Classification Of Emotions By Majority Voting Of Classifiers Based On Multimodal Physiological Features

Emotion recognition from physiological signals such as electroencephalogram, electrocardiogram, blood pressure, etc. is a difficult problem. In order to elevate correct classification rate, we tried majority voting method. Four different single classifiers were constructed based on multimodal physiological features. When we tried classification into 3 categories, each single classifier showed correct classification rate of 38.39%, 38.25%, 40.66%, and 35.15% respectively. Weighted majority voting of their result was adapted as final conclusions and its correct classification rate was 41.58%, which was improved by 0.92%. Although the effectiveness of majority voting method is confirmed, the correct classification rate is still far from satisfactory. The performance of emotion recognition should be improved much further.