Performance Monitoring of High-Speed NRZ Signals Using Machine Learning Techniques

Abstract

Advances in high-speed communication network technologies have spurred interest in signal performance monitoring. This study proposed a 25-Gb/s non-return-to-zero (NRZ) signal performance monitoring method using an artificial neural network (ANN), which can estimate the five parameters of Q factor, signal-to-noise ratio, time jitter, rise time, and fall time. Using 5000 data sets and adopting seven neurons in the hidden layer, the mean relative errors of the five estimated parameters are about 5.76% to 11.74%. This parameter extraction technique based on machine learning can apply to real-time optical network performance monitoring for high-speed NRZ signals.