Preprocessing

When pre-processing EDA data, one typically down-samples or filters the data with a low-pass filer (generally <10 Hz) [11]. The next step is to remove artifacts of electrodes displacements. Motion artifacts can be identified by visual inspection of the data [8], but this is cumbersome for very large datasets. [11] recommends setting a signal-change threshold criterion and reject changes that violate this threshold. Kleckner et al. [18] developed an automated quality assessment for ambulatory data, that can be used to remove motion artifacts from acquired EDA data. This tool is publicly available¹ and detects out-of-range values, fast changes in EDA, moments in which the device was not being worn, and data surrounding invalid data.

^1.EDAQA