Extracting features for a brain-computer interface by self-organising fuzzy neural network-based time series prediction

This paper presents a novel feature extraction procedure (FEP) for extracting features from the electroencephalogram (EEG) recorded from subjects producing right and left motor imagery. Four self-organizing fuzzy neural networks (SOFNNs) are coalesced to perform one-step-ahead predictions for the EEG time series data. Features are derived from the mean squared error (MSE) in prediction or the mean squared of the predicted signals (MSY). Classification is performed using linear discriminant analysis (LDA). This novel FEP is tested on three subjects offline and classification accuracy (CA) rates approach 94% with information transfer (IT) rates >10 bits/min. Minimum subject specific data analysis is required and the approach shows good potential for online feature extraction and autonomous system adaptation.