AUTHOR=Cai Lei , Liu Xiaoqian TITLE=Identifying Big Five personality traits based on facial behavior analysis JOURNAL=Frontiers in Public Health VOLUME=10 YEAR=2022 URL=https://www.frontiersin.org/journals/public-health/articles/10.3389/fpubh.2022.1001828 DOI=10.3389/fpubh.2022.1001828 ISSN=2296-2565 ABSTRACT=

The personality assessment is in high demand in various fields and is becoming increasingly more important in practice. In recent years, with the rapid development of machine learning technology, the integration research of machine learning and psychology has become a new trend. In addition, the technology of automatic personality identification based on facial analysis has become the most advanced research direction in large-scale personality identification technology. This study proposes a method to automatically identify the Big Five personality traits by analyzing the facial movement in ordinary videos. In this study, we collected a total of 82 sample data. First, through the correlation analysis between facial features and personality scores, we found that the points from the right jawline to the chin contour showed a significant negative correlation with agreeableness. Simultaneously, we found that the movements of the left cheek's outer contour points in the high openness group were significantly higher than those in the low openness group. This study used a variety of machine learning algorithms to build the identification model on 70 key points of the face. Among them, the CatBoost regression algorithm has the best performance in the five dimensions, and the correlation coefficients between the model prediction results and the scale evaluation results are about medium correlation (0.37–0.42). Simultaneously, we executed the Split-Half reliability test, and the results showed that the reliability of the experimental method reached a high-reliability standard (0.75–0.96). The experimental results further verify the feasibility and effectiveness of the automatic assessment method of Big Five personality traits based on individual facial video analysis.