Scholarlink

<p>This research aims to classify affective states of emotion based on facial landmarks using machine learning. It utilizes the ASCERTAIN dataset to capture facial landmark trajectory signals, which track the movement patterns of specific facial regions, such as the lips, nose, and eye corners, over time. By analyzing these trajectories, the study seeks to develop a robust emotion estimation system. Three machine learning algorithms were evaluated: Logistic Regression (LR), K-Nearest Neighbors (KNN), and Linear Discriminant Analysis (LDA). Among these, Logistic Regression showed superior performance, achieving 98.95% accuracy and an F1-score of 0.99 for both arousal and valence classifications. The effectiveness of the proposed system was assessed using multiple evaluation metrics, including Cohen’s Kappa, Precision, Recall, F1-score, Accuracy, and ROC-AUC (Receiver Operating Characteristic - Area Under the Curve). Findings indicate that the LR model outperformed state-of-the-art methods, suggesting that this approach could be a valuable tool for automated emotion recognition and decision-making based on variations in emotional states.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine Learning-Driven Emotion Recognition Through Facial Landmark Analysis

  • 期刊论文

  • 发表日期：

    2025年1月29日

查看全文
  • Akhilesh Kumar,
  • Awadhesh Kumar,
  • Sumit Gupta

摘要

This research aims to classify affective states of emotion based on facial landmarks using machine learning. It utilizes the ASCERTAIN dataset to capture facial landmark trajectory signals, which track the movement patterns of specific facial regions, such as the lips, nose, and eye corners, over time. By analyzing these trajectories, the study seeks to develop a robust emotion estimation system. Three machine learning algorithms were evaluated: Logistic Regression (LR), K-Nearest Neighbors (KNN), and Linear Discriminant Analysis (LDA). Among these, Logistic Regression showed superior performance, achieving 98.95% accuracy and an F1-score of 0.99 for both arousal and valence classifications. The effectiveness of the proposed system was assessed using multiple evaluation metrics, including Cohen’s Kappa, Precision, Recall, F1-score, Accuracy, and ROC-AUC (Receiver Operating Characteristic - Area Under the Curve). Findings indicate that the LR model outperformed state-of-the-art methods, suggesting that this approach could be a valuable tool for automated emotion recognition and decision-making based on variations in emotional states.