Home // INFOCOMP 2014, The Fourth International Conference on Advanced Communications and Computation // View article

Research on Classification of Fiber Intrusion Signal Based on Supported Vector Machines

Authors:
Jie Zhu

Keywords: fiber sensor; differential phrase demodulation; supported vector machine; Time-Frequency analysis; wavelet de-noising.

Abstract:
The widely use of optical fiber gives rise to the need of its protection and intrusion detection. An optical fiber system in which the optical path satisfies the structure of sagnac loop can easily form a distributed fiber senor. With the photo-elastic effect, when intrusion happens, there will be optical signals created in the fiber, and with optical and electrical methods, one can get the intrusion related signals for analysis. After obtaining the signals, we use differential phrase demodulation method to demodulate the signals. With the demodulated signals, the feature vector of the signals can be extracted through time-frequency analysis. Then, supported vector machine (SVM) is used to classify 3 different types of intrusion. For better accuracy of classification, we use wavelet de-noising to do the noise elimination. Field experiments showed that the system is reliable and of good accuracy.

Pages: 163 to 167

Copyright: Copyright (c) IARIA, 2014

Publication date: July 20, 2014

Published in: conference

ISSN: 2308-3484

ISBN: 978-1-61208-365-0

Location: Paris, France

Dates: from July 20, 2014 to July 24, 2014