Skip to Main Content
In Part I (Liu and Zarowski 2001), a method of connection splice event detection and location by the digital signal processing of noisy optical time-domain reflectometry data was given. It applied the Gabor series representation (GSR) for transient signal detection of Friedlander and Porat (1989) to the optical time-domain reflectometer (OTDR) data yielding Gabor coefficients. The Rissanen minimum description length (MDL) criterion was then applied to the Gabor coefficients to eliminate those that are most likely due to noise alone. The surviving coefficients are indexed in such a manner as to indicate the location of events in the OTDR data. Although this GSR/MDL method is computationally efficient, it only gives a coarse estimate of event location. Therefore, this paper (Part II) develops the rank-1 matched subspace detection and estimation algorithm that employs more computation to achieve a greater accuracy in event position estimation. It is based on the matched subspace detection theory of Scharf and Friedlander.