基于动态特征的真伪笑容表达与识别

笑容是人类最普遍、最频繁的表情。人类进化出伪装笑容的能力,也拥有部分识别伪装的能力。在表情的表达与识别上,动态信息起着重要的作用。一方面,笑容表达的动态特征可能为区分真伪笑容提供重要的信息,所以我们拟借助近年发展的计算机视觉的特征提取技术,系统地量化分析真伪笑容的动态特征(时长、方向、速度、流畅性、运动对称性、不同部位同步性、头动模式等),考察笑容在不同伪装方式及不同情境下的区别与一致性,深入理解人类笑容表达的特点。另一方面,通过探索有效动态特征与正确识别率的关系,检验知觉-注意假说,了解真伪笑容的识别特点及研究识别机制。通过比较动态真伪笑容的表达特点与识别特点,进一步理解人类表情信号编码与解码之间的关系。

The expression and recognition of genuine and disguised smiles based on dynamic information

The smile is the most common and frequently expressed facial expression. Human beings have evolved ways to fake smiles, and developed the ability to detect the disguise. Dynamic information of facial movements has been shown to be essential in both expressing and recognition of facial expressions. The underlying dynamic information of smiles may be investigated to better differentiate between fake and genuine smiles. With the aids of feature extraction methods from Computer Vision, we investigate the dynamic features of smiles (such as the duration, direction, velocity, smoothness, dynamic symmetry, and synchronicity and head motion patterns). Furthermore, we investigate how various dynamic features of genuine and fake smiles may differ across different situations to further our understanding of the human smile. We also aim to investigate the underlying mechanism of individual's ability to distinguish between genuine and fake smiles. We examined whether paying attention to effective dynamic features would help improve recognition performance, and test the perception-attention hypothesis. By comparing the features of smiles and examining the underlying mechanism behind people’s ability to recognize the veracity of smiles. We can obtain a better understanding of the relationship between encoding and decoding of the facial expressions.