Distributed φ-OTDR signal classification based on VMD and SVM

Abstract

With the development of distributed fiber optic sensing, the recognition of different vibration modes has become increasingly important. In this paper, a distributed external-heterodyne φ-OTDR system is used for outdoor vibration acquisition and mode recognition. In the event recognition experiment, feature vectors are obtained using the Variational Mode Decomposition (VMD) algorithm. Support Vector Machine (SVM) is then used to classify the input feature vectors accordingly. Through this experimental method, it is possible to accurately identify five types of vibrations: stepping, tapping, wind blowing, passing subway, and passing car. Compared to traditional methods, the accuracy has improved from 72.50% to 91.25%, demonstrating promising application prospects.