FACIAL EXPRESSION ANALYSIS BY IMAGE PROCESSING

描述（改编自研究者摘要）：面部表情提供 敏感的情绪反应线索，并发挥关键作用， 人际行为的调节。基于人类观察者的方法 测量面部表情是劳动密集型的、定性的，并且难以 在不同实验室和不同时间进行标准化。为了使可行性更严格， 面部表情的定量测量在不同的应用中， 调查人员组成了一个跨学科的研究小组， 面部表情分析和计算机视觉。1995年8月，研究 该小组获得了计算机图像面部表情分析的初始资金 处理（NIMH#IRO1MH51435）。在NIMH的支持下，他们创建了一个大型的 方法开发的代表性数据库，并开发和验证了 跟踪数字化图像中的注视和面部特征的面部分析系统 婴儿、儿童和成人受试者的序列。在近正面视图中， 分析系统与手动FACS编码的同时有效性为16 44个FAGS动作单元和超过30种组合。为 竞争性更新，研究人员将（1）增加系统的鲁棒性， 头部方向和头部运动（2）增加可识别动作的数量 包括所有30个具有特定解剖学基础的单位，以及（3） 对研制的Face进行系统的、大规模的、全面的试验 使用多个数据库的分析系统。这些数据库涵盖了 不同的年龄和背景以及不同类型的情感诱导， 方向、头部运动、以像素为单位的面部大小以及语音的存在。 最初收集数据库是为了回答以下实质性问题： 情绪过程;它们代表了面部分析 系统在心理学研究和临床应用中会遇到。

DESCRIPTION (adapted from investigator's abstract): Facial expression provides sensitive cues about emotional response and plays a critical role in the regulation of interpersonal behavior. Human-observer based methods for measuring facial expression are labor intensive, qualitative, and difficult to standardize across laboratories and over time. To make feasible more rigorous, quantitative measurement of facial expression in diverse applications, the investigators formed an interdisciplinary research group that covers expertise in facial expression analysis and computer vision. In August 1995, the research group received initial funding for Facial Expression Analysis by Computer Image Processing (NIMH #IRO1MH51435). With NIMH support, they created a large representative database for method development and developed and validated a face analysis system that tracks gaze and facial features in digitized image sequences of infant, child, and adult subjects. In near frontal views, the Face Analysis System has achieved concurrent validity with manual FACS coding for 16 of 44 FAGS action units and more than 30 combinations in which they occur. For the competing renewal, the investigators will (1) Increase system robustness to head orientation and head motion (2) Increase the number of recognizable action units to include all 30 of those that have a specific anatomic basis, and (3) Conduct a systematic, large-scale, comprehensive test of the developed Face Analysis System by using multiple databases. These databases encompass subjects of various ages and backgrounds and varying types of emotion induction, head orientation, head motion, size of the face in pixels, and presence of speech. The databases were initially collected to answer substantive questions about emotion processes; they represent the types of data that the Face Analysis System will encounter in psychology research and clinical applications.