PROJECT TITLE :

Event Discrimination of Fiber Disturbance Based on Filter Bank in DMZI Sensing System

ABSTRACT:

To meet the growing demand of event discrimination in a very dual Mach–Zehnder Interferometry (DMZI) vibration sensing system, this paper proposes a completely unique scheme of distinguishing invasion events. This scheme consists of 3 stages: endpoint detection, filter-bank-based Feature Extraction, and radial-basis-function neural network classification. Because the laborious core, the proposed filter bank, which comes from the classical frequency-sampling-based filter design technique, greatly suppresses the interchannel interference and, thus, provides correct feature description for the radial basis operate (RBF)-primarily based classifier. Moreover, we tend to additionally derive the closed-type formula of the filter coefficients and, so, style a pipeline structure for the filter bank, that brings the benefits of high accuracy, high flexibility, nice rapidity, and low price. Simulation verifies the proposed filter bank's superiority in separating different frequency bands. Field experiments additionally show that the proposed event discrimination scheme cannot only eliminate some false alarms caused by noninvasion events but can discriminate 2 common invasion events (climbing the fence and knocking the cable) with high recognition accuracy furthermore.