Wavelet transform based multiple features extraction for detection of epileptic/ non-epileptic multichannel EEG

摘要

Electroencephalogram is the process of capturing spontaneous neural activity of brain by placing electrodes along the scalp. EEG plays a vital role in analyzing neuronal disorder like epilepsy. In this research work, we use wavelet transform method to extract and analyze multiple features of epileptic multichannel EEG. Instead of discrete wavelet transform or continuous wavelet transform, here we used Stationary wavelet transform to perform five level decomposition of multichannel EEG signal with sampling rate 114 Hz. The advantages of SWT are good directionality and retention of phase information in the EEG signal. Four statistical wavelet features (Max, Min, Variance and Energy) are evaluated from the wavelet coefficients of each decomposed sub-band for multichannel EEG dataset of 10 subjects. Wavelet features analysis and visual inspection shows the significant difference for epileptic and non-epileptic EEG signals. Also the affected lobe for epileptic subject is identically verified by them. Finally we cross validated the decision through wavelet feature analysis and identification of affected lobe for epileptic multichannel EEG subjects' with Neurophysician remark.