Classifying Emotion Using Convolutional Neural Networks

Jonathan L Moran
2019 UC Merced Undergraduate Research Journal  
Despite the computer's historical success as a communication tool, machines themselves have yet to fully master the most basic forms of nonverbal communication that we humans use daily. Gender, ethnicity, age and emotional state is often perceived immediately by most humans engaging in conversation. In face-to-face interactions, humans can form broad generalizations about an individual's social status, health, and well-being within a blink of an eye. Training a classifier algorithm to
more » ... this form of human behavior is a rather difficult task. While the accuracy of exchange of non-verbal messages may be questioned, the vast amount of information humans can generalize from these thinly-sliced events is a true feat of human intelligence. In this paper, we will be exploring the concepts of object recognition and deep learning neural networks to ultimately train a classification model to recognize universal human emotion from the FER-2013 facial expression dataset (Kaggle, 2013) .
doi:10.5070/m4111041558 fatcat:tzn3lg4lr5g6tdm47yafcqd3su