Purpose Several recent studies have employed recordings of error potentials as communication channels within a brain-computer interface. Decoding an error and thus facilitating direct control of an external device via neural signals can be deployed for a medical application for people with severe motoric deficits. To this purpose, the reaction time and accuracy of the system are elements of first importance. Methods We introduced a straightforward method of classifying error potentials with a strategy that combines two classifiers on different levels, quadratic discriminant analysis (QDA) followed by the support vector machine (SVM) algorithm. More than that we used an accessible and portable device, Emotiv EPOC(+)that allows experiments in close to real-life scenarios. Results Across n = 3 subjects, we found perceptible (on average with 24.5% for the unpreprocessed data and 23.4% for the preprocessed data), but statistically not significant (Kruskal-Wallis Test p = 0.09 for unpreprocessed data and p = 0.27 for preprocessed data) improvements in classification performance of a data set using the QDA algorithm as a low-level classification and subsequently SVM algorithm as a meta-classifier. Conclusion Our approach might reduce the classification time period optimizing the results of the analysis.     «

Purpose Several recent studies have employed recordings of error potentials as communication channels within a brain-computer interface. Decoding an error and thus facilitating direct control of an external device via neural signals can be deployed for a medical application for people with severe motoric deficits. To this purpose, the reaction time and accuracy of the system are elements of first importance. Methods We introduced a straightforward method of classifying error potentials with a...     »