Reconhecimento facial e FNIRS para a detecção de traços de emoção aplicados a jogos sérios

• Autor(a) principal: Souza Júnior, Milton Machado de
• Data de Publicação: 2019
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da PUC_RS
• Texto Completo: http://tede2.pucrs.br/tede2/handle/tede/8452
• topic: Jogos Sérios; Reconhecimento Facial; FNIRS; Emoçôes; Classificadores; Serius Games; Facial Recognition; FNIRS; Emotions; Classifiers; ENGENHARIAS

This work aims to improve the recognition of emotional responses to different stimuli that provoke alterations into humans. Said alterations can be measured through several tools such as Magnetic Resonance Imaging, Functional Near Infrared Spectroscopy (fNIRS), Electroencephalography (EEG) and Facial Recognition. In this work, we used a combination of the fNIRS system, which measures the variation of hemoglobin oxygenation, and facial recognition tools. The participants were stimulated with a sequence of images from the IAPS database, which are labeled with the weighted emotions they provoke. This allowed the training of a classifier that was capable of predicting the emotion that was experienced by the user during the activity performed. This article covers the recognition of positive, negative and neutral emotions using classifiers, created in Matlab, that are based on participant physiological responses. The combination of fNIRS, facial recognition and machine learning supported the creation of a predictor with 77.2% correct classification rate.