基于人脸特征与深度学习的学生人格特质分析

发布时间：2018-04-30 13:05

本文选题：大五人格特质 + 人脸特征点定位　；参考：《江西师范大学》2017年硕士论文

【摘要】：随着人工智能的高速发展,加速了全球经济一体化进程,竞争已经跨越国界,并对我国的经济和社会产生了巨大的影响。全球竞争的实质是人力资源的竞争,因此,社会对高素质人才的需求与日俱增。高校作为各类人才的重要输出地,所承载的使命也显得越来越重要。如何在面对种种压力和矛盾的背景下,培养一支高素质的就业队伍亦成为高校亟待解决的问题之一。基于深度学习的大五人格特质与人脸特征分析旨在运用心理学和人力资源管理等专业背景,采用人格测验的方法,结合人脸识别技术和机器学习方法,来挖掘大五人格特质与人脸特征之间存在的关系,作为教育管理者初步判别学生大五特质的辅助工具,并依据大五人格理论对学生进行人格特质分析提供进一步的参考。为解决上述问题,实现研究目标,主要工作如下:依据心理学人格测试理论,认为人的人格特质与人脸的某些特征存在潜在关联,并选用大五人格量表,采用统计分析法对大五人格特质进行测验,进而提出了一种基于人脸特征与深度学习的大五人格特质分析研究。大五人格量表能够得到相对真实可靠的人格特质,为实验的准确性奠定了基础。为了将被测试者的人格特质与脸部特征关联起来,需要对每个被测试者的头像进行人脸特征提取。利用主动形状模型(ASM)可忽略输入图像尺寸大小的优势,故采用改进后的ASM进行人脸特征点的提取,并选取特征之间的距离占整个脸部长或宽的比例、五官特征的长宽比等32个具有代表性的特征数据,为下一步深度学习提供训练依据。采用深度学习的方法,通过深度置信网络,对学生的大五人格特质与人脸面部特征进行训练和分类,从而找到大五人格特质与人脸特征之间的关系。利用深度学习无监督学习分类的特性,提出基于人脸特征的深度学习方法,并通过训练样本对其相关权值进行训练更新。将实验结果与传统的人格测验方式得到的结果进行对比分析,结果表明,该方法在时效性和精确度方面都具有更好的效能。
[Abstract]:With the rapid development of artificial intelligence, the process of global economic integration has been accelerated, competition has crossed national boundaries, and has had a great impact on the economy and society of our country. The essence of global competition is the competition of human resources. As an important export place of all kinds of talents, the mission carried by colleges and universities is becoming more and more important. Under the background of various pressures and contradictions, how to cultivate a high-quality employment team has also become one of the problems to be solved urgently in colleges and universities. The analysis of Big five personality traits and facial features based on deep learning aims at applying psychology and human resource management, personality test, face recognition and machine learning. To explore the relationship between Big five personality traits and facial features, as an auxiliary tool for education administrators to preliminarily judge students' Big five traits, and to provide further reference for students' personality trait analysis based on Big five personality theory. In order to solve the above problems and achieve the research goal, the main work is as follows: according to the psychological personality test theory, it is considered that there is a potential correlation between personality traits and some features of human face, and the Big five Personality scale is selected. This paper uses statistical analysis method to test Big five Personality traits, and then puts forward a kind of Big five Personality trait Analysis Research based on face feature and depth learning. The Big five Personality scale can obtain relatively true and reliable personality traits, which lays a foundation for the accuracy of the experiment. In order to correlate the personality traits of the subjects with the facial features, we need to extract the face features from the heads of each subject. The advantage of the size of input image can be ignored by using active shape model (ASM), so the improved ASM is used to extract facial feature points, and the distance between features is chosen to be the proportion of the length or width of the whole face. There are 32 representative feature data such as aspect ratio of features, which provides training basis for further study. By using the method of deep learning and using the deep confidence network, this paper trains and classifies the students' facial features and personality traits of Big five, and finds out the relationship between the personality traits of Big five and facial features. Based on the feature of unsupervised learning classification, a method of depth learning based on facial features is proposed, and its relevant weights are updated by training samples. The experimental results are compared with those obtained by the traditional personality test. The results show that the proposed method is more effective in terms of timeliness and accuracy.
【学位授予单位】：江西师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：B848

【参考文献】