Question about Savitzky–Golay filter

💬 Support & Discussions
, ,
1
  • MNE version: e.g. 0.24.1
  • operating system: Windows 10

Dear all

For enhancing SNR, I want to do temporal smoothing, and I decide to use Savitzky–Golay filter. However, I am not sure how to set the “window_length” and “polyorder” for the filter. Here Eric Larson suggested that a 5th-order polynomial with a window length twice as large as the wavelength of the freq of interest would be OK. Does anyone have any idea why these parameters are OK? Also, I could not find any studies that used this filter for temporal smoothing. Does anyone have any study in his/her mind that used Savitzky–Golay filter?

Best

2

IIRC my suggestion was based on the gist linked in the Notes for example in: mne.Evoked — MNE 1.0.dev0 documentation

Personally I don’t bother to use savgol filtering because using FIR filter design for low-pass / smoothing I can immediately characterize the effects on the signals. With savgol it’s a bit more of a guessing game for me at least (not that familiar with the technique)

I don’t know of any that have used it. From a quick google search of “Savitzky-Golay electroencephalography” I found at least:

1 Like