Fig. 4

Block diagram of classification process for personalized average model: all trials of each subject were proceeded to the main classification block. Random downsampling was performed to remove data imbalance from face and no-face classes. Then typical machine learning classification process was executed with sixfold nested cross-validation technique. Here simple filter feature selection technique (t test) was followed by artificial neural network for the two class problem. Finally, the outcomes are classification accuracy, sensitivity and specificity of each subject