A Combined Events Recognition Scheme Using Hybrid Features in Distributed Optical Fiber Vibration Sensing System

Abstract

In this paper, a high efficiency multiple events recognition scheme based on a hybrid feature extraction algorithm and a combined classifier for distributed optical fiber vibration sensing (DOFVS) system has been proposed and demonstrated. The hybrid feature vectors are extracted by using zero crossing rate, sample entropy, wavelet packet energy entropy, kurtosis, and multiscale permutation entropy. A combined classifier of support vector machine and radial basis function neural network is proposed to improve the reliability of the recognition results. The recognition result is given only when both of the two classifiers output same event types. The experimental results demonstrated that the average identification rate of five typical patterns (no intrusion, waggling the fence, climbing the fence, kicking the fence, and cutting the fence) over 97% is achieved through the combined classifier. Moreover, the whole recognition processing speed of the combined scheme is also good of real time performance, which can be limited in 1.1 s. Therefore, this kind of events recognition scheme has a quite promising application prospects in DOFVS system.

Keywords

• Distributed vibration sensing
• optical fiber
• events recognition
• hybrid feature vectors
• combined classifier
• signal analysis