ANN Based Automatic Identification and Classification of Ocular Artifacts and Non-Artifactual EEG

The electroencephalogram (EEG) signal plays an eminent and consequential part in the diagnosis and evaluation of treatment in neurological disorders. EEG artifact is a sequence with higher amplitude and different shape, size, morphology, statistics compared with non-artifactual EEG and other signals (ECG, EMG, etc.). This paper presents first time, ANN incorporated with three classes for better classification and the new way for automatic identification of ocular artifacts through variance feature. Training and testing is done using variance of ocular artifacts (Eye blinks artifacts-EBA/Eye movements artifacts-EMA) and non-artifacual EEG signals. These will fed to the three classes ANN model. Supervised ANN is used to build an automatic classification for Eye blink, Eye movement artifacts and non-artifactual EEG. The variance of different OAs and non-artifactual EEG will create three classes target tables. These variance values are assigned as the input labels for the training data sets. Sensitivity, specificity and classification accuracy for EBA is estimated as 96.66%, 96.66%, and 96.66% while for EMA is figure out as 100%, 96.66% and 98.33 %. With more EEG datasets for all classes the classification accuracy can be further improve.