Autor(es): Assunção, Gustavo ; Patrão, Bruno ; Menezes, Paulo
Data: 2022
Identificador Persistente: https://hdl.handle.net/10316/100593
Origem: Estudo Geral - Universidade de Coimbra
Assunto(s): emotion recognition; machine learning; human perception
Emotionally pleasant experiences trigger repetition in humans whilst their emotional opposite lead to avoidance/refusal of activities. Human interest, a key factor in everyday human life, can be quite useful in the evaluation of participant engagement on activities and their optimization, so as to maximize interest and motivation. A glaring issue however comes from subtle demonstrators associated with human interest, which converge onto its complex assessment/quantization. Yet, there is an inherent correlation between human interest and its provoked emotional response which can be explored in tandem with emotion recognition for the development of an engagement metrics tool. Such a mechanism would be highly beneficial for the improvement of several activities involved with learning and security, by enabling precise control over participant enthusiasm. In this paper we present an interest mapping technique which provides the user with spatio-temporal information extracted from a participant crowd. The technique aims to extract emotional cues from participant facial data, assessing its spatial and temporal distributions over the course of scenarios such as therapy sessions and lectures. The goal is to demonstrate where and when activities must be improved in order to retain attention, maximize efficacy and assure emotional pleasantness in participants. For validation, this study makes use of data collected over a college lecture so as to provide readers with a real demonstration of the technique's advantages.
