To cite this paper use one of the standards below:
Automatic facial emotion recognition has emerged as a complementary tool to traditional subjective assessment methods, offering greater objectivity in identifying emotional reactions. In the field of sensory analysis, this resource can broaden the understanding of the relationship between taste stimuli and affective responses, reducing reliance on self-reports and minimizing individual biases. Therefore, this study aimed to investigate the feasibility of applying computer vision models to recognize facial expressions in sensory analysis training, with an emphasis on hedonic evaluations. The study was conducted with students from the Food Technology program at Fatec/Marília-SP, during a practical class focused on recognizing basic tastes (sweet, salty, umami, sour, and bitter). The computer system used was developed in Python, using the PyTorch framework to build a convolutional neural network trained with the public AffectNet database, composed of thousands of images annotated into distinct emotional categories. For the experimental phase, facial expressions were recorded on video and processed with the OpenCV library, which employs computer vision modules for accurate face detection. The extracted images were analyzed by the proposed model, whose output was subjected to a Gaussian mixture model with 14 emotional components, categorized as positive (happy, hopeful, surprised, proud, loved, euphoric) and negative (angry, sad, fearful, disgusted, fatigued), as well as neutral states. The results revealed a predominance of positive emotions (59.3%), indicating that the sensory tests were perceived as pleasant experiences, in line with the hedonic nature of the activity. We also observed the sporadic occurrence of negative reactions, such as disgust (0.89%), and a significant presence of neutral expressions (10.94%), which is considered expected in this context. The analysis suggests that the model is capable of reflecting emotional tendencies consistently with the experimental observations, although adjustments are necessary to fully adapt it to the specificities of sensory analysis. It is noteworthy that, more than simply linking absolute emotional states, the tool proved effective as a reference for distinguishing acceptance or rejection reactions to taste stimuli. The conclusion is that integrating computer vision models with sensory training represents a promising, automated, and non-invasive tool for monitoring human emotional responses. This approach has potential for application in consumer perception research, contributing to greater objectivity, robustness, and innovation in food sensory analysis methodologies.
Views per month
Galoá count, excluding known robots. Updated on 30/09/2026.
Views per month
Galoá count, excluding known robots. Updated on 30/09/2026.
With nearly 200,000 papers published, Galoá empowers scholars to share and discover cutting-edge research through our streamlined and accessible academic publishing platform.
Learn more about our products:
This proceedings is identified by a DOI , for use in citations or bibliographic references. Attention: this is not a DOI for the paper and as such cannot be used in Lattes to identify a particular work.
Check the link "How to cite" in the paper's page, to see how to properly cite the paper