眼睛是判别人脸自发表情的重要线索，心理学及相关研究也表明眼睛与面部表情动作之间存在一定的协同机制。本项目通过视频诱发与访谈形式相结合来构建包含眼部信息的人脸自发表情视频数据库，在遵循眼球与眼周部件不同运动规律的前提下，建立眼部动作的语义特征描述标准（EPA）及其提取方法，以有效刻画眼部多种表情动作行为；引入公理化模糊集理论（AFS）挖掘自发表情中眼睛与面部其它部件之间协同机制，以语义规则集形式对自发表情动作构成规律进行知识表达；针对面部遮挡条件下表情理解问题，将上述语义规则转化为基于贝叶斯网的情感推理模型，通过融合眼部信息探索复杂场景中表情识别的新途径。本项目拟建立的EPA以及获取的眼睛与面部动作协同规则集，在丰富人脸自发表情识别的相关理论与方法手段的同时，还可促进信息科学与心理学等学科交叉，并在对于尊重多民族习惯下的表情理解与合成、安防、心理医疗等领域具有较好的应用前景。

Eye is a crucial clue for recognizing spontaneous expression. Relevant research works in psychology indicate that there exists a certain collaborative mechanism between the expression actions between eye region and face area. In this project, firstly, video and interview are both employed to induce someone’s true emotion, and thus to construct a spontaneous database including useful eye information. On the premise of obeying the action rules of eyeball and periocular respectively, the project plans to build a description criteria of semantic features to illustrated the action behavior in eye region, named Eyeball-Periocular Action (EPA); then, the extraction method of EPA is also studied. Axiomatic fuzzy set (AFS) theory is introduced to mine the collaborative mechanism between eye and other facial components in the process of spontaneous expression, which lays a solid foundation for knowledgeable representation of emotional action in the form of fuzzy rules. Aiming at recognizing the expression in a masked condition, we intend to convert above mentioned semantic rules into a Bayesian network model for reasoning one’s emotion, and explore a new approach of expression recognition in complex scenario by integrating eye information. The semantic description criteria of eye action and collaborative rules between eyes and other facial actions to be studied in this project will enrich the relevant theories and methods of spontaneous expression recognition; boost the interdisciplinary research between information science and psychology; be applied in expression synthesis, security and mental health.