Discerning Genuine and Acted Smiles Using Neural Networks

Deception recognition is a growing area of interest in studies involving electroencephalograms (EEGs), as it paves the way for further research in demystifying and understanding the human brain. Furthermore, it has other practical applications that can prove to be beneficial to several fields of research and industry. A part of deception recognition is identifying fake smiles and true smiles given in response to certain stimuli. Here, we develop a neural network to predict whether a smile is fake simply by going through the EEG signals, given an example dataset. The network used, obtained an average accuracy of 72.38 % before principle component analysis was applied, and 78.29 % after for all subjects. In addition, the average sensitivity for all subjects was 77.35 % and the average specificity was 88.29 %.