Intelligent Pattern Recognition Using Distributed Fiber Optic Sensors for Smart Environment

Area of Science:

• Optoelectronics and Sensor Technology
• Artificial Intelligence and Machine Learning
• Cybersecurity and Surveillance

Background:

• Distributed fiber optic sensors (DFOSs) are vital for intrusion detection in sensitive areas.
• Conventional methods face challenges with complexity, computational load, and signal path inefficiency.
• Advancements in signal processing and deep learning are needed to enhance DFOS performance.

Purpose of the Study:

• To introduce an innovative interferometric sensing approach for intrusion detection.
• To improve the accuracy and efficiency of DFOS systems.
• To explore the application of a Mach-Zehnder interferometer (MZI) combined with a time forest neural network (TFNN).

Main Methods:

• Utilized a Mach-Zehnder interferometer (MZI) for signal acquisition.
• Employed a time forest neural network (TFNN) for pattern recognition and intrusion detection.
• Compared performance against conventional one-dimensional convolutional neural networks (1D-CNN).

Main Results:

• The proposed MZI-TFNN approach demonstrated superior performance in intrusion detection.
• Achieved an 8.43% higher accuracy compared to the 1D-CNN.
• Indicated enhanced efficiency in real-time signal processing.

Conclusions:

• The novel interferometric sensing method with TFNN offers a significant improvement over existing DFOS techniques.
• This approach holds substantial potential for real-time intrusion detection in smart environments.
• Highlights the effectiveness of integrating advanced optical sensing with deep learning for enhanced security applications.