MODELING FACE PERCEPTION

人脸感知建模

• 批准号: 2890940
• 负责人: GARRISON W COTTRELL
• 金额: $ 14.98万
• 依托单位: UNIVERSITY OF CALIFORNIA, SAN DIEGO
• 依托单位国家: 美国
• 财政年份: 1998
• 资助国家: 美国
• 起止时间: 1998-08-01 至 2001-07-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/2890940
• 关键词: behavior test; bias; brain injury; clinical research; computational neuroscience; emotions; face expression; human data; learning; model design /development; neural information processing; neuropsychology; perception; performance; photostimulus; psychological models

DESCRIPTION (Applicant's abstract): There has been a great deal of progress in understanding how complex objects, in particular, faces, are processed by humans. The aim of this proposal is to extend recent work on computational models of emotion and face processing to account for recent data concerning face processing in brain-damaged patients and normal controls. In all of these studies, the goal is to understand to what extent face processing must be considered "special", through modeling and experiment. Our specific goals are to: I) analyze the relationship between categorical perception of emotion in our models and in human subjects as elucidated by parallel experiments on the models and humans using identical stimuli; 2) understand further how specialized "modules" may arise for face, object and expression recognition through combinations of competitive learning mechanisms and innate biases; 3) understand the effects of various kinds of 'brain damage" on our model, as a route to understanding patient behavior. These goals will be addressed through careful modeling studies guided by the data. Our hypotheses will be tested by constructing neural network models that process the same face images used in human studies.

描述（申请人摘要）：已经取得了很大的进展 在理解复杂物体，特别是面部，是如何被 人类 这项建议的目的是扩大最近的工作计算 情绪和面部处理的模型，以解释最近的数据， 脑损伤患者和正常对照者的面孔加工。 在所有 这些研究的目标是了解面部处理在多大程度上 通过建模和实验，被认为是“特殊的”。 我们的具体 目标是：I）分析分类知觉之间的关系 情绪在我们的模型和人类的主题，阐明了平行 使用相同刺激的模型和人类实验; 2）理解 更进一步地说，面部、物体和表情的专门“模块”是如何产生的 通过竞争性学习机制的组合， 先天偏见; 3）了解各种“脑损伤”的影响 在我们的模型上，作为理解病人行为的途径。 这些目标 将通过由数据指导的仔细建模研究来解决。 我们 假设将通过构建神经网络模型进行测试， 与人类研究中使用的人脸图像相同。