Understanding Human Implicit Intention based on Electroencephalography (EEG)

Abstract

There are many approaches that understand human intention. Although previous study has mainly used behavioral data which are explicitly expressed such as speech, gesture, and touch strokes, but human intention cannot be revealed explicitly in real life. Whether it is intended or by accident, there are many situations that we do not disclose our minds. In this study, EEG was examined to investigate the human implicit intention. Subjects showed their intention by their voices whether they agree or disagree toward given obvious and non-obvious sentences. Experiment focuses on some situation that human may not want to express their real intention to others and may answer differently. It is assumed that brain activation may not be the same when they show differently with their real intention, so it can be a measurement of the implicit intention. ICA is applied to extract independent components of the recorded EEG data. And only few components were selected based on Fisher Linear Discriminant (FLD) which can discriminate between agreement and disagreement state. Using selected components, support vector machine trained with obvious condition identified the validation sample from the classifier output. The results showed that SVM output of selected independent components can discriminate implicit intention states, and recognize non-obvious condition. It may be used to understand implicit intention.