Toward number recognition system: A nonstationary signal analyzing approach through SVM algorithm

Non-stationary signal analysis based on visual stimulation has drawn extensive attention in BCI system to provide the promising services. The main task of this paper tries to evaluate specific pattern of each decimal number created in human brain using the specific features of EEG. For differentiating among the decimal numbers, salient features are extracted using time, frequency and time-frequency domain analysis and SVM classifiers are used to demonstrate the primitive features. It is observed that sigmoid kernel provides the highest accuracy than the other used classifiers and numbers are differentiated clearly by the best distinguishable features of PSD analysis.