Neural representations underlying visual perception of objects and faces

视觉对象是如何在人脑中表现和识别的？我的研究项目旨在回答这个广泛的问题。在过去的40年里，心理物理学、电生理学和最近的神经成像研究已经破译了视觉处理早期阶段的许多方面。我们现在知道，视觉感知始于将图像分解为最基本的属性，如空间频率和方向。对于能够识别复杂视觉物体的视觉处理的后期阶段，我们所知甚少。我们建议使用在低级视觉研究中有用的方法的方法学等效方法来检查涉及不变视觉对象识别的高级过程。识别一个对象而忽略与其身份无关的变化的能力被称为不变性。例如，人类观察者在识别字母时是尺度不变的。因此，他们能够阅读各种视觉尺寸的文本，从非常大的（如广告牌）到非常小的（如精细印刷）。没有不变性，就不可能有真正的识别。人类大脑完成不变识别的精妙能力与最先进的人工识别系统的能力形成鲜明对比。因此，发现如何在人脑中实现不变和鲁棒的目标识别是神经科学中最令人兴奋的开放问题之一。这一领域的突破将对与医学界合作开发可能破坏面部处理的疾病（如面孔失认症（脸盲症）和自闭症）的康复计划以及与工程界合作实现强大的自动识别系统产生深远的影响。

How are visual objects represented and recognized in the human brain? My research program is targeted to answer this broad question. In the last 40 years, psychophysics, electrophysiology and, more recently, neural imaging studies have deciphered many aspects of the early stages of visual processing. We now know that visual perception starts with a decomposition of images into their most basic properties such as spatial frequency and orientation. A lot less is known about the later stages of visual processing that enable recognition complex visual objects. We propose to use methodological equivalents of approaches that have been instrumental in studies of lower-level vision to examine higher-level processes involving invariant visual object recognition.The ability to recognize an object while ignoring variations that do not pertain to its identity is referred to as invariance. For example, human observers are scale-invariant in recognizing letters. Thus they are able to read text at a wide range of visual sizes, from very large, e.g., on a billboard, to very small, e.g., fine-print. Without invariance, there can be no real recognition. The exquisite ability of the human brain to accomplish invariant recognition stands in stark contrast to the capabilities of state-of-the-art artificial recognition systems. It is therefore one of the most exciting open problems in neuroscience to discover how invariant and robust object recognition is achieved in the human brain. Breakthroughs in this area will have far-reaching implications for collaborative efforts with the medical community to develop rehabilitation programs for diseases that may disrupt face processing such as prosopagnosia (face blindness) and autism, and with the engineering community towards realizing robust automated recognition systems.