Abstract:
Emotions play a significant role in information processing, attitude formation, and decision-making in real-world situations. Facial Expression Recognition (FER) systems have been widely studied to accurately and stably detect emotions. However, due to the diversity of human faces and image variations, developing such systems remains challenging. Previous research has focused on single-network or ensemble models, with ensemble models achieving higher accuracy but requiring more models, datasets, and computational complexity. This work addresses the challenge of handling multiple emotions displayed on a person's face, where single-label sentiment analysis can be noisy. To tackle this, we propose a standalone Convolutional Neural Network (CNN) model implemented in a real-time Intelligent System for Sentiment Recognition. This system combines face detection, sentiment classification, and real-time probabilistic labeling from a webcam feed. The proposed model outperforms existing standalone approaches, including VGG16, VGG19, EfficientNetB7, with good accuracy when applied to the noisy FER2013 dataset. We also offer complete support for final year engineering projects, including project centers in Bangalore, IEEE projects, IEEE final year projects, BTech final year projects, IEEE CSE projects, and final year IEEE projects for CSE.