Research Article
Automatic Modulation Classification with Multi-domain Feature Fusion
@INPROCEEDINGS{10.1007/978-3-031-71716-1_2, author={Guangyang Li and Xiaofeng Wang and Mengting Jiang and Yun Chen and Hengliang Liu and Daying Quan}, title={Automatic Modulation Classification with Multi-domain Feature Fusion}, proceedings={Machine Learning and Intelligent Communication. 8th EAI International Conference, MLICOM 2023, Beijing, China, December 17, 2023, Proceedings}, proceedings_a={MLICOM}, year={2024}, month={9}, keywords={Automatic modulation classification (AMC) deep residual network AlexNet dual channel model}, doi={10.1007/978-3-031-71716-1_2} }- Guangyang Li
Xiaofeng Wang
Mengting Jiang
Yun Chen
Hengliang Liu
Daying Quan
Year: 2024
Automatic Modulation Classification with Multi-domain Feature Fusion
MLICOM
Springer
DOI: 10.1007/978-3-031-71716-1_2
Abstract
Automatic Modulation Classification (AMC) is a crucial technology that empowers communication systems to adapt to different signal environments, thus ensuring efficient signal detection and demodulation in a wide range of civil and military applications. The use of Convolutional Neural Networks (CNNs), with the emergence of deep learning techniques, has become prevalent in AMC. However, the signal features that are extracted from a single domain by the CNNs may be missing some of the essential characteristics of the raw signal. In this paper, a dual-channel model combining a deep residual network and an improved AlexNet has been proposed for AMC. The model makes use of features of the radar signals from both the time domain and the time-frequency domain to improve the detection performance. In the experiment, the model achieved a recognition accuracy of 91.70% on the RadioML 2016.10A dataset when the signal-to-noise ratio (SNR) was 2 dB, demonstrating its effectiveness in AMC. Experimental results demonstrate our proposed model outperforms the state-of-the-art methods.
