CRI: Neuromorphic VLSI Modelling of Attention-Based Visual Search

为了避免信息过载，到目前为止，输入到我们感觉器官的大部分信息必须被大脑丢弃，只有经过精心挑选的部分才被允许影响行为并达到长期记忆。这个过程被称为选择性注意，它是心理学、神经生物学和计算科学研究的焦点。在这个项目中，我们将设计和测试模拟VLSI芯片，这些芯片实现了我们之前开发的注意力选择过程的神经生物学合理模型。该系统将在标准的视觉搜索范式上进行测试，就像在猴子和人类受试者的心理学实验中经常使用的那样。构建这样的“神经形态”设备有望极大地促进对生物感知的理解，因为一个工作和健壮系统的设计者必须克服，从而理解生物系统面临的许多相同的功能限制（噪音，有限的连接可能性等）。此外，机器视觉也将从这项工作中受益匪浅，因为一旦计算机视觉（与生物视觉相比）仍然处于原始阶段，技术图像理解系统将面临与生物有机体相同的信息过载问题。——恩斯特·尼伯博士，克里格精神/大脑研究所助理教授，约翰·霍普金斯大学神经科学教授niebur@jhu.edu 3400 N. Charles Street (410)516-8643, -8640 (secr), -8648 (fax) Baltimore, MD 21218

IBN: 9634357 PI: Niebur In order to avoid information overload, by far the largest part of the input to our sensory organs has to be discarded by the brain, and only carefully selected parts are allowed to influence behavior and to reach long-term memory. This process is called selective attention and it is the focus of intense research in psychology, neurobiology and the computational sciences. In this project, we will design and test analog VLSI chips which implement neurobiologically plausible models of the attentional selection process that we have developed previously. The system will be tested on standard visual search paradigms as used routinely in psychological experiments with monkey and human subjects. Building such "neuromorphic" devices is expected to greatly benefit the understanding of biological perception since the designer of a working and robust system has to overcome, and thus to understand, many of the same functional constraints biological systems are faced with (noise, limited connection possibilities etc). Also machine vision should benefit significantly from this work since technical image understanding systems will be confronted with the same problem of information overload as biological organisms are, once the still primitive stage of computer vision (compared to biological vision) will be surpassed. -- Ernst Niebur, PhD Krieger Mind/Brain Institute Asst. Prof. of Neuroscience Johns Hopkins University niebur@jhu.edu 3400 N. Charles Street (410)516-8643, -8640 (secr), -8648 (fax) Baltimore, MD 21218