Positive Emotion Recognition—A Survey of Computational Models

In this paper, we provide a systematic survey of works on positive emotion recognition and detection. We queried major research paper databases for 12 different emotional states: admiration, amusement, awe, compassion, contentment, elation, enthusiasm, excitement, gratitude, pride, relief, and sympathy. From an initial pool of more than 800 papers, we selected 81 that propose recognition models, and categorized them according to the modality and data type used for recognition. According to the results, the most frequently occurring labels are amusement and excitement, followed by compassion and contentment. In terms of data type, the highest number of solutions utilize various physiological signals, with visual data being the second most common. As the first survey to focus on a large number of distinct positive emotions, our paper contributes to Affective Computing and, more broadly, to Human-Computer Interaction by demonstrating that: 1) the topic of positive emotion recognition remains largely unexplored, despite numerous potential applications; 2) there are significant shortcomings in the research on positive emotion recognition, including the lack of relevant datasets, contextual information, and adequate data collection procedures. Finally, we provide suggestions and guidelines to support future research in this area.