A dual stage wavelet workflow for motion artifact localization and correction in reflected light fNIRS signals

Motion artifacts in functional near-infrared spectroscopy (fNIRS) signals can distort reflected light waveforms and affect derived physiological estimates. This study presents a dual stage, joint ROI guided wavelet workflow for motion artifact localization and correction in reflected light fNIRS signals from the PhysioNet databaseMotion Artifact Contaminated fNIRS and EEG Data. The method combines continuous wavelet transform based joint ROI detection across the 690 and 830 nm signals with local dual threshold discrete wavelet correction and smooth boundary blending. Performance was assessed by mean squared error (MSE) within artifact contaminated ROIs and by root mean square error (RMSE) of a ratio based oxygen saturation estimate relative to the clean reference channel. Compared with spline-Savitzky-Golay and global dual threshold wavelet benchmarks, the proposed workflow achieved the lowest estimate domain RMSE in five of nine trials. These findings support joint ROI guided local wavelet correction as a competitive strategy for preserving ratio derived reflected light estimates in this dataset.