Deep Learning-Based Automatic Face Expression Recognition Framework

One of the most important characteristics that might reveal a person's mental condition is their facial expression. Humans can communicate around 55% of their message nonverbally and about 45% audibly. One of the most challenging topics in computer science is now automatic facial expression detection. Some types of facial expression recognition (FER) are not simply for interpreting human behavior and keeping track of people's moods and mental states. It is also making inroads into industries including criminology, hologram, smart healthcare, security, and entertainment, as well as industries like robotics and entertainment and stress detection. Medical sciences highly value facial expressions, particularly for bipolar patients whose moods shift dramatically regularly. In this research, a Convolutional Neural Network (CNN) based method, and automation framework for facial detection, is suggested with 4 convolution layers and 2 concealed layers to increase accuracy. Various male and female faces with expressions including rage, dread, disgust, contempt, neutral, joyful, sad, and surprise are included in an enhanced Cohn-Kanade (CK+) dataset. The three stages of FD-CNN in this work are processing, extraction of features, and classification. With the help of this suggested procedure, a FER performance of 94% is achieved. K-fold cross-validation is used to validate the suggested algorithm. After validation, the computed specificity and sensitivity are 94.02% & 99.34%, respectively. Additionally, the model's f1 rating, recall, and precision - which are determined-are calculated and are correspondingly 84.07%, 78.22 %, and 94.09%.