On events recognition in optical sensing systems

Aleksey Fedorov, Maxim Anufriev, Andrey Zhirnov, Konstantin Stepanov, Evgeniy Nesterov, Dmitry Namiot, Valery Karasik, Alexey Pnev "Gaussian mixture model for events recognition in optical time-domain reflectometry based sensing systems"

The novel approach for recognition of particular classes of non-conventional events in signals from phase-sensitive optical time-domain reflectometry is proposed. The proposed algorithmic solution is based on the adaptive filtering for de-nosing of signals and Gaussian Mixture Model with the feature space formed by the cepstral coefficients for their clustering. We use experimentally measured signals from phase-sensitive optical time-domain reflectometry based sensing systems for evidence of the suggested algorithm. Our results show that two classes of events can be detected and distinguished between two classes with the probability being close to 0.9. Proposed algorithmic solution can be used in real-time distributed fiber optic sensing systems for control of protected areas.