complementary nature of different modal features for improved classification accuracy.

Main Methods:

• The proposed DMFF-CNN utilizes Gram Angular Field (GAF) image coding and In-phase/Quadrature (IQ) data.
• GAF images are processed by ResNet50, while IQ data is fed into a Complex Value Convolutional Neural Network (CV-CNN).
• A Dual-Modal Feature Fusion (DMFF) mechanism integrates features from both pathways for final classification.

Main Results:

• The DMFF-CNN demonstrated superior performance compared to existing methods, including state-of-the-art approaches.
• The method exhibits excellent robustness, particularly at low signal-to-noise ratios (SNRs).
• An average classification accuracy of 92.1% was achieved on the dataset signals.

Conclusions:

• The proposed DMFF mechanism effectively utilizes feature complementarity for AMC.
• DMFF-CNN offers a promising new direction for advancing automatic modulation classification techniques.