Toward letter recognition system: Determination of best wavelet and best rhythm using EEG

Abstract

The paper describes the application of different wavelet analysis together with machine learning algorithm for the recognition of English alphabet from EEG signal. Decision making was executed in two stages. At first important features such as maximum, minimum, delta value, moment, kurtosis, skew, median, mean and standard deviation at different sub-bands are computed using the wavelet functions — Daubechies 8, Coiflet 6, Biorthogonal 4.4, Symlet 4. Finally, a learning-based algorithm like support Vector Machine (SVM) classifier is implemented for classifying letters. From the analysis, Daubechies 8 is found the most suitable candidate among the wavelet families in this proposed research for accurate recognition of different letters. So the focus of this work is to recognize different letters through SVM classifier. In this analysis, among different rhythms of EEG signal delta rhythm shows best performance in recognizing letters and the recognition rate is 80%.