Unsupervised, Modulation-Independent Detection of Harmful Fiber Events from SOP Signatures in a Live Metro Ring
Paper in proceeding, 2027
We propose a stand-alone, unsupervised SOP-monitoring framework, requiring no labels or hyper-parameter tuning, with channel alignment to mitigate modulation-format dependence. On a live metro ring, it achieves 97–99% accuracy for CW/DP-16QAM and 79–88% for QPSK.
Optical Fiber Monitoring
Unsupervised Learning
State of Polarization
Modulation-Independent Detection
Anomaly Detection
Machine Learning
Optical Networks
Author
Leyla Sadighi
Trinity College Dublin
Marija Furdek Prekratic
Chalmers, Electrical Engineering, Communication, Antennas and Optical Networks
Stefan Karlsson
Micropol Fiberoptics AB
Devika Dass
Trinity College Dublin
Marco Ruffini
Trinity College Dublin
Optical Fiber Communication Conference (OFC 2027)
Optical Fiber Communication Conference (OFC 2027)
Los Angeles, USA,
Los Angeles, USA,
Subject Categories (SSIF 2025)
Computer Sciences
Telecommunications