TABLE 2.
Summary of preprocessing methods for PPG.
| Preprocessing method | Details | Purpose |
| Frequency filtering | Bandpass filter | Reduction for high-frequency noise, baseline movement reduction |
| - 1st order Butterworth [(0.5 – 5) Hz] (Sukor et al., 2011) | ||
| - 2nd order Butterworth [(0.2 – 10) Hz] (Liu et al., 2020b) | ||
| - 3rd order Butterworth [(0.4 – 10) Hz] (Papini et al., 2018) | ||
| - 4th order Butterworth [(0.5 – 50) Hz] (Pradhan et al., 2019) | ||
| - 4th Chebychev I [(0.5 – 16) Hz] (Ferro et al., 2015) | ||
| - 4th order Butterworth [(0.5 – 10) Hz] (Canac et al., 2019) | ||
| - 64th order FIR [(0.1 – 10) Hz] (Selvaraj et al., 2011) | ||
| - Discrete cosine transform filtering [(0.5 – 10) Hz] (Shin et al., 2010) High pass filter |
||
| - 4th order Butterworth, cut-off: 0.01 Hz (Fischer et al., 2017) Low pass filter |
||
| - 2nd order Butterworth, cut-off 10 Hz (Liu et al., 2020a) | ||
| - 4th order Butterworth, cut-off 15 Hz (Fischer et al., 2017) | ||
| Empirical mode decomposition | Waveform reconstruction using intrinsic mode functions whose dominent frequency is > 0.5 Hz (Lu et al., 2008) |
Reduction for low-frequency (<0.5 Hz) noise and baseline noise reduction |
| Wavelet transform | Signal reconstruction using specific sub-bands after stationary wavelet transform (Vadrevu and Manikandan, 2018) | Suppression of background artifacts and noises |
| Independent component analysis | Reducing motion artifact using frequency domain independent component analysis based on red and infrared signal (Krishnan et al., 2008) | Motion artifacts reduction |
| Moving difference filter | Calculating the difference with the sample after a window size of a moving window (Canac et al., 2019) | Enhancing upslope of the photoplethysmogram |
| Curve fitting | Amplitude normalization | Eliminating non-stationary dynamics |
| - Amplitude compensation curve (Kim et al., 2019) | ||
| Detrending | ||
| - 32nd-order polynomial fitting (Selvaraj et al., 2011) |