A Study On Spatial Filtering And Feature Extraction Methods In Multi-Task Brain-Computer Interfaces

The low information transfer rate (ITR) is an intrinsic problem that binary BCIs face, and restricts their practical application. The most effective method to increase ITR is to extend two mental tasks to multiple tasks. In this paper two algorithms are proposed to generalize binary CSP algorithm to multiple task condition.The first algorithm proposed is multi-class feature extraction method based on Common Spatial Pattern, which is a highly successful method for the motor imagery based on brain-computer interfaces (BCIs) in the case of two task conditions. The traditional method called OVR is still a binary in nature. The paper generalizes binary CSP algorithm to multiple task conditions by approximate joint diagonalization based on quadratic optimization. Several spatial filters corresponding to each task are obtained the electrophysiology feature, event-related desynchronizing (ERD). Another algorithm is based on independent component analysis (ICA) algorithm. ICA is one statistical method closely related to the method called blind signal separation. In this paper, a typical ICA algorithm FastICA is utilized for feature extraction by decomposing EEG signals into independent sources. At last, a support vector machine (SVM) is used as classifier in these two algorithms.These algorithms are respectively applied to three-task and four-task data sets recorded during motor imagery based on BCI experiments. The results demonstrate that the performance of the algorithm based on approximate joint diagonalization is better than traditional method, and the method based on FastICA is also satisfactory.