Abstract

Facial expression recognition is a critical area of computer vision research in the current era. In order to improve the performance of the existing system, many researchers have used handcrafted features and machine learning (ML) classification methods. However, machine learning performs well on large data sets, whereas the handcrafted system performs well on datasets captured under predefined conditions. The ML method will not perform well on small datasets like CK+ and JAFFE due to overfitting. We propose a genetic algorithm (GA) optimization architecture based on fuzzy C-means clustering (GAO-FCM) to address these problems. To identify and recognize facial expressions quickly, accurately, and reliably, the proposed GAO-FCM organizes several characteristics of individual facial expressions. In the Japanese Female Facial Expression (JAFFE) dataset as well as in the Extended Cohn-Kanade (CK+) dataset, face detection and pre-processing techniques are explored in order to achieve the best accuracy of 99.23% for CK+ and 96.31 % for JAFFE datasets.