چكيده به لاتين
In recent decades, using of Electroencephalogram (EEG) and processing of it for communicating to surroundings, caused Brain Computer Interface (BCI) appear. P300 signal is a type of EEG signal that used in BCIs system. P300 is a visual evoked potential that usually is evoked using specific pattern stimulation. In this research, vehicle control using P300 signal done. For this work, a visual stimulation pattern for evoking P300, containing six keys designed and EEG recorded simultaneously from five subjects. Any experiment has two states: offline and online. In offline experiment, after EEG recording and preprocessing, four feature extraction methods; CSP, CTP, FLD and proposed method, and three classification methods; LDA, SVM and Bayes, used for classification P300 and non-P300 signals. Proposed method in this research introduced as cost function using FLD and SNR, and for maximization of it, Genetic Algorithm used. Results of different feature extraction and classification methods compare and among them SVM classifier using proposed feature extraction method (SVM+GA) has the best result and so, in online experiments used from it. In offline results, SVM+GA mean accuracy of all subjects, for single and double trials, respectively, was 87.71% ± 1.11 , 95.80% ± 0.84. These results averaged on 10-fold cross validation. Using Leave-One-Out validation method, mean accuracy of all subjects for double trials was 95.49% ± 1.20 for SVM+GA. In online results, mean accuracy of all subjects was 89.28% ± 4.73 and mean Information Transfer Rate (ITR) was 86.03 ± 5.87 bits/min.
