Innovative Method for Face Emotion Recognition using Hybrid Deep Neural Networks

Automatic facial expression recognition opens up new possibilities for human-computer interaction. To a large extent, face detection is the crux of many systems that aim to recognize people. Face detection is a challenging task because of the wide variety of human faces in terms of color, stance, expression, location, and orientation. It is therefore convenient to distinguish different facial expressions using various modelling techniques. Recognizing human gestures in images is a fascinating area of research because of the variety of expressions that may be conveyed by the position and movement of the eyes, mouth, nose, etc. Any facial expression-recognition neural network can also be used to classify faces, identify faces, and match faces to tokens. The FER system is entirely automated, with separate modules for finding faces, finding faces, extracting features, choosing the best features, and classifying data. After a face has been detected using the AdaBoost algorithm, the frame with the highest intensity of emotion is extracted using the inter-frame mutual information criterion. Several techniques, such as the Gabor filter and GAN-CNN model training are applied to the selected frames to extract distinctive characteristics. According to the findings of this study, a GAN-CNN based method can be used to identify the face expressions. As compared to the CNN and SVM models, this approach achieves higher accuracy (about 99.7 % ).