基于三维数字图像相关方法的面部表情变形测量研究
On the Facial Expression Deformation Measurement Based on 3-Dimentional Digital Image Correlation Method

目前关于人脸面部表情的相关研究正在逐渐应用到各领域，而其研究大多为基于数据库的定性分析。本文将三维数字图像相关方法用于人脸面部表情研究。首先，针对人脸面部图像的具体特点，在深入研究方法原理的基础上，提出以参数优化和消除刚体位移来提高实验测量精度。在此基础上，对普通表情和微表情状态下面部肌肉变形进行精确的计算和定量的研究，分析特定表情的形成原因。实验采用自传式回忆的方法唤起被测试者的基本情绪，随之进行模仿来诱发面部普通表情；采用指导性表达抑制的方法诱发被测试者的微表情。通过对特定表情状态下的肌肉运动进行全场和局部的计算，获得精确的三维位移场和位移矢量场。结果表明，面部微表情状态与普通表情状态的运动规律基本一致，只是位移幅度存在显著差异。实验测量结果与实际情况基本吻合，符合常规认知。不同的是本文将特定表情状态下面部肌肉的运动规律从以往的感性认知上升到精确计算和定量分析的水平，为面部表情的自动识别及形成机理的深入研究提供了良好的基础。
At present, relevant research about facial expression is gradually applied to various fields, while this research is mostly qualitative analysis based on database. In this paper, 3-dimentional digital image correlation method is used to study facial expression. First of all, according to specific features of face image, based on deep study of method principle, two means including parameter optimization and elimination of rigid body displacement were put forward to improve the accuracy of experimental measurement. On this basis, precise calculation and quantitative study of facial muscle deformation was carried under conditions of normal expression and micro expression state, the cause of specific facial expression was analyzed. In experiment, the basic emotion of the subject being tested was evoked by using the method of autobiographical memory, followed by imitation to induce facial expressions. The micro expression of subject being tested was induced by guiding expression inhibition method. Accurate 3-dimensional displacement field and displacement vector field were obtained by calculating the whole field and local muscle motion in specific expression state. Results show that the facial micro expression state was consistent with that of normal expression state, but there is significant difference in displacement amplitude. Experimental results are in good agreement with actual situation, which is consistent with conventional cognition. But the difference is that the understanding of muscle movement of facial specific expression was improved, from previous perceptual cognition to level of accurate calculation and quantitative analysis, which provides the foundation for automatic identification of facial micro expression and deeper study of micro expression formation mechanism