|
AUTHOR(S): Mohammad Ali Alavianmehr, Mansoor Bagheri, Faezeh Teymourirad
|
TITLE |
 PDF |
ABSTRACT This paper presents an advanced nonlinearity compensation scheme for dual‑polarization 16‑QAM optical transmission systems operating over long‑haul fiber links. Our approach combines a highly accurate split‑step Fourier method (SSFM) for solving the Manakov equations—accounting for realistic impairments such as polarization mode dispersion (PMD) and inline EDFA noise—with cutting‑edge digital backpropagation techniques. We propose a novel deep learning framework that integrates convolutional neural networks (CNNs) for spatial feature extraction with a hybrid architecture that fuses attention mechanisms and recurrent neural networks—specifically gated recurrent units (GRU) and long short‑term memory (LSTM) networks—enhanced by conditional batch normalization (CBN). By incorporating an attention module, the network is enabled to dynamically focus on the most informative features, effectively mitigating both deterministic nonlinear distortions and stochastic signal–noise interactions. Simulation results demonstrate that our hybrid approach significantly reduces the bit error rate (BER), improves the Q²‑factor, and lowers the error vector magnitude (EVM) compared to conventional digital backpropagation (DBP) and other perturbation‑based equalization techniques, while maintaining a favorable balance between performance and computational complexity. These findings underscore the potential of our deep learning‑based method for real‑time implementation in next‑generation optical communication networks. |
KEYWORDS perturbation‑based post‑equalization, Nonlinearity Compensation, convolutional neural networks, conditional batch normalization, long short-term memory |
|
Cite this paper Mohammad Ali Alavianmehr, Mansoor Bagheri, Faezeh Teymourirad. (2025) Hybrid Attention-Enhanced GRU–LSTM Networks with Conditional Batch Normalization for Advanced Nonlinearity Compensation in Dual-Polarization Optical Systems. International Journal of Applied Physics, 10, 10-19 |
|
