We exploit a self-supervised deep multi-task learning framework for
electrocardiogram (ECG) -based emotion recognition. The proposed solution
consists of two stages of learning a) learning ECG representations and b)
learning to classify emotions. ECG representations are learned by a si