Facial Expression Embrace: A CNN-SVM Approach for Multi-Class Emotional Classification

This study introduces a new CNN-SVM model that can categorize facial expressions into seven distinct emotional states: Happy, fear, angry and neutral became happy; disgust and surprise also changed to an active form. To evaluate the model in detail, we employ F1 Score (harmonic mean of precision and recall), support, and total accuracy. The model’s most impressive results are recall scores and accuracy rates indicating how effective it is at distinguishing emotions such as happiness or fear. Afraid performed well on all counts, achieving a score of 94.5 % according to the F1 Score which factors in both recall and accuracy. The wide applicability of the model to handling events within individual emotional categories is verified by its strong performance in both Happy and Anger (the two largest Support groupings). According to results with 0.97 for Happiness, 0.96 each for Fear on the one hand and Anger or Neutral moods on the other, while Disgust comes in at only three-tenths of that value, accuracy is usually very high indeed. Afraid’s 0.98 accuracy rate is remarkable, clearly setting it apart from the pack. The work has implications not only in the development of emotion recognition technology but also offers many insights into HCI and affective computing. The results made here provide a solid foundation for further progress in this field, which is somewhat obscure but has begun to make headway through the categorization of human emotions based on facial expressions.