An optic-fiber fence intrusion recognition system using mixture Gaussian hidden Markov models

Abstract

Since the development of optic-fiber interferometers and design of outstanding speech recognition models, the study on perimeter intrusion detection systems (PIDS) becomes a field of interest. In this paper, an optic-fiber based fence intrusion detection and recognition system that uses Sagnac interferometers and mixture Gaussian hidden Markov models (GMM-HMMs) is proposed. Experiments on real fence intrusions are performed, and comparisons are also carried out between our approach and the SVM based method, which prove our system more robust and accurate.