Recognition of Music-Induced Emotions Based on Heart-Brain Connectivity

摘要

Emotion recognition using physiological signals has received a great deal of attention in recent years. Local features of signals have been more commonly used in this field of research. Brain connectivity measures have also been applied to emotion recognition in last few years. Considering the extensive neural interactions between the brain and heart and the intrinsic nervous system of the heart, and bearing in mind the different autonomic nervous system activity in various emotional states, we expect that the connectivity between the heart and brain can be measured as a discriminative feature for the purpose of emotion recognition. In the present paper, we investigate the possibility of recognizing four music-induced emotions on the basis of functional connectivity between the heart and brain. One of the discriminating features in multichannel/multisource signal processing is Mutual Information. Herein, mutual information is estimated between the power of three channels of forehead biosignals in six sub-bands and the normalized high frequency component of HRV. Subject-dependent and subject-independent classification studies were carried out using SVM classifier, which resulted in an average accuracy of 95.7% and 86.9%, respectively. The promising results suggest that the amount of information which our brain and heart provide about each other is different in each emotional state and may be used to distinguish between them.