NEURAL NETWORK ASSESSMENT OF FACIAL EMOTIONAL EXPRESSION

面部情绪表达的神经网络评估

• 批准号: 3503375
• 负责人: R. RICHARD AVENT
• 金额: $ 4.85万
• 依托单位: AMERICAN RESEARCH CORP OF VIRGINIA
• 依托单位国家: 美国
• 财政年份: 1992
• 资助国家: 美国
• 起止时间: 1992-05-01 至 1992-10-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/3503375
• 关键词: biomedical equipment development; computational neuroscience; computer graphics /printing; computer human interaction; cues; diagnosis design /evaluation; emotions; eye; face expression; mathematical model; mental disorder diagnosis; mouth; videotape /videodisc

The National Institute of Mental Health has identified the need for methods of assessing emotional behavior to assist the tracking of normal and deviant developmental processes. In response to this need, American Research Corporation of Virginia (ARCOVA) proposes the development of a neural network-based machine vision system for evaluation of emotional state from facial expression. Research comparing the relative contribution of expressive channels in humans has shown that observers use information from facial cues more than any other source to determine an individual's emotional frame of mind. Specific aims of the research include extraction of edge features from black and white video imagery of human faces, selection of optimal neural network paradigms for extraction and analysis of facial features (such as eyes, eye brows and mouth), training and testing of the selected neural networks, and design of an interactive user interface. The significance of this effort is the innovative use of neural networks for facial feature extraction and analysis. Since neural networks have been shown to be more effective at pattern recognition tasks than traditional methods in similar applications, ARCOVA will demonstrate a proof-of-concept system for automated measurement of human emotion which is superior to present techniques.

美国国家心理健康研究所已经确定了对方法的需求 评估情绪行为以帮助跟踪正常和 偏差的发展过程。 为了应对这种需求，美国 弗吉尼亚州研究公司（Arcova）提出了 基于神经网络的机器视觉系统，用于评估情绪 面部表达状态。 比较相对贡献的研究 人类表达渠道的表明，观察者使用信息 从面部提示比任何其他来源都要确定一个人的 情感思维框架。 研究的具体目的包括提取 黑色和白色视频图像的边缘功能， 选择最佳神经网络范例用于提取和分析 面部特征（例如眼睛，眼睛眉毛和嘴），训练和 测试选定的神经网络以及交互式用户的设计 界面。 这项工作的意义是神经的创新使用 面部特征提取和分析的网络。 由于神经网络 已证明比模式识别任务更有效 在类似应用中的传统方法，Arcova将证明 概念验证系统，用于自动测量人类情绪 优于当前技术。