Automatic Selection of Multi-view Learning Techniques and Views for Pattern Recognition in Electroencephalogram Signals

Abstract

The present work explores six different Multi-view learning (MVL) techniques for the classification of electroencephalogram (EEG) signals in order to take advantage of complementary descriptive information from different representations of the same object. We worked with four views of EEG signals extracted by applying two different feature extraction methods in time domain and two in the frequency domain. We propose a model for automatic selection of view combination, using the total number of views, then three views and finally two views with each MVL approach explored, based on classification performance. The classification accuracy achieved by the Multi-view learning approach and the subset of views selected by our model exceeds the results achieved in single view works where the same databases are used for pattern recognition in EEG signals.