Feature Domain Transform Filter for the Removal of Inherent Noise Bound to the Absorption Signal

Area of Science:

• Signal processing
• Chemical sensing

Background:

• Traditional time-frequency domain filtering methods face limitations in separating noise with identical signal characteristics.
• Accurate detection of gases like methane is crucial for environmental and safety monitoring.

Purpose of the Study:

• To introduce a novel filtering algorithm, the feature domain transform filter (FDTF), for enhanced signal processing.
• To improve the performance of methane gas sensors through advanced noise reduction techniques.

Main Methods:

• Developed an FDTF combining principal component analysis (PCA) for feature transformation, deep learning for information extraction, and time domain transformation.
• Validated the FDTF using simulated and experimental datasets for methane gas detection.

Main Results:

• The FDTF effectively filters noise with the same frequency and phase as the target signal.
• FDTF demonstrated superior performance compared to existing time-frequency filtering algorithms.
• The FDTF-enhanced methane sensor exhibited good linearity and achieved a minimum detectable column density of 2.50 ppm·m.

Conclusions:

• Feature domain filtering represents a significant innovation over traditional time-frequency approaches.
• The FDTF significantly improves the precision, stability, and sensitivity of methane gas sensors.
• This study validates the efficacy of signal processing in a new domain for improved data analysis.