Automatic Emotion Detection from Facial Features Extraction and Machine Learning Models
Raghav Shankar, Bhupendra Singh Kirar, Pratham Upadhyay, Shree Govind Jee Mishra · 2025
In the domain of Interactive Systems (IS), machine learning (ML) emerges as a profound, user-friendly, and accurate approach to enhancing emotional intelligence in digital systems. This research work focuses on facial emotion detection, an important but often-neglected aspect of affective computing. Models used for the assessment include Multinomial Naïve Bayes (MNB), Decision Trees (DT), K-Nearest Neighbours (KNN), Random Forest (RF), Support Vector Machines (SVMs), XG Boost (XGB), and Logistic Regression (LR). By employing the FER2013 dataset and MediaPipe’s (MP’s) facial recognition technology, this research extracts 478 3D facial landmarks to test. The best Acc in emotion detection is found using LR with an accuracy of 65.57% followed by XGB with an accuracy of 61.24%. This study thus will help as a practical application toward making improvements in IS in matters regarding health, marketing, or even security.