Title: A scalable multimodal ensemble learning framework for automatic modulation recognition
Authors: Jian Shi; Guangxue Yue; Shengyu Ma; Tianjun Peng; Bolin Ma
Addresses: College of Information Science and Engineering, Jiaxing University, Jiaxing, Zhejiang, China; Key Laboratory of Medical Electronics and Digital Health of Zhejiang Province, Jiaxing University, Jiaxing, Zhejiang, China ' College of Information Science and Engineering, Jiaxing University, Jiaxing, Zhejiang, China; Key Laboratory of Medical Electronics and Digital Health of Zhejiang Province, Jiaxing University, Jiaxing, Zhejiang, China ' College of Information Science and Engineering, Jiaxing University, Jiaxing, Zhejiang, China; Key Laboratory of Medical Electronics and Digital Health of Zhejiang Province, Jiaxing University, Jiaxing, Zhejiang, China ' College of Information Science and Engineering, Jiaxing University, Jiaxing, Zhejiang, China; Key Laboratory of Medical Electronics and Digital Health of Zhejiang Province, Jiaxing University, Jiaxing, Zhejiang, China ' College of Data Science, Jiaxing University, Jiaxing, Zhejiang, China
Abstract: The Automatic Modulation Recognition (AMR) method based on Deep Learning (DL) has achieved excellent performance and gradually become a hot spot of research. Most researches have designed complex structures or supplemented feature information to achieve the recognition of modulation signals, which cannot fully combine the advantages of different models to extract features, resulting in poor recognition accuracy of modulated signals. To solve the problem, we propose a Scalable Multimodal Ensemble Learning Framework (SMELF), which trains various models with multimodal information including In-phase Quadrature (I/Q) and Amplitude Phase (A/P) information to supplement feature information. The meta-model is used as a combined strategy to correlate the feature extraction advantages of each model. The simulation results show that SMELF not only achieves superior classification accuracy, but also is the most robust under different Signal-to-Noise Ratios (SNRs) environments and the training sample sizes. In addition, our method can further improve the classification accuracy by combining more diverse and better performance models, which reflects the great potential of the framework.
Keywords: automatic modulation recognition; multimodal information; ensemble learning; vision transformer.
DOI: 10.1504/IJWMC.2024.137175
International Journal of Wireless and Mobile Computing, 2024 Vol.26 No.2, pp.182 - 197
Received: 05 May 2023
Accepted: 02 Sep 2023
Published online: 04 Mar 2024 *