Intrusion Discrimination in Terms of LMD and ICA With Combined Features in The Fiber-Optic Perimeter System

Abstract

For improving the performance of intrusion discrimination in the dual Mach-zehnder interferometric (DMZI) perimeter system, we propose a novel method based upon local mean decomposition (LMD), independent component analysis (ICA) and features combination. By the LMD-ICA, the original signal is processed to construct a virtual noise, thereby obtaining the sensitive information of the signal. With multiple features from the sensitive information, the type of intrusions can be discriminated by the method of serial feature fusion (SFF). The experiments are performed with real data for the case of the single-vibration and the single-vibration under the rain interference. The results demonstrate that the proposed method is superior to the traditional discrimination one, with an average recognition rate of over 96%.

Keywords

• Intrusion discrimination
• local mean decomposition (LMD)
• independent component analysis (ICA)
• features combination