Identification of external media using a long-period grating and optical time-domain reflectometry.

The application of signal processing and feature extraction to reflectometry traces aiming at sensor interrogation is not sufficiently explored. In this work, traces produced by an optical time-domain reflectometer in experiments using a long-period grating in different external media are analyzed using signal processing techniques inspired by audio processing. The objective is to demonstrate that, using this type of analysis, it is possible to correctly identify the external medium through the characteristics of the reflectometry trace. Results show that the features extracted from the traces were able to produce good classifiers, one of them achieving 100% correct classification for the data set presently considered. This technology could be applied in scenarios where it is necessary to distinguish among a given set of gases or liquids nondestructively.