Double wavelet threshold filter design for real-time adaptive denoising in OA-ICOS
Yixin Deng, Zairan Wang, Yangyang Chen, Qin Liu, Xiangyu Wang · IET conference proceedings. · 2025
In the Off-Axis Integrated Cavity Output Spectroscopy (OA-ICOS) technique, single-stage filters are commonly used to reduce noise and improve the signal-to-noise ratio, enabling more precise CO2 concentration detection. However, these filters often underperform under real-time conditions. To enhance real-time performance, this paper proposes a novel framework based on Double Wavelet Threshold Filter (DWTF) that adaptively suppresses noise while preserving key features of CO2 concentration signal. Firstly, one wavelet threshold filter (WT) is introduced to calculate the mean of the current sequence, providing an initial denoising step. Secondly, another WT is employed to extract features from a longer sequence for adaptive computation of the discount factor. Finally, an exponential moving average filter (EMA) is applied to combine both current and historical data to produce a smoother aggregated output. To validate the effectiveness, DWTF was compared with other conventional algorithms for a CO2 dataset with three distinct states: smooth, ship and mutant, obtained by OA-ICOS. The experimental results demonstrate that DWTF performs effectively in all three states.