Identification of Mood in Early Childhood with Face Recognition

Abstract

There are many factors that affect the mood of children during the teaching and learning process in class. Some of the causes of children having a bad mood are that the child is still sleepy when going to school, getting unpleasant treatment, an uncomfortable environment at home, and less attention in the classroom due to the teacher’s limitations in monitoring students during learning. At the same time, a good mood is imperative to maintain concentration and enthusiasm in the learning process in class. This research proposes the Convolutional Neural Network (CNN) method, which is also one of the methods in Deep Learning. CNN can receive input in the form of images which machines can then use to learn to recognize images, determine what objects or aspects are contained in the image, and so on. The moods to be identified consist of angry, neutral, sad, and happy, with a total of 26,215 FER2013 data. Several metrics are used to evaluate the model’s performance, such as Accuracy, Precision 79%, Recall 50%, and F1-score 61%. CNN performance in the classroom for early childhood mood identification achieved 84% accuracy. The system created in this study works by detecting human objects through videos taken using a camera, after which the face recognition stage is carried out then feature extraction is carried out, then classified according to the available data model. The last stage is to save the results of mood identification in the form of a report so that it can be accessed at any time.

Access to Document

Other files and links

• Link to publication in Scopus

Fingerprint