Study of pattern recognition based on SVM algorithm for φ-OTDR distributed optical fiber disturbance sensing system

Currently, phase sensitive optical time-domain reflectometer (-OTDR) distributed optical fiber sensing system is difficult to accurately determine current position of disturbance and distinguish the model of disturbance effectively. A method was proposed based on support vector machine (SVM) which can accurately distinguish disturbance and the model of disturbance. With the technique of the binary tree, a categorizer based on SVM was set up by extracting the various signal characteristics of the mean, the variance, the mean square deviation and energy of the time-domain and frequency-domain. Thus the disturbance and disturbance mode can be distinguished. In terms of the sensing signal feature, the categorizer I was determined if the sensing signals was disturbance signals or not firstly. Then, mode of disturbance can be recognized by the following categorizers. Experiments were carried out to validate the proposed method by 600 groups of data. The correct discrimination rate is better than 96%. The rate of missing report and the rate of false positives is less than 4%. The rate of correct pattern recognition is greater than 94%.