Performance Comparison of Different Convolutional Neural Network Approaches for Facial Expression Recognition

Facial expression is a non-verbal way of communication to express the human state of mind using facial muscles. Happiness, sadness, anger, surprise, disgust, fear, and neutral expressions are widely used in the field of medical rehabilitation, sentiment analysis, counseling, and so on inspiring researchers to develop effective models to classify the expressions effectively. LeNet5, AlexNet, Deep Model, Shallow Model, Deep CNN Model are some commonly used models that have been developed to recognize facial expressions using machine learning and deep learning. In this research, a new convolutional neural network model has been proposed and compared with the existing models. The FER-2013 dataset has been used to evaluate the performance using different metrics to find the efficiency of the models. The proposed model provides comparatively better accuracy than most of the existing models, which is 64.4%. Keywords: Facial Expression, Image Classification, Convolutional Neural Network, Deep Learning, Non-verbal communication.