In distributed optical fiber perimeter security system

in order to recognize intrusion and interference events effectively in realtime

a vibration signal pattern recognition algorithm based on time-domain features in actual optical fiber perimeter security system is proposed. Original vibration signal is processed to reduce the noise

short-term energy and short-term threshold rate are extracted and formed feature vector. Next

the feature vector is input into support vector machine（ SVM） for training and test and establish classification model

which is used to security system and recognize intrusion and interference event. Experimental results show that the pattern recognition algorithm classification is high-efficiency

high-accurate

effectiveness and feasibility.