Adapting the Interplay Between Personalized and Generalized Affect Recognition Based on an Unsupervised Neural Framework

Recent emotion recognition models, most of them being based on strongly supervised deep learning solutions, are rather successful in recognizing instantaneous expressions. However, when applied to continuous interactions, these models show a weaker adaptation person-specific and long-term appraisal. In this article, we present an unsupervised neural framework that improves by how describe affective behavior individual persons. Our is composed three self-organizing mechanisms: (1) recurrent growing layer cluster general expressions, (2) set associative layers, acting as affective memories model specific emotional persons, (3) online which provides contextual modeling We propose different strategies integrate all mechanisms improve the performance arousal valence OMG-Emotion dataset. evaluate our with series experiments ranging from ablation studies assessing contributions each component objective comparison state-of-the-art solutions. The results evaluations good emotions monologue videos. Furthermore, discuss self-regulates interplay between generalized personalized perception influences model’s reliability unseen