光纤监测系统入侵信号的检测和识别

Detection and Identification of Intrusion Signals in Fiber Optic Monitoring System

为改善光纤监测系统监测信号特征复杂及入侵信号检测效率和识别精度低等问题,提出了一种改进的谱减法进行去噪处理,从而有效减少噪声干扰并获得具有高信噪比的监测信号;其次,由于光纤监测信号大部分为静默信号,因此对其进行端点检测从而有效提取入侵信号;最后,基于提取的入侵信号的时频域特征,提出了一种多特征融合深度学习识别模型,从而准确识别入侵类型.

In order to improve the complexity of monitor signals′characteristics and low detection efficiency and recognition accuracy of intrusion signals in fiber optic monitoring system,an improved spectral subtraction method was proposed for denoising to effectively reduce noise interference and obtain monitor signals with high signal-to-noise ratio.Secondly,since most of the fiber optic monitor signals are silent signals,endpoint detection was performed to effectively extract intrusion signals.Finally,based on the time-frequency domain features of the intrusion signals extracted,a multi-feature fusion recognition deep learning model was proposed to accurately identify the intrusion signals.