Optical Fiber Vibration Signal Recognition Based on the EMD Algorithm and CNN-LSTM

Area of Science:

• Optoelectronics
• Signal Processing
• Machine Learning

Background:

• Accurate identification of optical fiber vibration signals is essential for perimeter security systems.
• Distributed acoustic sensing (DAS) using phase-sensitive optical time-domain reflectometry (φ-OTDR) is a key technology for intrusion detection.
• Improving the recognition accuracy of intrusion events in φ-OTDR systems remains a challenge.

Purpose of the Study:

• To enhance the recognition accuracy of intrusion events detected by φ-OTDR systems.
• To propose a novel identification method combining EMD, CNN, and LSTM.
• To validate the effectiveness of the proposed method in real-world environments.

Main Methods:

• Optical fiber vibration signals were decomposed using empirical mode decomposition (EMD).
• Effective intrinsic mode functions (IMFs) were selected based on correlation coefficients and reconstructed.
• Convolutional neural networks (CNNs) extracted time-series features, and long short-term memory (LSTM) networks classified the signals.

Main Results:

• The proposed EMD-CNN-LSTM method effectively identified three different types of vibration signals.
• A recognition accuracy of 97.3% was achieved for intrusion signals.
• The method demonstrated successful pattern recognition for φ-OTDR systems.

Conclusions:

• The combined EMD-CNN-LSTM approach significantly improves intrusion detection accuracy in φ-OTDR systems.
• This method offers valuable insights for developing practical engineering products in perimeter security.
• The study successfully addresses challenges in φ-OTDR pattern recognition.