Performance Analysis of Classification Models in Multiclass Facial Expression Recognition Based on Eigenface Features

Abstract

Facial Expression Recognition (FER) is currently widely explored by researchers in the field of Computer Vision. The application of Machine Learning and Deep Learning methods is useful in developing an intelligent system that is accurate in recognizing facial expressions such as emotions. This is inseparable from the type of dataset and classification method used which certainly affects the desired results. To choose the right method, it is necessary to compare the performance of these methods. This study focuses on comparing the performance results of four classification methods namely, Convolutional Neural Network (CNN), Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Naïve Bayes Classifier (NBC) on a multiclass dataset for seven classes of facial emotion labels based on Eigenface feature selection uses the Personal Component Analysis (PCA) algorithm. The test parameters used to perform method comparisons are accuracy, recall, precision, f1-score, as well as the Receiving Operating Characteristic (ROC) and Area Under Curve (AUC) curves. The results of the analysis state that the SVM method has the highest accuracy value, while other methods show varying performance based on recall, precision, f1-score, and ROC and AUC analysis. This research was conducted on the FER 2013 dataset which showed that the classification method tested had quite good performance according to the test parameters.