An Efficient Separability of Motor Imagery Using Least Square Support Vector Machine for Brain Computer Interface

Although brain–computer interface (BCI) methods have been evolving quickly in recent decades, there still live a number of unsolved difficulties, such as enhancement of engine motor imagery (MI) pointer classification. In this paper, we suggest a hybrid algorithm to advance the classification achievement rate of MI-based electroencephalogram (EEG) in BCIs. The suggested design evolves a novel cross-correlation based characteristic extractor, which is aided with a least square support vector machine (LS-SVM) for two-class MI pointers acknowledgement. To verify the effectiveness of the suggested classifier, we restore the LS-SVM classifier by a logistic regression classifier and a kernel logistic regression classifier, separately, with the identical features extracted from the cross-correlation method for the classification. The suggested approach is tested on datasets; IVa and IVb of BCI affray III. The performances of those procedures are assessed with classification correctness through a 10-fold cross-validation procedure. We furthermore consider the performance of the suggested procedure by comparing it with eight recently described algorithms. Untested results on the two datasets display that the proposed LS-SVM classifier supplies an enhancement contrasted to the logistic regression and kernel logistic regression classifiers. The outcomes furthermore show that the proposed approach outperforms the most recently described eight procedures and accomplishes a 7.40% improvement over the best results of the other eight investigations.