Personalization of Emotion Recognition for Everyday Life Using Physiological Signals From Wearables

Emotion recognition from physiological signals has many promising applications, but it remains an unsolved challenge. There are many scientific papers trying to construct general systems suitable for the entire population. Our research focuses on the subject-oriented approach and strategies for personalization of emotion recognition. Our solutions will fuse generalized and individualized deep learning architectures in different ways. For experimental purposes, we will collect unique datasets gathered in everyday life using various wearables: smartwatches and chest straps. We will also develop appropriate validation procedures capable of measuring both the personalization and generalization properties of the considered methods. This paper describes the problem's background, its scope, and our research methodology. Additionally, we present our achievements so far and our future work.