Proposal for the Optimal Scheme of Savitzky–Golay Filter for the fNIRS Signals of Different Sampling Rate

This paper introduces an approach for determining optimal parameters to effectively apply the Savitzky–Golay (SG) filter to remove physiological noises in functional near-infrared spectroscopy (fNIRS) signals. The most common method for fNIRS signal filtering is a window-based FIR filter with a 0.1Hz cut-off frequency which is erroneous to some extent considering its higher Euclidean distance between the original and filtered signal. Although the SG filter doesn’t have this issue, achieving the selective cut-off frequency has no specific guidelines. This paper investigated systematically to suggest the optimal conditions of SG filter instead of FIR filter to achieve the targeted cut-off frequency with the high signal-to-noise ratio (SNR). With different possible conditions of the SG filter, the raw fNIRS signals were filtered and compared to the signals that are filtered by an FIR filter. To ensure the robustness of the obtained results, various established statistical measures, including Pearson’s correlation coefficient, Signal-to-Noise Ratio (SNR), average absolute deviation, and others, were thoroughly examined. The experimental findings put forth a recommendation, highlighting the optimal parameters for the Savitzky-Golay (SG) filter to effectively eliminate physiological noises from functional Near-Infrared Spectroscopy (fNIRS) signals.