Exploring the Uncanny Valley

日本著名机器人学家森正弘在1970年探索神秘谷时，提出了“神秘谷”曲线来描述人类对非人类代理人的情感反应。他的假设是，当特工变得更像人类时，观察者的熟悉度并不会像人们直觉的那样增加，而是落入“诡异之谷”，当特工密切但不完美地模仿人类时。随着计算机图形学的进步，人们可以越来越逼真地呈现形式和运动，这个诡异的山谷已经成为娱乐业的一个高风险问题。诡异谷假说还提出了一个基本的科学问题：当人们看待人类运动时，他们感知到什么？这些感知如何受到现实的、接近现实的或讽刺的人类运动的影响？随着功能磁共振成像(FMRI)测量和分析技术的准确性不断提高，有了一种新的工具来回答这些问题。PI们正在使用功能磁共振成像、眼球跟踪和传统的感知指标来探索诡异山谷的存在和原因。运动捕捉数据(包括皮肤和肌肉变形)、关键帧动画、高速视频和故事片中的片段被用来构建一组刺激来回答这些问题。这项研究的结果是对动画人体运动的感知有了更好的理解，这将为计算机图形学的实践者和研究人员提供信息，使他们能够专注于人类动画将产生最大影响的方面。这项研究对科学和社会都有广泛的影响。例如，这里进行的工作适用于对自闭症的科学理解，自闭症是一种毁灭性的神经发育障碍，会在多个领域影响社会功能，包括对他人行为和意图的感知。

Exploring the Uncanny ValleyIn 1970, an eminent Japanese roboticist, Masahiro Mori, proposed the "uncanny valley" curve to describe the emotional response of humans to nonhuman agents. His hypothesis was that as an agent is made more humanlike, the observer's familiarity does not increase as one would intuit, but falls into a ``valley of eeriness,'' when the agent closely yet imperfectly impersonates a human being. With progress in computer graphics allowing increasingly realistic rendering of forms and motion, the uncanny valley has become a high-stakes concern of the entertainment industry. The uncanny valley hypothesis also poses a fundamental scientific question: What do people perceive when they view human motion and how are those perceptions affected by realistic, nearly realistic, or caricatured human motion? With the growing accuracy of functional magnetic resonance imaging (fMRI) measurement and analysis techniques, there is a new tool with which to answer these questions. The PIs are using fMRI, eye tracking, and traditional perceptual metrics to explore the existence of and causes of the uncanny valley. Motion capture data (including skin and muscle deformation), keyframed animations, high speed video, and clips from feature films are used to construct a set of stimuli with which to answer these questions.The outcomes of this research is a greater understanding of the perception of animated human motion which will inform practitioners and researchers in computer graphics and allow them to focus on the aspects of human animation that will have the greatest impact. The research has broad impacts on science and society. For example, the work conducted here has applicability to the scientific understanding of autism, a devastating neurodevelopmental disorder that effects social functioning in multiple domains including the perception of other people's actions and intentions.