• affecting computing
• convolutional neural network
• deep learning
• electroencephalogram
• emotion recognition
• explainable artificial intelligence
• grad-cam
• interpretable machine learning
• transfer learning

In recent years, it has become more and more important to detect and recognize the emotion in computer systems which people interact with via BRI (Brain Computer Interface). Among the various types of BCI, one of the most common is the EEG (Electroencephalogram).
In this thesis, we perform emotion recognition using EEG features and BHI (Brain-Heart Interplay) features as input for a subject-adactive Convolutional Neural Network. We test our method on the publicly available DEAP and MAHNOB databases, for each of which we perform experiments subject-specific, subject-independent and fine-tuned leave-one-subject-out.
The grad-CAM visualization technology is also applied to show an intutive representation of the feature importance.