Selecting the optimal conditions of Savitzky–Golay filter for fNIRS signal

Abstract This paper proposes a method to find the best conditions for applying Savitzky–Golay (SG) filter to remove physiological noises from the functional near-infrared spectroscopy (fNIRS) signal. A narrative review on existing physiological noise reduction techniques from fNIRS signal demonstrates that the most common methods are window based finite impulse response (FIR) and SG filters. However, these filters did not clarify why and how it is able to remove noises from the fNIRS signal. This paper shows a systemic investigation of works performed with window based FIR filter and SG filter and found very convincing results to use SG filter with specific conditions. Three main frequency bands (0–0.1 Hz, 0–0.14 Hz, and 0.03–0.1 Hz) have been considered as standard for fNIRS signal filtering and filtered the signals by window-based FIR filter. With a number of conditions of SG filter, the raw fNIRS signals were filtered again and checked the correlation between filtered signals by FIR and SG. By check and trial basis, the best correlations were revealed. To validate the proposed results, several golden standard statistical investigations were analyzed. The experimental results propose a recommendation which indicates the best conditions of the SG filter to remove physiological noises from the fNIRS signals.