Research On Reducing Calibration Time In Motor Imagery-Based Bcis

Brain-computer interface(BCI)is a communication system that does not rely on peripheral nerves and muscles,and uses external devices to communicate with the outside world.One of the main reasons that limit the practical application of BCIs is long calibration time.In this dissertation,based on BCIs of motor imagery(MI),we proposed a new method to reduce the training(calibration)time of BCIs without sacrificing classification accuracy.The method aims to augment the subject’s training set data by generating artificial Electroencephalography(EEG)data set using a small number of initially available training trials.First,we use the Euclidean-space Alignment(EA)method to align a small amount of available training trails and test trails to a common reference point.The aligned training trails are then subjected to Empirical Mode Decomposition(EMD)to obtain Intrinsic Mode Functions(IMFs),and the selected IMFs are mixed to obtain an artificial EEG data.Repeating several times to obtain the required artificial EEG data set.Finally,the artificial EEG data set are used as an expansion set and together with the original small data set as a new data set,which is used to build a training model.In this thesis,a Linear Discriminate Analysis(LDA)classifier or a Logistic Regression(LR)classifier is used to discriminate the test trials.The performance of the proposed algorithm is evaluated on two motor imagery(MI)data sets and compared with that of the algorithm trained with only real EEG data(Baseline)and the algorithm trained with expanded EEG data by EMD without data alignment.The experimental results showed that the proposed algorithm can significantly reduce the amount of training data needed to achieve a given performance level and thus is expected to facilitate the real-world applications of MI-BCIs.