Optical signal to noise ratio monitoring and modulation format identification based on multi-scale channel attention network

This paper proposes an optical signal-to-noise ratio (OSNR) monitoring and modulation format identification (MFI) technique for elastic optical networks, based on constellation diagrams and the multi-scale channel attention network (MCA-Net). MCA-Net integrates depthwise separable convolution for efficiency, Inception multi-scale parallel structure for robust feature extraction, and Squeeze-and-Excitation (SE) channel attention mechanism for adaptive feature weighting from constellation diagrams. The simulation results demonstrate that the proposed MCA-Net-aided technique can effectively achieve OSNR monitoring and MFI based on the constellation diagrams. Within the OSNR range of 15 to 24 dB, the OSNR monitoring accuracy of MCA-Net for 56 Gbit/s QPSK, 8PSK, 16QAM, and 32QAM modulated signals reaches 100%, 100%, 99.5%, and 98.0% respectively. Meanwhile, the MFI accuracy of this technique for all four types of signals reaches 100%. We also explore the impact of the training data volume, the OSNR step size, and the network structure on the performance of MCA-Net. The research results show that the proposed MCA-Net-aided technique delivers excellent OSNR monitoring and MFI performance and could be embedded into elastic optical network performance monitoring instruments for future intelligent signal analysis.