Design and Implementation of a Learning Emotion Recognition System

The purpose of this study was to construct an instant and short-term immediate learning emotion prediction model based on the facial expression that automatically evaluates learners’ learning activities. First, the “Learning Emotional Facial Image Database” was built as the training data to build the emotion classification model for the teaching and learning activity. The “Learning Emotional Facial Image Database” now has 3211 facial images and labels learning emotions. Then facial features were captured by the image processing technique and compute the action units. Two kinds of input features, feature value and action units, are used to build prediction model. Finally, three supervised machine learning algorithms were used: Support Vector Machine, Multi-layer Perceptron, and Random Forest to link the relation between facial features and learning emotions. In this study, the Random Forest algorithm has the highest learning emotion recognition rate, 88.20%, with the feature value as the input vector.