EEG based emotion classification using "Correlation Based Subset Selection"

Emotion detection is one of the popular research topics in "Brain-Computer Interfacing" where researchers are trying to find the various emotional states of people. EEG signal is widely used for detecting different categories of emotions. The EEG signal is captured through multiple electrode channels, very few of them are useful for emotion detection. In our paper, a "Correlation-based subset selection" technique is introduced for dimension reduction. Then we proceed with classification process using "Higher Order Statistics" features of the reduced set of channels. However, we have classified four classes of emotions (positive, negative, angry and harmony) in our paper. The execution time of our proposed algorithm is O(n(2) + 2n). The classification accuracy of this model with the reduced set of channels is 82%. Finally, we compare our proposed model with some popular emotion classification models and the result shows that our model substantially outperforms all the previous models. However, the proposed model helps physically disabled people to express their feelings with minimum time and cost-effectively.