Multichannel Epileptic EEG Classification Using Quaternions and Neural Network

Epilepsy is a common chronic neurological disorder characterized by recurrent unprovoked seizures. Electroencephalogram (EEG) monitoring is a typical evaluation method. Multi-channel EEG signals have much more discrimination information than what only one channel has. But traditional EEG signal recognition algorithms are lack of effective fusion of multi-channel EEG signals. In this paper we propose the quaternion representation of multichannel EEG signals. We also make use of the quaternion principle component analysis (QPCA) method to extract multichannel EEG features. New representation method is compared with the traditional method and the experimental results show the better ability of the quaternion approach.