Related Experiment Video
Updated: Jun 16, 2025
09:03
A Silicon-tipped Fiber-optic Sensing Platform with High Resolution and Fast Response
Published on: January 7, 2019
7.1K
Layer normalized OpenGAN framework for human intrusion activity recognition in a φ-OTDR optical fiber sensing system
Optics Express
|June 14, 2025
Summary
This study introduces ResEff-OpenGAN-LN, an advanced framework for distributed optical fiber vibration sensing systems. It significantly improves the accurate classification of known events and detection of unknown events, enhancing perimeter security.
Area of Science:
- Engineering
- Computer Science
Background:
- Distributed optical fiber vibration sensing systems (DVS) are crucial for perimeter security due to their sensitivity and robustness.
- Current DVS systems struggle with accurately classifying known (closed-set) and detecting unknown (open-set) events.
Purpose of the Study:
- To develop an open-set recognition framework to address the limitations of current DVS systems.
- To enhance both closed-set classification accuracy and open-set detection capabilities.
Main Methods:
- Proposed an open-set recognition framework named ResEff-OpenGAN-LN.
- Integrated layer normalization into the OpenGAN architecture to stabilize training.
- Utilized ResEff for efficient feature extraction to improve closed-set classification.
Main Results:
- ResEff achieved 99.92% accuracy on closed-set classification tasks.
- ResEff-OpenGAN-LN demonstrated an AUROC of 0.9900 on mixed datasets.
- Achieved 96.63% overall accuracy in detecting both open-set and closed-set signals.
Conclusions:
- The ResEff-OpenGAN-LN framework effectively enhances DVS system performance.
- This approach shows significant potential for improving intrusion detection and reducing false alarms in perimeter security.
