چكيده به لاتين
A Brain- Computer Interface (BCI) is a kind of human-computer interface that collects and classifies Electroencephalogram (EEG) data to communication between human and computer. The P300 wave is an Event-related potential occurs as a response to familiar visual in an oddball paradigm, one of the common brain- computer interface application is the P300 speller that designed to comminucate language between human and computer, it uses the 6×6 matrix of characters, the subject must focus on one character and P300 occurs when the target rows or columns are flashed.
To do that this study EEG recorded from 10 channels of three subjects in offline mode and after preprosessing and denoising we try to detect P300 wave.
The other datasets we used them in this study the datasets provided by the BCI competition III.
This study goals to detect the P300 wave as accurate as possible, So in this research, we proposed method to feature extraction and used linear discriminant without averaging from all trials for classification achieve to better performance.
Results of different feature extraction methods (Fisher’s Linear Discriminant (FLD), Common Sparse Spectral Spatial Pattern (CSSSP),Time features) and classification methods (Linear Discriminant Analysis (LDA), Artificial Neural Network (ANN), support vector machine (SVM) with optimize parameters) compare and the proposed method yeilds better performance compared to the other methods.
Results demonstrate a classification accuracy (P300 & Non-P300) of %98.51 and %99.44 for subject A and B, respectively, and word prediction accuracy for 5 and 15 trials was %73.78 and %96.50, respectively.
In addition, our results indicate a significant improvement and progress in classification accuracy compared to the other feature extraction and classification approaches.
Keywords: Brain-Computer Interface (BCI), P300 speller, Artificial Neural Network (ANN), support vector machine (SVM), Common Sparse Spectral Spatial Pattern(CSSSP)