Abstract
In the recent years, there has been a significant growth in the area of brain computer interference. The main aim of such area is to read the brain activities, formulate a specific/desired output and power a specific approach using such output. Electroencephalography (EEG) may provide an insight into the analysis procedure of the human behavior and the level of the attention. Using the deep learning based neural network has a great success in different applications recently, such as making a decision, classifying a pattern and predicting an outcome by learning from a set of data and build the right weight matrices to represent the prediction outcome or the learning patterns. This research work proposes a novel model based on long short-term memory network to predict the standard and the deviant cases within EEG data sets. The EEG signals are acquired utilizing all the 128 electrodes that represent the 128 channels from infants aged between 5 and 7 months. Statistical approaches, principal component analysis (PCA) and autoregressive (AR) power spectral density estimate have been employed to extract the features from the EEG data sets. The proposed deep learning based model has shown great robustness dealing with different types of features extracted from the processed data sets. Very promising results have been achieved in predicting the standard and deviant cases. The standard case was presented with frequent, repetitive stimulus and the deviant case was presented with infrequent sounds.