Facial Expression Classification for User Experience Testing Using K-Nearest Neighbor

One of the important steps of testing out applications such as video game is getting the information regarding user experience. Emotion from the testers while playing can be used as a parameter of the user experience. Emotions such as anger, happiness, sadness, or surprise can be seen from changes in facial expressions. These emotional parameters can be used as feedback for satisfaction or deficiency in the video game so that developers can increase the improvement of the final product of the game. This project discusses the human facial expression classification system to test video games using the K-Nearest Neighbor (KNN) classification method and using the Indonesia Mixed Emotion Dataset (IMED) as training data and trial data. In this system, there are several processes, namely preprocessing, feature extraction, and classification. Finally, this system issues a classification of facial expressions detected in the form of chart that can be used in user experience testing. The result of this research is that the K-Nearest Neighbor (KNN) algorithm results in training model accuracy rate of 98.24% and real-time human facial expressions with up to 56% accuracy.