基于面部肌电的微表情表达特性和识别机制研究
文献类型：学位论文
| 作者 | 鲁绍愿 |
| 答辩日期 | 2024-06 |
| 文献子类 | 硕士 |
| 授予单位 | 中国科学院大学 |
| 授予地点 | 中国科学院心理研究所 |
| 其他责任者 | 王甦菁 |
| 关键词 | 微表情 面部肌电 情绪识别 微表情数据库 |
| 学位名称 | 理学硕士 |
| 学位专业 | 基础心理学 |
| 其他题名 | Study of the expression characteristics and recognition mechanism of micro-expressions based on facial electromyography |
| 中文摘要 | Micro-expressions (MEs) are facial actions that leak out when people try to hide their genuine emotions. These MEs can be used as behavioral cues in various fields, such as clinical diagnosis, security screening, and criminal interrogation. MEs are facial movements that are difficult to suppress in response to external stimuli and have the characteristics of short duration, low movement intensity, and local appearance. However, previous studies on the expressive characteristics of MEs have relied heavily on subjective manual coding, lacking objective and quantifiable metrics. This subjective encoding method leads to inconsistent definitions of MEs in different studies, affecting in-depth analysis and recognition. At the same time, studying the expression rules of MEs in various situations is crucial. The recognition and application of MEs will be promoted by exploring the differences in facial muscle movements between camouflage and real reactions. In addition, the emotion recognition mechanism of MEs has not been determined, and the lack of verification of facial imitation behavior in objective indicators has limited the accuracy of MEs recognition. Therefore, this study aims to reveal the nature of MEs more comprehensively by deeply exploring the facial muscle movement patterns during MEs expression and recognition. The results help us better understand MEs and provide solid theoretical support and practical guidance for improving the ability to recognize MEs. In this paper, we delve into the expressive characteristics and recognition mechanisms of MEs through two primary studies. Study 1 explored the characteristics of MEs under the suppression of real emotions and the camouflage of real emotions by experiment 1 and experiment 2, respectively. In experiment 1, the percentage of facial electromyography (EMG) in maximum voluntary contraction (MVC), MVC%, was used to quantify the intensity of MEs. Through interval estimation, the intensity of ME ranges from 7% to 9.2% MVC, and the duration varies from 0.307 seconds to 0.327 seconds. The experimental results showed that the muscle activity of MEs in the upper face was greater than that in the lower face, which proved the difference between MEs and macro-expression (MaEs) in different muscle channels. In experiment 2, the degree of differentiation of facial muscle movement to camouflage response and real response was investigated by combining the MEs and facial EMG. The results showed that facial muscle activity was smaller in camouflage response, revealing the rule of facial muscle movement during camouflage. In other words, individuals suppressed and controlled facial expressions during camouflage. Study 2 focuses on the recognition mechanism of MEs and explores the influence of emotion types on emotion recognition through behavioral and physiological data. In Experiment 3, the expressions in the EMG-MEs database constructed in this paper are recognized. The results of behavioral indicators and facial EMG signal analysis show that compared with MaEs, individuals have lower accuracy, longer reaction time, and smaller facial muscle activity when recognizing MEs. There are also differences in behavioral and physiological indicators when recognizing different emotions. The experiment also verifies the facial feedback hypothesis during ME recognition. It discusses the difference in facial imitation behavior between upper and lower faces, which proves that the feedback of the upper face is stronger. Overall, this paper verified the expression characteristics of MEs through facial EMG signals and found the corresponding facial imitation behavior in ME recognition. The results reveal the correlation between the expression and recognition of MEs and promote the recognition by studying the expression characteristics of MEs. In addition, this paper proposes an ME coding method based on facial EMG signal-assisted artificial, which provides convenient coding tools and technical support for ME research and related applications. |
| 英文摘要 | 微表情是人们试图隐藏真实情绪时的面部表情，可以为临床诊断、安全检查 和刑事审讯等各个领域提供必要的行为线索。微表情是在外部刺激下难以压抑的 面部动作，具有持续时间短、运动强度低和局部出现等特性。然而，以往对微表 情的表达特性的研究主要依靠人工编码，缺乏客观量化的指标。这种主观编码方 式导致不同研究中的微表情定义不一致，从而影响微表情的深入分析和识别。同 时，微表情在不同情境下的表达规律研究至关重要，通过探究伪装和真实反应下 面部肌肉运动差异，将促进微表情的识别和应用。此外，微表情的情绪识别机制 也尚未确定，缺乏在客观指标上对面部模仿行为的验证，微表情识别的准确性受 到了一定的限制。因此，本研究旨在通过深入探究微表情表达和识别过程中的面 部肌肉运动模式，以期更全面地揭示微表情的本质。研究结果不仅有助于我们更 好地理解微表情，更将为提升微表情识别能力提供坚实的理论支持和实践指导。 本文通过两个研究分别探讨了微表情的表达特性和识别机制。研究一分别通 过实验 1 和实验 2 探究压抑真实情绪和伪装真实情绪下微表情的表达特性。实验1 使用面部肌电占最大随意收缩（maximum voluntary contraction，MVC）的百 分比 MVC%量化微表情的运动强度，通过区间估计得到微表情运动强度在 7%~ 9.2%MVC 之间、持续时间在 0.307 秒~0.327 秒之间，实验结果发现微表情在上 面部的肌肉活动大于下面部，证明了微表情和宏表情在不同肌肉通道中运动的差 异。实验 2 则通过结合微表情和面部肌电信号，考察面部肌肉运动对伪装反应与 真实反应的区分度，结果发现了伪装反应中面部肌肉活动更小，个体伪装时存在 对面部表情的压抑和控制，进一步揭示伪装时面部肌肉运动的规律。 研究二关注微表情的识别机制，通过行为和生理数据探讨情绪类型在微表情 情绪识别中的影响作用。实验 3 对本文构建的肌电微表情数据库中表情进行识别 任务，对行为指标和面部肌电信号分析结果表明，相比宏表情，个体在识别微表 情时准确率更低、反应时更长、面部肌肉活动更小，识别不同情绪时也存在行为 和生理指标上的差异。实验还验证了微表情识别的面部反馈假说，并探讨不同情 绪的面部模仿行为在上、下面部的差异性，证明了上面部的反馈更强烈。 总体上，本研究通过面部肌电信号验证了微表情表达特性，且在微表情识别 中也发现了对应的面部模仿行为，揭示了微表情的表达和识别过程中的关联性， 通过研究微表情的表达特性促进了微表情的识别。不仅如此，本文还提出了基于 面部肌电信号辅助人工的微表情编码方式，为微表情研究及其相关应用提供了便 利的编码工具和技术支持。 |
| 语种 | 中文 |
| 源URL | [http://ir.psych.ac.cn/handle/311026/47963]  |
| 专题 | 心理研究所_认知与发展心理学研究室 |
| 推荐引用方式 GB/T 7714 | 鲁绍愿. 基于面部肌电的微表情表达特性和识别机制研究[D]. 中国科学院心理研究所. 中国科学院大学. 2024. |
入库方式： OAI收割
来源：心理研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。