Adaptive multi-source separation strategy for a distributed optical fiber sensor based on a mask network.

Area of Science:

• Optical Engineering
• Signal Processing
• Sensing Technology

Background:

• Separating overlapping multi-source events in φ-OTDR systems is challenging, especially with limited training data.
• Data scarcity can cause overfitting in mask estimators, hindering the isolation of desired signal components.
• Conventional methods often rely on data augmentation to address limited datasets.

Purpose of the Study:

• To propose a novel adaptive separation strategy for multi-source events in φ-OTDR systems.
• To overcome the limitations of data scarcity and improve the separation of overlapping signals.
• To enhance the quality of separated results without extensive data augmentation.

Main Methods:

• Developed an adaptive separation strategy that iteratively refines poorly separated signals.
• Leveraged existing datasets and mask estimators within the iterative process.
• Validated the approach through theoretical analysis and experimental testing.

Main Results:

• The adaptive strategy effectively separates overlapping multi-source events in φ-OTDR systems.
• Demonstrated improved signal isolation quality under data-limited conditions.
• Showcased the feasibility and portability of the strategy across various scenarios.

Conclusions:

• The proposed adaptive separation strategy offers a novel solution for multi-source event separation in φ-OTDR.
• The method is effective even with limited training samples, addressing a key challenge in the field.
• The strategy shows promising potential for practical applications in optical sensing.