Intrusion recognition method based on echo state network for optical fiber perimeter security systems

Abstract

In order to accurately and efficiently identify different types of intrusion signals of optical fiber perimeter security systems, this paper proposes a novel intrusion signal recognition method based on an echo state network (ESN). A perimeter security system based on an in-line Sagnac interferometer is employed to simulate in lab two laying situations of the sensing fiber: attached to a physical fence or buried under ground and acquire various event signals. The preprocessed signals are input into a trained well ESN to identify different types of events. The final recognition result of an event signal segment is determined according to the most dominant classification label corresponding to the signal segment. As a result, the average identification rates are 98.75% and 100% for the two laying situations, respectively. The proposed method has no need of extracting signal features and a large number of samples to train the classifier model. Therefore, more accurate and more effective intrusion identification can be achieved by the method than by others. The method is expected to satisfy the requirements of the practical application in the security field.