Classification technique for real-time emotion detection using machine learning models

Abstrak

This study aimed to explore models to identify a human by using face recognition techniques. Data were collected from Cohn-Kanade dataset composed of 398 photos having face emotion labeled with eight emotions (i.e., neutral, angry, disgusted, fearful, happy, sad, and surprised). Multi-layer perceptron (MLP), support vector machine (SVM), and random forest were used in model accuracy comparisons. Model validation and evaluation were performed using Python programming. The results on F1 scores for each class in the dataset revealed that predictive classifiers do not perform well for some classes. The support vector machine (RBF kernel) and random forest showed the highest accuracies in both datasets. The results could be used to extract and identify emotional expressions from the Cohn-Kanade dataset. Furthermore, the approach could be applied in other contexts to enhance monitoring activities or facial assessments.