Analysis of Facial Expression for Emotion Recognition using CNN-SVM

Facial expressions are a crucial source of information regarding an emotional state of the person. Accurately interpreting facial expressions remains a daunting challenge in the realm of interpersonal communication. The term “face mining” refers to the process of identifying and analyzing patterns of facial images within a particular set or collection of images. This study proposes a CNN-SVM hybrid model to recognize emotions, which achieved a high level of accuracy in categorizing emotions into eight classes. The performance of the CNN-SVM model was assessed using F1 score, precision, and recall metrics, for each emotion category. The top-performing emotion categories were Neutral, Sad, and Happy, with F1 scores of 94.33%, 95.2%, and 94.03%, respectively. The model's overall weighted average F1 score, precision, and recall were approximately 93.81%, indicating good classification performance while accounting for class imbalance. The micro average F1 score, precision, and recall were also 93.81%, demonstrating that the model accurately classified instances across all emotions. These findings suggest that the CNN-SVM hybrid model is a dependable and effective method for emotion classification, with potential applications in various domains such as healthcare, psychology, and social media analysis.