Multi-class classification of EEG spectral data for artifact detection
Fragment książki (Rozdział monografii pokonferencyjnej)
MNiSW
20
Poziom I
| Status: | |
| Autorzy: | Tokovarov Mikhail, Plechawska-Wójcik Małgorzata, Kaczorowska Monika |
| Dyscypliny: | |
| Aby zobaczyć szczegóły należy się zalogować. | |
| Wersja dokumentu: | Drukowana | Elektroniczna |
| Język: | angielski |
| Strony: | 305 - 316 |
| Web of Science® Times Cited: | 1 |
| Scopus® Cytowania: | 1 |
| Bazy: | Web of Science | Scopus |
| Efekt badań statutowych | NIE |
| Materiał konferencyjny: | NIE |
| Publikacja OA: | NIE |
| Abstrakty: | angielski |
| Electroencephalographic signals are known to be highly sensitive tovarious types of noise originating from external and internal sources. Externalsources are usually related to experiment conditions whereas internal artifactsare usually generated by the persons examined. Internal artifacts, in contrast toexternal ones, are characterised by non-stationarity, which results in higherdetection complexity. Typical internal artifacts are related to eye blinking, eyemovement and muscle activity.The paper presents the comparison results of various approaches to classifi-cation of EEG-related features. Another aspect reviewed in the paper is com-paring normalisation methods of dealing with inter-subject variability.The analysis process covered several parts. Preprocessing included signalfiltering and bad-channel removal. Signal epoching with 50% overlap wasapplied in order to achieve better time resolution. Feature extraction was basedon frequency analysis performed with the Welch method. Due to the highnumber of obtained features, the feature selection procedure was the essentialpart of the processing. Selected features were used to train and validate super-vised classifiers. The accuracy was the main measure used in classifier perfor-mance assessment. |
