A Novel Approach for Accurate Signal Classification and Robust Performance Using an Enhanced Modulation Identification in MIMO Systems via Diffusion Kernel Attention Network With Green Anaconda Optimization

ABSTRACT Accurate signal classification and modulation identification are pivotal challenges in multiple‐input multiple‐output (MIMO) systems, especially under dynamic and noisy conditions. This paper proposes a novel framework—diffusion kernel attention graph convolutional neural network (DKAGCNN)—...

Celý popis

Uloženo v:

Podrobná bibliografie
Vydáno v:	International journal of communication systems Ročník 38; číslo 12
Hlavní autoři:	Balaram, Allam, Victoria Jancee, B., Roy, Manidipa, Kapila, Dhiraj
Médium:	Journal Article
Jazyk:	angličtina
Vydáno:	Chichester Wiley Subscription Services, Inc 01.08.2025
Témata:	Artificial neural networks Communications systems Complexity diffusion kernel attention network Graph neural networks green anaconda optimization MIMO communication Modulation modulation identification Neural networks Optimization Parameter robustness Performance enhancement Signal classification sparse graph convolutional neural network
ISSN:	1074-5351, 1099-1131
On-line přístup:	Získat plný text
Tagy:	Přidat tag Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!

Popis
Shrnutí:	ABSTRACT Accurate signal classification and modulation identification are pivotal challenges in multiple‐input multiple‐output (MIMO) systems, especially under dynamic and noisy conditions. This paper proposes a novel framework—diffusion kernel attention graph convolutional neural network (DKAGCNN)—which combines the strengths of diffusion kernel attention networks (DKANs) and sparse graph convolutional neural networks (SGCNNs), optimized through the green anaconda optimization (GAO) algorithm. The diffusion kernel effectively captures complex spatial–temporal dependencies in signal data, while the attention mechanism dynamically emphasizes informative features, enhancing classification precision. The use of sparse graph convolution reduces computational complexity without sacrificing essential information, making the system scalable to high‐dimensional inputs. To further improve performance, GAO fine‐tunes network parameters adaptively, ensuring robustness and convergence in diverse environments. Experimental evaluations demonstrate that the proposed DKAGCNN surpasses current cutting‐edge techniques in terms of categorization accuracy (99.76%), F1 score (98%), and computational efficiency (12 s). These results confirm the model's suitability for real‐time and high‐precision modulation identification in advanced MIMO communication systems. The proposed DKAGCNN framework integrates diffusion kernel attention networks and sparse graph convolutional neural networks, optimized via green anaconda optimization, to enhance signal classification in MIMO systems. It captures spatial–temporal features efficiently, emphasizes critical data, and achieves superior accuracy, F1 score, and speed for real‐time modulation identification.
Bibliografie:	The authors received no specific funding for this work. Funding ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14
ISSN:	1074-5351 1099-1131
DOI:	10.1002/dac.70157