Comparative Analysis of CNN and SVM Models for Facial Emotion Recognition

The proposed study investigates the application of Convolutional Neural Networks (CNN) and Support Vector Machines (SVM) in the recognition and classification of five basic human emotions: happiness, sadness, anger, fear, and surprise. Affective computing with emotion recognition “a great element of it “looks forward to better human-machine interfaces that get machines to understand and echo human emotions via reading and responding to the emotional states of people. By retrieving the data from the vast array of facial expressions, the data was preprocessed to be able to meet the model's expectations. After that CNN and SVM models were designed and trained with the dataset of these some performance evaluation criteria i.e. accuracy, precision, recall, and F1 score were taken into account. The CNN model generated higher performance in the overall evaluation, particularly as indicated by accuracy and F1 score. This was mainly attributed to its optimally visual data feature extraction capability. The SVM model has also revealed a good performance, however, mainly in precision accuracy which represents its success in classification tasks. The undertaking is part of the new emerging field of affective computing, which provides comparative insights into the capabilities of CNN and SVM models in facial emotion recognition and thus a possible use of the strength of both in hybrid models. The remaining issues to be addressed are regarding the fusion models and the sensory data integration to achieve higher accuracies and reliabilities of emotion recognition.