인간정서 디코딩을 위한 비침습형 뇌파 분석의 방법론적 상관관계 연구

Objective: This paper aims to quantify the features derived from non-invasive brain wave elicited by human emotions, and to investigate correlations between the features based on information theory. Background: Numerous researches have aimed to decode human emotion - divided into dimensional and discrete emotional states – by utilizing neurophysiological analysis methods. Decoding emotions from electroencephalogram (EEG), in particular, is widely studied since it is considered to be a direct method. But, the accuracy of such analysis may be inconsistent depending on the correlation between emotion-related EEG features. Thus, a study that performs quantitative analysis of the correlation between emotions and EEG features, and analyze information index for emotion decoding would be required. Method: Based on emotion decoding literatures, correlation between accuracy and the number of EEG channels or features was quantified in advance. And Hick’s law was utilized in order to quantify information level of the brain wave in each emotional state. Results: For decoding of emotions, 10~20 channel signals are likely to yield 1.1 bit of information. Conclusion: The greatest limit in current level of emotion decoding from EEG is that the level of information gained is highly dependent on the number of recorded channels. Many studies these days are aimed at reducing the number of channels required and, furthermore, at eliminating such dependency.