A Regular Composite Feature Extraction Method of Vibration Signal Pattern Recognition in Optical Fiber Link System

Optical fiber link system based on phase optical time domain reflectometer uses optical fiber as sensor media. It can detect, recognize and locate invasive signal. The module of pattern recognition is an important part of the system. It intelligently and instantly distinguishes danger disturbances from the safe ones. This paper proposes a regular composite feature extraction method of vibration signal pattern recognition in optical fiber link system. This method applies improved double threshold method to detect the start-stop positions of the valid segment of the signal, then combines the max energy and max signal noise ratio to select the main regular and valid feature segment of invasive disturbance signal within sampling period. Finally, the composite eigenvectors can be extracted by using the last time of the feature segment and wavelet packet energy spectrum. Support vector machine is used to do signal pattern recognition. The experiment results show that the pattern recognition accuracy rate has been much improved. The method proposed in this paper is able to effectively recognize different types of optical fiber vibration signal in real time.