Answers are grounded in this paper and cite their sources.
Accurate modulation format identification (MFI) and optical signal-to-noise ratio (OSNR) estimation are essential for intelligent coherent optical receivers. This study proposes a temporal–radial feature fusion framework that jointly performs MFI and OSNR estimation using dual-polarization in-phase and quadrature signals after constant-modulus algorithm equalization. Unlike approaches relying on a single signal representation, the proposed design explicitly exploits complementary feature domains and carefully evaluates fusion benefits under matched parameter budgets. The architecture combines a temporal convolutional neural network with polarization-resolved radial amplitude histograms, integrating complementary temporal and statistical information through shared representations. Performance is assessed using simulated 30-Gbaud coherent signals with quadrature phase-shift keying and 16-, 32-, 64-, and 128-ary quadrature amplitude modulation. Under back-to-back conditions, the proposed model achieves 100 % MFI accuracy and an OSNR mean absolute error (MAE) of 0.1516 ± 0.0163 dB over 5 training seeds. Against a parameter-matched temporal-only convolutional baseline with 91,990 trainable parameters, feature fusion reduces OSNR estimation error by 78.2 % while improving MFI performance. Generalization is further examined without retraining on 9,000 simulated records from a 160-km linear-fiber link, yielding 99.778 ± 0.076 % MFI accuracy and 0.6038 ± 0.0134 dB OSNR MAE. Some high-OSNR conditions exhibit larger estimation errors. These findings indicate that complementary temporal and radial features enhance joint optical performance monitoring while maintaining a controlled model complexity.
No takes yet. Share an insight, caveat, or question.
Alhussain et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: