Abstract:

The difference of degree of nonlinearity between normal and epileptic electroencephalogram (EEG) signals was investigated by quantified delay vector variance (DVV) method. The quantified DVV method can be used to determine the degree of nonlinearity of the analyzed time series based on surrogate data. We compare the results of the degree of nonlinearity of EEG signals obtained by different surrogate data generation methods, namely Iterative Amplitude Adjusted Fourier Transformation (IAAFT) and Phase Randomization (PR). It is shown that there exists difference between the degrees of nonlinearity of the EEG signals obtained by the two surrogate data generation methods. However, results obtained by the two surrogate data generation methods show that both normal and epileptic signals are nonlinear, and the degree of nonlinearity of epileptic EEG signals is higher than that of normal EEG signals. Therefore, it is proposed that the difference of degree nonlinearity between normal and epileptic EEG signals can be used as the feature for the detection of epileptic seizure.

Key words: information processing, degree of nonlinearity, quantified delay vector variance, electroencephalogram (EEG) signal