Machine learning approach for epileptic seizure detection using wavelet analysis of EEG signals

Analysis of EEG is the primary method for diagnosis of epilepsy. In this paper discrete wavelet transform is used for the time-frequency analysis of EEG signal. Using discrete wavelet transform, EEG signal is decomposed into five different frequency bands namely delta, theta, alpha, beta and gamma. Only theta, alpha and beta carry seizure information. Statistical feature like energy, variance and zero crossing rate and nonlinear feature like fractal dimension is extracted from each of the three sub bands and fed to support vector machine classifier. Support vector machine classifies the input EEG signal into seizure free and seizure signal. Experimental results show that the proposed method classifies EEG signals with excellent accuracy, sensitivity and specificity compared to the existing methods.