{"title":"一种基于变压器的复值卷积和增强Bi-LSTM的调制自动识别框架","authors":"Shenping Wu, Chao Wang, Jiakai Liang, Mayue Wang, Keqiang Yue, Wenjun Li","doi":"10.1016/j.phycom.2025.102824","DOIUrl":null,"url":null,"abstract":"<div><div>Automatic Modulation Classification (AMC) is critical for modern wireless communication systems. Despite significant progress in deep learning-based AMC, existing methods still struggle to jointly capture local features, model temporal dependencies, and extract global representations. We propose CBADNN, an end-to-end architecture that combines Transformer self-attention for global context modeling, complex-valued convolutions for local spatial feature extraction, and bidirectional stacked LSTMs (Bi-sLSTMs) for temporal dependency learning. CBADNN achieves state-of-the-art performance, with overall accuracies of 64.02% and 65.50% on the widely used RadioML 2016.10a and RadioML 2016.10b datasets, respectively. On RadioML 2016.10a, it outperforms the best baseline by 0.81% under high SNR (4 dB–18 dB) and 0.33% under medium SNR (−8 dB–2 dB), consistently demonstrating superiority across diverse SNR conditions.Furthermore, a detailed evaluation of classification accuracy across various modulation types under different SNR conditions is conducted, and ablation experiments on the RadioML2016.10a dataset are performed to rigorously validate the effectiveness of each module in the proposed network.</div></div>","PeriodicalId":48707,"journal":{"name":"Physical Communication","volume":"73 ","pages":"Article 102824"},"PeriodicalIF":2.2000,"publicationDate":"2025-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"A transformer-based framework with complex-valued convolution and enhanced Bi-LSTM for automatic modulation recognition\",\"authors\":\"Shenping Wu, Chao Wang, Jiakai Liang, Mayue Wang, Keqiang Yue, Wenjun Li\",\"doi\":\"10.1016/j.phycom.2025.102824\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><div>Automatic Modulation Classification (AMC) is critical for modern wireless communication systems. Despite significant progress in deep learning-based AMC, existing methods still struggle to jointly capture local features, model temporal dependencies, and extract global representations. We propose CBADNN, an end-to-end architecture that combines Transformer self-attention for global context modeling, complex-valued convolutions for local spatial feature extraction, and bidirectional stacked LSTMs (Bi-sLSTMs) for temporal dependency learning. CBADNN achieves state-of-the-art performance, with overall accuracies of 64.02% and 65.50% on the widely used RadioML 2016.10a and RadioML 2016.10b datasets, respectively. On RadioML 2016.10a, it outperforms the best baseline by 0.81% under high SNR (4 dB–18 dB) and 0.33% under medium SNR (−8 dB–2 dB), consistently demonstrating superiority across diverse SNR conditions.Furthermore, a detailed evaluation of classification accuracy across various modulation types under different SNR conditions is conducted, and ablation experiments on the RadioML2016.10a dataset are performed to rigorously validate the effectiveness of each module in the proposed network.</div></div>\",\"PeriodicalId\":48707,\"journal\":{\"name\":\"Physical Communication\",\"volume\":\"73 \",\"pages\":\"Article 102824\"},\"PeriodicalIF\":2.2000,\"publicationDate\":\"2025-09-02\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Physical Communication\",\"FirstCategoryId\":\"94\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S1874490725002277\",\"RegionNum\":4,\"RegionCategory\":\"计算机科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"ENGINEERING, ELECTRICAL & ELECTRONIC\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Physical Communication","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1874490725002277","RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"ENGINEERING, ELECTRICAL & ELECTRONIC","Score":null,"Total":0}
引用次数: 0
摘要
自动调制分类是现代无线通信系统的重要组成部分。尽管基于深度学习的AMC取得了重大进展,但现有的方法仍然难以共同捕获局部特征、建模时间依赖性和提取全局表示。我们提出了CBADNN,这是一种端到端架构,结合了用于全局上下文建模的Transformer自关注,用于局部空间特征提取的复杂值卷积,以及用于时间依赖性学习的双向堆叠lstm (bi - slstm)。CBADNN达到了最先进的性能,在广泛使用的RadioML 2016.10a和RadioML 2016.10b数据集上,总体准确率分别为64.02%和65.50%。在RadioML 2016.10a上,它在高信噪比(4 dB - 18 dB)和中信噪比(- 8 dB - 2 dB)下的性能分别优于最佳基线0.81%和0.33%,在不同信噪比条件下始终表现出优势。此外,对不同信噪比条件下不同调制类型的分类精度进行了详细评估,并在RadioML2016.10a数据集上进行了烧蚀实验,严格验证了所提网络中每个模块的有效性。
A transformer-based framework with complex-valued convolution and enhanced Bi-LSTM for automatic modulation recognition
Automatic Modulation Classification (AMC) is critical for modern wireless communication systems. Despite significant progress in deep learning-based AMC, existing methods still struggle to jointly capture local features, model temporal dependencies, and extract global representations. We propose CBADNN, an end-to-end architecture that combines Transformer self-attention for global context modeling, complex-valued convolutions for local spatial feature extraction, and bidirectional stacked LSTMs (Bi-sLSTMs) for temporal dependency learning. CBADNN achieves state-of-the-art performance, with overall accuracies of 64.02% and 65.50% on the widely used RadioML 2016.10a and RadioML 2016.10b datasets, respectively. On RadioML 2016.10a, it outperforms the best baseline by 0.81% under high SNR (4 dB–18 dB) and 0.33% under medium SNR (−8 dB–2 dB), consistently demonstrating superiority across diverse SNR conditions.Furthermore, a detailed evaluation of classification accuracy across various modulation types under different SNR conditions is conducted, and ablation experiments on the RadioML2016.10a dataset are performed to rigorously validate the effectiveness of each module in the proposed network.
期刊介绍:
PHYCOM: Physical Communication is an international and archival journal providing complete coverage of all topics of interest to those involved in all aspects of physical layer communications. Theoretical research contributions presenting new techniques, concepts or analyses, applied contributions reporting on experiences and experiments, and tutorials are published.
Topics of interest include but are not limited to:
Physical layer issues of Wireless Local Area Networks, WiMAX, Wireless Mesh Networks, Sensor and Ad Hoc Networks, PCS Systems; Radio access protocols and algorithms for the physical layer; Spread Spectrum Communications; Channel Modeling; Detection and Estimation; Modulation and Coding; Multiplexing and Carrier Techniques; Broadband Wireless Communications; Wireless Personal Communications; Multi-user Detection; Signal Separation and Interference rejection: Multimedia Communications over Wireless; DSP Applications to Wireless Systems; Experimental and Prototype Results; Multiple Access Techniques; Space-time Processing; Synchronization Techniques; Error Control Techniques; Cryptography; Software Radios; Tracking; Resource Allocation and Inference Management; Multi-rate and Multi-carrier Communications; Cross layer Design and Optimization; Propagation and Channel Characterization; OFDM Systems; MIMO Systems; Ultra-Wideband Communications; Cognitive Radio System Architectures; Platforms and Hardware Implementations for the Support of Cognitive, Radio Systems; Cognitive Radio Resource Management and Dynamic Spectrum Sharing.