Exploring dimensionality reduction of EEG features in motor imagery task classification

摘要: A Brain-Computer Interface (BCI) system based on motor imagery (MI) identifies patterns of electrical brain activity to predict the user intention while certain movement imagination tasks are performed. Currently, one most important challenges is adaptive design a BCI system. For solving it, this work explores dimensionality reduction techniques: once features have been extracted from Electroencephalogram (EEG) signals, high-dimensional EEG data has be mapped onto new reduced feature space make easier classification stage. Besides standard sequential selection methods, paper analyzes two unsupervised transformation-based approaches – Principal Component Analysis and Locality Preserving Projections Local Fisher Discriminant (LFDA), which works in supervised manner. The projected chosen following wrapper-based approach by an efficient leave-one-out estimation. Experiments conducted five novice subjects during their first sessions with MI-based systems order show that appropriate use methods allows increasing performance. In particular, obtained results LFDA gives significant enhancement terms without computational complexity and, then, it promising technique for designing