Simultaneous baud rate/modulation format identification and multi-parameter optical performance monitoring using multi-task learning with enhanced picture of Radon transform

Abstract

We propose a simultaneous baud rate identification (BRI), modulation format identification (MFI) and multi-parameter optical performance monitoring (OPM) scheme for elastic optical networks (EONs). Firstly, based on the advantage that enhanced picture of Radon transform (EPRT) contains more phase, amplitude, and density information of signals, the proposed scheme takes EPRT, which is obtained from the original constellation diagram of EON signals after pretreatment of grayscale processing, edge detection and Radon transform, as the key feature for signal parameter identification and monitoring. Subsequently, the EPRTs are fed into a multi-task learning (MTL) neural network to perform BRI, MFI, chromatic dispersion identification (CDI), differential group delay (DGD) monitoring and optical signal-to-noise ratio (OSNR) estimation simultaneously. The numerical results of 14/28 GBaud polarization division multiplexing (PDM)-EON transmission system demonstrate that the accuracies of BRI, MFI and CDI could achieve almost 100% after 105 epochs, 114 epochs and 174 epochs respectively. For the task of OSNR estimation, its mean absolute errors (MAE) converges to 0.468 dB after 187 epochs, and the MAE of DGD monitoring is stable at 0.021 times symbol period after 170 epochs. Furthermore, the effectiveness of the proposed scheme is further verified by 14/28 GBaud PDM-EON proof-of-concept experimental system. The results manifest that the MFI and BRI accuracy achieve almost 100% after 118 epochs and 175 epochs, respectively. For the OSNR estimation task, its MAE converges to 0.542 dB after 200 epochs.