Face Expressions Recognition Based on Image Processing using Convolutional Neural Network for Human Computer Interface

Abstract

Feelings, communicated in various structures, are a specific relational correspondence. The emotional state helps in an independent direction, helps inventiveness, and oversees human comprehension and human-machine correspondence. In a couple of years, the need to recognize an individual’s feelings is expanding, and interest in human feeling acknowledgment in different fields has been expanding, however not restricted to human-PC interfaces and metropolitan sound discernment. This study proposed a new self-constructed architecture named NNN-Net to compare it with famous AlexNet architecture. We use the same parameters, input size, optimizer, and learning rate in both architectures to find the best combinations that will perform the best result. The dataset that we use is CK+48, one of the famous datasets to study face expression recognition. We also augmented the dataset to increase the number of images for each class and balance the dataset. Furthermore, we found that our NNN-Net shows better results with an exact combination of parameters. The best accuracy result is 98.63%. at last, this study can be helpful as a foundation to classify students’ expression using an online meeting platform.