Classification of Error-Related Negativity (ERN) and Positivity (Pe) potentials using kNN and Support Vector Machines

Authors:
Errikos M. Ventouras;Pantelis Asvestas;Irene Karanasiou;George K. Matsopoulos
Affiliations:
Department of Medical Instrumentation Technology, Technological Educational Institution of Athens, Ag. Spyridonos Street, Egaleo, Athens 12210, Greece;Department of Medical Instrumentation Technology, Technological Educational Institution of Athens, Ag. Spyridonos Street, Egaleo, Athens 12210, Greece;Institute of Communications and Computer Systems, National Technical University of Athens, 9, Iroon Polytechneiou Street, Zografou Campus, Athens 15773, Greece;School of Electrical and Computer Engineering, National Technical University of Athens, 9, Iroon Polytechneiou Street, Zografou Campus, Athens 15773, Greece
Venue:
Computers in Biology and Medicine
Year:
2011

Citing 5
Cited 0

Floating search methods in feature selection

Pattern Recognition Letters
Learning and Soft Computing: Support Vector Machines, Neural Networks, and Fuzzy Logic Models

Learning and Soft Computing: Support Vector Machines, Neural Networks, and Fuzzy Logic Models
New modifications and applications of fuzzy C-means methodology

Computational Statistics & Data Analysis
You are wrong!: automatic detection of interaction errors from brain waves

IJCAI'05 Proceedings of the 19th international joint conference on Artificial intelligence
Brief communication: Efficiency analysis of KNN and minimum distance-based classifiers in enzyme family prediction

Computational Biology and Chemistry

Quantified Score

Hi-index	0.00

Visualization

Abstract

Error processing in subjects performing actions has been associated with the Event-Related Potential (ERP) components called Error-Related Negativity (ERN) and Error Positivity (Pe). In this paper, features based on statistical measures of the sample of averaged ERP recordings are used for classifying correct from incorrect actions. Three feature selection techniques were used and compared. Classification was done by means of a kNN and a Support Vector Machines (SVM) classifier. The use of a leave-one-out approach in the feature selection provided sensitivity and specificity values concurrently higher than or equal to 87.5%, for both classifiers. The classification results were significantly better for the time window that included only the ERN, as compared to time windows including also Pe.