Classification of EEG signal from imagined writing using a combined Autoregressive model and multi-layer perceptron

EEG signal contain massive information on brain activities which can be extracted by filtering and processing the signal at specific frequency. The similarity in the EEG signals obtained during actual and imagined writing exists and can be revealed using good representation of the signals. A technique called Autoregressive (AR) is able to model the EEG signals which can be used as input feature for Multi Layer Perceptron. In this study, the EEG signals recorded during actual and imagined writing was analyzed and classified to find the frequency range where similarity in both signals exists. The results obtained indicate that there is similarity in the signals especially at frequency of 8-13 Hz (Mu region).