| 其他摘要 | Personality refers to an individual's inherent tendency and psychological characteristics to others, things, and his own behavior in social adaptation. Personality affects various aspects of people such as personal consumption habits, performance ability, interpersonal communication, mental health, and even political stance. Personality theory is also widely used in various fields and industries. For example, in recruitment, accurate assessment of personality can help to fully understand the personality characteristics of candidates and assist in predicting their job matching ability. The study of personality has always been a very important subject, and personality assessment techniques are an indispensable method for studying personality. Most of the traditional personality assessment methods are mainly in the form of self-report(such as scales). Although it has a high evaluation accuracy, it still has shortcomings in terms of measurement efficiency.
In recent years, with the rapid development of computing technology, the research on the cross-border integration of artificial intelligence and psychology has provided a new idea and method for psychological research. Because of the popularity of social media platforms nowadays, there is a large amount of user public data(such as text, pictures or videos) in the network. It has brought rich data resources to psychological research; at the same time, some studies have shown that user behavior can reflect the psychological characteristics of individuals. Consequently, this paper attempts to use individual videos to study the Big Five personality recognition method based on facial motion analysis. This method does not depend on the autonomous participation of users, and the data sources used for identification are rich. It has the advantages of high efficiency in large一scale measurement, non-invasiveness, and ecologization. It makes up for the disadvantages of traditional personality assessment technology in these aspects.
This paper contains the following three studies:
Study 1:Correlation analysis of the Big Five personality and individual facial changes. By calculating the correlation between facial key point movement changes and the scores of the five dimensions of the Big Five, It was found that the intensity of the change of the facial key points from the contour of the lower half of the right face to the contour of the chin showed a significant negative correlation with agreeableness; The T-test of facial key point movement changes between the high and low groups of dimensions revealed that the left cheek contour of the subjects in the high group was significantly higher than that in the low group in the openness dimension. The results obtained in this study verifies the close correlation between facial activities and personality from the data level, and also shows that the automatic recognition technology of Big Five personality based on facial video analysis is feasible.
Study 2: Research on the construction technology of automatic recognition model of Big Five personality based on facial video analysis. This paper attempts to use a variety of machine learning algorithms to build an automatic recognition model of the Big Five based on the analysis of the movement of 70 key points of the face, and evaluate the effectiveness of the model. The study found that the calibration validity of the personality recognition model based on the CatBoost regression algorithm was 0.37 to 0.42, and the split-half reliability was 0.75 to 0.96. The experimental results further verify the feasibility and effectiveness of the big five personality automatic recognition technology based on facial video analysis.
Study 3:Research on the effect of individual facial video duration on the effect of the Big Five personality recognition model. This paper uses 5 seconds as an incremental interval, slices the original data according to different time lengths, and deeply explores the influence of data volume on the prediction effect of the model. The study found that when the video duration was 45 seconds, the model showed a high recognition rate in all five dimensions. The three dimensions of extraversion, neuroticism and openness also showed higher recognition rates when the video duration was shorter (10 seconds, 15 seconds). Simultaneously, the recognition effect of the five dimensions of personality increased from 0.37 to 0.42 in Study 2 to 0.478 to 0.529 based on 45 seconds of video data. The research results provide data support for the application research of automatic recognition technology of Big Five personality based on facial video analysis.
Through the above three studies, this paper has carried out in-depth exploration on the automatic recognition technology of Big Five personality characteristics based on facial video analysis, and proposed a new efficient and feasible personality assessment model. |
修改评论