Events in fiber optics given noisy OTDR data. I. GSR/MDL method

This paper proposes a novel method of detecting and locating connection splice faults (events) in fiber optics by the digital signal processing (DSP) of noisy optical time-domain reflectometry (OTDR) data. This is motivated by the fact that as fiber becomes more widely adopted as a communications medium, methods of automated fault detection/location will become more important. The approach taken is to use Gabor series expansion coefficients to coarsely localize the faults. Due to the presence of measurement noise, these coefficients are random variables, and it is Gabor coefficients with a nonzero mean that determine fault presence and location. Coefficients with nonzero mean are found with the aid of Rissanen´s minimum description length (MDL) criterion for model order estimation. The results show that the method is able to distinguish connection splice events from noise and the Rayleigh component in the OTDR data