search

TomasLenc / acf_tools

Utilities for working with autocorrelation to capture periodicity in a signal.
GNU General Public License v3.0
2 stars 0 forks source link

keep narrow band rather than single frequency bin in last step of noise correction #14

Closed TomasLenc closed 9 months ago

TomasLenc commented 9 months ago

Based on Dounia's suggestion, zeroing out all frequencies except of harmonics of the response may add "oscillatory" sinc activity to the time-domain signal. We do have a good reason to do this step, given that our response is expected to be periodic with an exact integer number of cycles completed within the time span of the analysed signal. However, it may be worth giving the user an option to keep a small band (with ramps) around each response frequency. So instead of hard-zeroing out all non-response frequency bins, we would have something more gradual, kind of comb-filter(ish).

cedlenoir commented 9 months ago

Maybe in the docstring of get_acf it could be clearer that in the example [20, 30] +/-20 bins around are kept untouched and bins from 20 to 30 are progressively linearly set to zero. I do not get the difference between "keep_band_around_f0_harmonics : [int, int], default=[0, 1]" and to keep only the exact bin this param should be [1, 1].