Cognitive Systems Foresight: Human and computer face recognition from video sequences

这个为期三年的项目是计算机视觉技术识别人脸的专家和人类解读人脸的心理学专家之间的合作。该项目的目标是建立一个人类面部处理的生成性计算模型，能够合成和解释人脸图像序列。这个项目将成为建立一个计算模型的核心部分，该模型将解释熟悉和陌生的运动人脸识别现象。这将对这一领域的心理学知识有所借鉴和贡献。它将与运动对识别的影响及其与面部学习过程的相互作用具有特别的相关性。众所周知，运动有两种不同的心理效应，第一种是表征增强，它与面孔的三维静态表征的构建有关，并与相对陌生的面孔有关。另一种是动态特征‘’，是个体特有的运动模式。这与更熟悉的面孔有关；以前的相关计算工作集中在表征增强问题上。本研究的特定目标是构建一个能够应对熟悉度变化的人类面部加工的生成性计算模型。为了实现这一点，我们需要(A)建立一个能够从多个帧中学习人脸的静态特征的系统，以及(B)建立对行为及其变化的描述。此外，在人类中，我们需要(C)研究面部运动学习的过程和(D)研究动态签名的表征。为了支持这项研究，将收集一个适合人类和计算机实验的跨文化运动人脸数据库。计算模型将被用来产生心理学实验的数据，以调查运动对识别的影响。这些实验的结果将为自动人脸识别算法的发展提供参考。

This 3-year project is a collaboration between experts in computer vision techniques for recognising faces and experts in the psychology of how humans interpret faces. The aim is to produce a generative computational model of human facial processing, capable of both synthesizing and interpreting sequences of face images.This project will form the central part of an effort to build a computational model which will account for both familiar and unfamiliar moving face-recognition phenomena. This will draw from and contribute to the psychological knowledge in this area. It will have particular relevance to the effects of movement on recognition and its interaction with the process of facial learning. It is known that there are two separate psychological effects of movement, firstly representational enhancement'', which relates to the construction of 3-D static representations of the face and is associated with relatively unfamiliar faces. The other, a dynamic signature'', is the characteristic motion pattern of an individual. This is associated with more familiar faces; previous relevant computational work has concentrated upon representational enhancement issues.The particular aim of this research will be to construct a generative computational model of human facial processing, able to cope with variations in familiarity. To achieve this, we need to (a) build a system capable of learning the static characteristics of faces from multiple frames and (b) build a description of behaviour and its variation. In addition, in humans, we need to (c) investigate the process of facial movement learning and (d) investigate the representation of dynamic signatures. To support this research, a cross-cultural database of moving faces suitable for human and computer experiments will be collected. The computational model will be used to generate data for psychological experiments investigating the effects of movement on recognition. The results of these experiments will inform development of automatic face interpretation algorithms.