Performance Comparison of Convolutional Neural Networks (CNN) and Support Vector Machine (SVM) Algorithms in Human Face Classification

Abstract

Facial expression recognition is crucial in fields like mental health monitoring and human-computer interaction. This study compares Convolutional Neural Networks (CNN) and Support Vector Machine (SVM) in classifying facial images into stress and non-stress categories. Using a preprocessed dataset of labeled facial expressions, CNN was employed for its strength in automatic spatial feature extraction, while SVM served as a traditional machine learning benchmark. Both models were trained and tested on the same dataset split. Results showed CNN outperformed SVM in all performance metrics: CNN achieved 88.94% accuracy, 94.42% precision, 93.25% recall, and an F1-score of 89.85%, while SVM recorded 76.53% accuracy, 77.14% precision, 85.72% recall, and an F1-score of 80.67%. Despite its lower performance, SVM had faster training and a simpler structure, making it suitable for resource-limited scenarios. The study emphasizes the superiority of deep learning for complex image classification tasks.Facial expression recognition is crucial in fields like mental health monitoring and human-computer interaction. This study compares Convolutional Neural Networks (CNN) and Support Vector Machine (SVM) in classifying facial images into stress and non-stress categories. Using a preprocessed dataset of labeled facial expressions, CNN was employed for its strength in automatic spatial feature extraction, while SVM served as a traditional machine learning benchmark. Both models were trained and tested on the same dataset split. Results showed CNN outperformed SVM in all performance metrics: CNN achieved 88.94% accuracy, 94.42% precision, 93.25% recall, and an F1-score of 89.85%, while SVM recorded 76.53% accuracy, 77.14% precision, 85.72% recall, and an F1-score of 80.67%. Despite its lower performance, SVM had faster training and a simpler structure, making it suitable for resource-limited scenarios. The study emphasizes the superiority of deep learning for complex image classification tasks.