A New Improved Feature Extraction Method in Memory EEG Data

various papers and conferences about EEG data can be found at present. There are various feature extraction methods reviewed oversimplified in section one, such as zerocrossing, low zero-crossing rates, coherence analysis, subspace methods, the mean absolute amplitude, standard variance, kurtosis and so on. The feature extraction methods such as self-produced mother wavelet feature extraction method, best basis-based wavelet packet entropy feature extraction, empirical mode decomposition and non-linear feature extraction using correlation dimension and Hurst exponent are detailed introduced in section two. Those feature extraction methods are complex and limited, which often used in some specific fields. In this paper, a new feature extraction is proposed named incremental value, which considers the changes in brain waves. Next LDA and classification tree are used to analyze the results of feature extraction and to predict with unequal memory error compared with the feature extraction methods, such as mean absolute amplitude, standard variance and kurtosis. The method that we proposed is concise and accurate than other methods.