Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
6853639 | Cognitive Systems Research | 2018 | 27 Pages |
Abstract
The electroencephalogram (EEG) is a powerful method for investigation of different cognitive processes. Recently, EEG analysis became very popular and important, with classification of these signals standing out as one of the mostly used methodologies. Emotion recognition is one of the most challenging tasks in EEG analysis since not much is known about the representation of different emotions in EEG signals. In addition, inducing of desired emotion is by itself difficult, since various individuals react differently to external stimuli (audio, video, etc.). In this article, we explore the task of emotion recognition from EEG signals using distance-based time-series classification techniques, involving different individuals exposed to audio stimuli. Furthermore, since some of the participants in the experiment do not understand the language of the stimuli, we also investigate the impact of language understanding on emotion perception. Using time-series distances as features for the construction of new data representations, applied here for the first time to emotion recognition and related tasks, lead to excellent classification performance, indicating that differences between EEG signals can be used to build successful models for recognition of emotions, individuals, and other related tasks. In the process, we observed that cultural differences between the subjects did not have a significant impact on the recognition tasks and models.
Keywords
Related Topics
Physical Sciences and Engineering
Computer Science
Artificial Intelligence
Authors
Vladimir Kurbalija, Mirjana IvanoviÄ, MiloÅ¡ RadovanoviÄ, Zoltan Geler, Weihui Dai, Weidong Zhao,