Fuzzy classification of human emotions using Fuzzy C-Mean (FCFCM)

Emotion classification using electroencephalogram (EEG) has been actively researched in many Human-Computer Interaction (HCI) applications such as healthcare, recreation and education. In the last decade, studies in affective Brain-Computer Interfaces (aBCI) have increased especially in the discipline of affective computing. One of the topics is fuzzy classification of human emotions. Fuzzy set theory has been applied due to its potential benefit of human comprehensibility. Furthermore, the approach exhibits shorter computation time and it is faster than other machine learning techniques if it has only a number of manageable rules. It is also suitable for online applications. However, it is important to have an efficient means for extracting appropriate rules. In this study, it is proposed to use Fuzzy C-Mean (FCM) for extracting fuzzy rules to be used in the fuzzy inference system (FIS) for classification purpose. This technique uses the EEG data for learning and to produce rules. Three advantages are suggested: accuracy, flexibility and computation time. The results show that the algorithm has an improved accuracy in comparison to fuzzy classification using fixed rules and Support Vector Machine (SVM), with values of 55.77%, 49.62% and 54% respectively. In addition the proposed technique also required less execution time.