Electroencephalogram Based Brain-Computer Interface: Key Techniques and Application Prospect

Several key techniques of brain computer-interface based on motor imagery are introduced. For the feature extraction, emphasized are the common spatial patterns (CSP) and discriminative spatial patterns (DSP) filters; for the pattern recognition, stressed are the Bayesian linear discriminant analysis (BLDA) employed large probabilistic test samples to expand the training set, the transductive support vector machines (TSVM), the manifold-based Laplacian support vector machine (LapSvm), and the hierarchical Bayesian linear discriminant analysis. For on-line system realization, amplifier designing and the idle-state detection are described. Finally, the potential future directions are discussed.