Use of CNN to assess emotions evoked by auditory stimuli in videos

Consumption of immersive audio content, e.g., via binaural reproduction through headphones, has increased over time. However, objective models that allow classifying audio systems through human perception are still poorly researched. Therefore, this study aims to evaluate the change in emotional state caused by immersive musical content. The evaluation was conducted subjectively, using a \textit{Mean Opinion Score - MOS} model and objectively, by classifying the subjects' facial reaction with a Convolutional Neural Network, based on the VGG-16. The results obtained are favorable to the expected objectives, and may serve as a trigger for further studies.