This study aims at finding the relationship between EEG-based biosignals and human emotions. Event Related Potentials (ERPs) are registered from 21 channels of EEG, while subjects were viewing affective pictures. 12 temporal features (amplitudes and latencies) were offline computed and used as descriptors of positive and negative emotional states across multiple subjects (inter-subject setting). In this paper we compare two discriminative approaches : i) a classification model based on all features of one channel and ii) a classification model based on one features over all channels. The results show that the occipital channels (for the first classification model) and the latency features (for the second classification model) have better discriminative capacity achieving 80% and 75% classification accuracy, respectively, for test data.
Paper
Full text
Classification models of emotional biosignals evoked while viewing affective pictures
Semantic Scholar · Computer Science · 2014
Abstract
This study aims at finding the relationship between EEG-based biosignals and human emotions. Event Related Potentials (ERPs) are registered from 21 channels of EEG, while subjects were viewing affective pictures. 12 temporal features (amplitudes and latencies) were offline computed and used as descriptors of positive and negative emotional states across multiple subjects (inter-subject setting). In this paper we compare two discriminative approaches : i) a classification model based on all features of one channel and ii) a classification model based on one features over all channels. The results show that the occipital channels (for the first classification model) and the latency features (for the second classification model) have better discriminative capacity achieving 80% and 75% classification accuracy, respectively, for test data.