Facial expression classification using deep learning

Facial expression recognition has grown in importance in scientific study, is being seen as a key component of human interfaces, and is crucial for the establishment of any system that connects humans and machines. Because of its applications in emotion, analysis, medicine, psychology, and robotics that enable intelligent human-computer interaction, facial expression recognition has been one of the fastest-growing fields. Two techniques for facial expression recognition of six fundamental prototype expressions—smile, pleased, furious, disgusted, sad, surprised, and afraid—have been described in this thesis. These techniques rely on the Facial Action Coding System (FACS), and the photos utilized were acquired from the Cohn-Kanade and CK+ databases. Using a combination of various transforms, the first method will try to extract features using the wavelet transform and the Discrete Cosine Transform (DCT). Based on the features gathered from the first method, patterns will be clustered using the Kohonen Self Organizing Map (SOM), and recognition will be done using Euclidean distance. The results demonstrate that this method achieves a 95.3% recognition ratio and has extremely good accuracy in facial expression recognition.