An approach for letter recognition system modeling based on prominent features of EEG

Letter recognition system is a novel approach in the field of communication with the external world using human brain activity. The system is based on temporal and spatial analysis to extract salient features of raw electroencephalogram (EEG) signal. Among various features amplitude, skewness, mean value of EEG signal are chosen which indicate the largest dispersion for different letters and help to evaluate letter recognition system. Then the raw EEG signal is analyzed using FFT and wavelet. Both wavelet transform and statistical analysis distinguish letters more precisely than FFT analysis. The overall recognition rate is 80% and 85.6% for statistical and wavelet analysis, respectively. It is shown that our proposed system is capable of recognizing English alphabet efficiently and reliably.