Application of a Voting-Based Ensemble Method for Recognizing Seven Basic Emotions in Real-Time Webcam Video Images

This study examines the automatic recognition of human emotions in real-time through facial expressions from webcams. Real-time emotion recognition is a crucial element in human-computer interaction and emotional computing. The study evaluates the effectiveness of various techniques in real-time facial emotion recognition using a custom CNN model, creating an ensemble with a voting mechanism, and integrating the system for real-time emotion recognition. The CNN model was trained on the FER2013 dataset, which consists of facial images labeled with different emotional states. It achieved a remarkable accuracy of 95%. In this study, we developed a dataset named ATS_FER2024, which consists of 184 images depicting seven distinct emotions. The tests conducted on this dataset yielded an accuracy rate of 89%. Despite its small size, the dataset's accuracy is noteworthy. The findings contribute to academic knowledge on developing emotion recognition systems, enhancing empathy, and creating context-sensitive interactions in real-world applications.