Classification of face expressions by deep learning method

Emotion is a psychological state of being in the human mind. Various studies in various fields support various viewpoints on the process of emotion formation. Emotion, according to some philosophers, is the outcome of changes in personal conditions or the environment. Some biologists, however, believe that the neurological and hormonal systems are principally responsible for the genesis of emotion. We were motivated to create a facial expression recognition approach based on convolutional neural networks (ConvNets). Convolutional neural network models are used to predict a facial expression label belonging to one of the following categories: neural, happiness, anger, surprise, sadness, or disgust. The label for the facial expression must correspond to one of the following categories: neural, happiness, anger, surprise, sadness, or disgust. The image serves as an input to the system. Consequently, these models undergo rigorous testing. The capture and classification module makes it possible to know the emotional impact that a computer system can cause on the user, and with this information to take measures to improve user interaction. Thus, we conclude that it is possible to use CNN to classify expressions applied to usability tests and can even provide real-time responses to those who perform the test. The main contribution of this thesis is that we modify the CNN algorithm with linear activation functions and replace the built-in SoftMax layer of the CNN with other classifiers to increase the accuracy of the detection, the modified CNN algorithm not only extracts the features using the Conv-2 layer but also uses feature templates and Haar like features to increase the speed of the workflow.