Skip to main content

Table 1 Ranking of features obtained by LPM.

From: Automatic Classification of Artifactual ICA-Components for Artifact Removal in EEG Signals

Feature

Weight

Current Density Norm

Range Within Pattern

Mean Local Skewness 15 s

λ

8-13 Hz

FitError

Border Activation

2DDFT

Spatial Mean Activation Central

Max First Derivative

Variance

k 2

Spatial Mean Activation Left

Spatial Mean Activation Left Frontal

Laplace-Filter

Mean Local Variance 15 s

14-30 Hz

4-7 Hz

Mean Local Variance 1 s

Spatial Distance of Extrema

Spatial Mean Activation Occipital

k 1

Maximum Amplitude

y

Spatial Mean Activation Right

Kurtosis

x

0-3 Hz

Deterministic Entropy

Spatial Mean Activation Frontal

z

Variance of Local Variance 1 s

Range

Spatial Mean Activation Right Frontal

Variance of Local Variance 15 s

Mean Local Skewness 1 s

31-45 Hz

Shannon Entropy

  1. The diameter of the black circles visualizes the absolute LPM weights |E(w i /||w||)| per feature after learning on the training data set. (The LPM-hyperparameter C had been set to C = 0.1 based on the cross-validation performance.)