Deep Learning Based on Face Emotion Recognition using an Artificial Neural Network

Facial Emotion Recognition has become a critical research domain in Artificial Intelligence due to its vital applications across multiple fields, including security and healthcare. This study aims to overcome prevailing limitations of current methodologies, which primarily rely on static image frames and neutral facial expressions, hindering optimal emotion recognition rates. Leveraging advanced deep learning techniques, this research focused on the identification of seven human emotions - happiness, anger, disgust, fear, sadness, surprise, and neutrality - using both static and dynamic images from CK+ and FER datasets. The research experimented with various models such as Lightweight MobileNet and Artificial Neural Networks, aiming for enhanced accuracy in real-time emotion recognition applications. A novel visualization technique was developed to pinpoint the crucial facial regions integral for detecting distinct emotions by analyzing classifier outputs, revealing that different emotions are sensitive to different facial areas. The results emphasize the value of specific facial regions in conveying and recognizing emotions and illustrate the substantial promise of deep learning methodologies in advancing Facial Emotion Recognition technology.