A Deep Learning Model for Automatic Recognition of Facial Expressions Using Haar Cascade Images

Facial emotional recognition has become most popular with in the industry because of the high demand of deep learning automatic models. Image processing, video gaming, welfare, and security are a few of the sectors where it might be used. In this paper, we have presented an approach centered on a convolutional neural network (CNN). First, the facial expression image is captured using Haar Cascade, and the convolution technique has been used to extract the edge of each layer of the image. The obtained edge information is placed on each feature image to retain the texture image’s edge structure information. The maximum pooling approach is then used to minimize the complexity of the retrieved implicit features. Finally, a CNN classifier consists of multiple convolutional layers, two max-pooling layers, and two fully connected layers make up to recognize the expression. It will predict if the individual is happy, sad, surprised, and so on, and it will display the result continually. This approach attained an accuracy of 60.6% after tweaking the different hyperparameters. Also, the model’s reliability is assessed with previous traditional methods that use CNN, RNN, and SVM on the same dataset (FER-2013).