Research On Channel Selection In Motor Imagery Based Brain Computer Interface

Brain computer interface (BCI) is a new type of human-computer interaction techonology. It does not rely on the human normal peripheral nerve pathway and muscle tissue, and can directly build a communication channel between brain and external environment via computer. Currently, motor imagery (MI) based brain computer interface is one of the most popular BCI paradigms, and the extraction of sensorimotor rhythms information and effective classification of motor imagery EEGs is the key technology of it. Channel selection is an important factor affecting the performance of the system.In this paper, we proposed several channel selection methods for MI based BCI experimental paradigm, and studied the relationship between the classification performance and the number of channels as well as the influence of the difference of experimental paradigms on the number of the optimal channels. The major work and the research results of this paper are as follows:1. Regarding to the MI classification problem, most of the exisiting algorithms only adopted the characteristics of mu and beta bands. In order to overcome this defect in classification, a method of combining multiple frequency bands and spatial pattern was proposed. First, EEG signal from single channel was divided into multiple frequency bands, then the signal envelopes of the sub-bands were extracted as characteristic information, and finally with the variation of channel topology, the EEG data was classified using Pearson correlation coefficient. In addition, the relationshoip of classification performance and the channel number was roughly evaluated by using the proposed method.2. Based on the above work, a novel algorithm Relief-SBS for channel selection was proposed. This algorithm performed EEG channel selection by combining the principle of Relief with the iterative idea of sequential backward selection (SBS) algorithm. And then correlation coefficent was employed for classification of EEG signals. The selected channels that achieved optimal classification accuracy were considered as optimal channels. The data recorded from motor imagery task experiments was analyzed, and the results show that the channels selected by our proposed method achieved excellent classification accuracy, and also outperformed other feature selection methods. This shows effectiveness of our proposed method.3. The existing channel selection researches mainly focused on the MI tasks BCI paradigm without real-time feedback. The present work aims to investigate the optimal channel selection in MI tasks BCI paradigms with real-time feedback (one dimensional and two dimensional cursor control BCI paradigms). In the present study, three datasets respectively recorded from MI tasks experiment, one dimensional cursor control and two dimensional cursor control experiments were analyzed offline. Multiple frequency-spatial synthesized features were comprehensively extracted from every channel, and a new enhanced method IterRelCen was proposed to perform channel selection. IterRelCen was constructed based on Relief algorithm, but was enhanced from two aspects:change of target sample selection strategy and adoption of the idea of iterative computation, and thus performed more robust in feature selection. Finally, a multi-class support vector machine was applied as the classifier. The analyzed results demonstrated that IterRelCen has a strong ability for feature selection, and outperformed other channel selection methods. In addition, the results also showed that from MI tasks BCI paradigm, to one dimensional cursor control paradigm, and to two dimensional cursor control paradigm, the number of required channels for optimizing the classification accuracy increased and seems to positively correlate with the complexity of BCI paradigm.The proposed algorithms and these findings of this paper may provide useful information to optimize EEG based BCI systems, and further improve the performance of noninvasive BCI.