Artifact Detection

Find artifacts in signals using spectral energy. Uses the BOTMpy library.

Functionality and Options

A period is considered an artifact when the energy of a single frequency contains a defined fraction of the total signal energy during the period. The fraction is defined by the option "Threshold".

The length of the considered periods can be set using the "FFT samples" option. The "Cutoff frequency" determines the maximum frequency at which artifacts are expected to occur. The "Overlap percentage" can be used to achieve more fine grained artifact borders at the cost of performance. It determines if and how much the periods used for the FFT overlap.

Finally, a heuristic is employed to merge and/or drop artifact epochs: First, artifact periods up to "Maximum merging distance" * "FFT samples" samples apart are merged into single epochs. Next, epochs shorter than "Minimum artifact length" * "FFT samples" are not considered artifacts (this is useful to exclude outliers when the artifact energy fluctuates strongly).

Neo epochs are created and added to the correct segment. Each epoch is labeled with the "Epoch label" and gets an annotation 'tag' defined by the "Annotation Tag".

All files

VersionViewer versionUpload dateSize
0.1 0.3.0 June 24, 2013 3.9 KB


New comment

Please log in to leave a comment.

Author: Philipp Meier


Download most recent version: 0.1

License: BSD