.png){kind=icon}
.png){kind=icon}
Return
MM-Net: A Multi-Modal Approach Toward Automatic Modulation Classification
.png){kind=icon}
PRE
AI
DOI：10.1109/LCOMM.2023.3342604.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Automatic Modulation Classification (AMC) has become an important component in communication systems for both civil and defense applications. The shortcomings of traditional approaches to AMC have led researchers to develop complex machine learning (ML)-based approaches. In this work, inspired by multi-modal approaches for general Computer Vision tasks like Semantic Segmentation, we propose MM-Net, a multimodal approach to AMC that uses domain-specific features in the form of Higher Order Cumulants (HOCs) to improve classification performance. Furthermore, we explore the usage of HOCs in existing Deep Learning (DL)-based applications for AMC. Simulation results show that for eight modulation classification, MM-Net achieves high classification accuracy even at low SNRs, demonstrating the robustness of the multimodal approach even under challenging channel conditions, while existing methods are improved by utilizing HOCs, especially at low SNR values.
Keywords:
Feature extraction
Modulation
Task analysis
Signal to noise ratio
Convolutional neural networks
Computer architecture
Robustness
Automatic modulation classification
machine learning
deep learning
cumulants
convolutional neural networks
transfer learning
Journal
IF:
4.4
Papers:
1.3W
Citations:
2.2W