【作者】 王玲；

【导师】 尧德中；

【作者基本信息】 电子科技大学 ， 生物医学工程， 2008， 博士

【摘要】 对深度运动(motion-in-depth,MID)过程的研究是视运动知觉研究中非常重要的课题。对自然界的动物来讲,判断接近的物体对自己是否具有威胁性及精确判断出飞来物体何时会到达(即计算碰撞时间TTC,time-to-collision)是攸关观察者生命与生活的关键问题。尽管在动物电生理方面有很多好的工作,但目前未见带有明显生理意义且符合神经元特性的成熟脑机制模型的报道。对人类来说,研究“人脑如何处理运动过程中的视觉信息”以及“怎样通知运动系统以躲避即将到来的危险”是非常有意义的工作。现代先进的神经成像技术的应用,如正电子发射断层成像(positron emission computerized topography,PET)、功能磁共振成像(functional magnetic resonance imaging,fMRI)、脑电图(electroencephalogram,EEG)和脑磁图(magnetoencephalography,MEG)等为人类的深度运动问题提供了一些解答,但仍有很多问题待解决。本论文围绕深度运动过程的视觉脑机制问题,从数学模型、脑电事件相关电位(event-related potential,ERP)时-空分析等几个方面进行了深入、细致地有创新性的研究。所作的主要工作与成果如下:1、动物检测碰撞过程的视觉脑机制研究用模型方法阐述了动物(鸽子)检测碰撞过程的脑机制,并用神经网络的方法进行仿真,得到了较好的结果。模型的简单性与动物的简单反应原则一致,且在生理性、稳健性、延展性、适用性等方面均有良好表现,证明在解释鸟类从初始的视网膜成像到大脑最终获得准确的碰撞时间(TTC)信息的整个视觉信息处理过程中,它可能是一种比较合理的视觉脑机制模型。2、人类知觉深度运动过程的视觉脑机制研究通过更接近人类视觉经验的实验范式,结合人的ERP与fMRI数据,系统地研究了人在知觉深度运动过程中的视觉脑机制问题。多种时-空分析方法的结果表明,人的深度运动感知与平面运动感知是有差异的,且人的深度运动ERP受运动方向、接近物体的大小和运动速度的强烈影响。深度运动的感知过程激活了三个主要的脑区,并呈现了五个主要的脑电时间成分,本文详细探讨了这些时-空动力学响应在深度运动感知中的意义与作用。在这些成果的基础上,我们提出了一个简单的多区域同步整合模型,以解释人类知觉深度运动的信息处理过程。3、大脑信息整合的研究从大脑信息整合的角度,探讨了动物与人在知觉深度运动过程中的视觉信息处理问题,渗透于以上两个有关的动物与人的视觉脑机制研究中,我们提出了动物检测碰撞过程的时-空整合模型和人知觉深度运动过程的多区域同步整合模型。这些模型都具有明确的生理意义,在解释知觉深度运动过程中的大脑视觉信息处理中显示了较好的效果,并可望用于人工智能机器人的视觉系统中,以仿生的方式达到深度运动视觉中的感知和碰撞避免。更多还原

【Abstract】 Motion-in-depth (MID) is an important issue in study of visual motion perception. For animal, it is a critic problem to observer’s survival and life in judging whether a looming object be dangerous and computing accurately when the looming object arrives, i.e. computing time-to-collision (TTC) exactly as soon as possible. Although there were many good works on animal electrophysiological studies, there was no reported mature brain mechanism model with significant physiological meanings and be consistent with the neurons characteristics. For human, to investigate how the brain processes the visual information of this motion and how it informs the motion system for escaping the impending danger object is also a meaningful project. Modern advanced neuroimaging techniques such as positron emission computerized topography (PET)、functional magnetic resonance imaging (fMRI)、electroencephalogram (EEG) and magnetoencephalography (MEG) provide some answers to questions in motion perception. But as a current important and hot issue, there are still many problems to be solved.This paper descibes deeply and detaily the novel studys on brain mechanism of MID course in many aspects in cognitive neural science, including mathematical model, event-related potential (ERP) spatio-temporal analysis and so on. Our main works and achievements are follows:1. The study on animal’s brain mechanism on detecting collision courseIt expatiate the animal’s (pigeon) brain mechanism on detecting collision course from model field simulated by neural network method with good results. The simple and convenient property of the model is consistent with animal’s simple response principal. And the good representations of the model on physiology, robustness, extensibility and applicability manifest it might be a very potential and reliable brain mechanism on explaining the whole visual information process on pigeon brain from initial retinal imaging to obtaining final accurate TTC information.2. The study on human’s brain mechanism on percepting MID courseThrough an experimental paradigm more close to human’s visual experience, combined with human’s ERP and fMRI data, we study the human’s brain mechanism on MID course systematically. Multiple spatial and temporal analysis methods are adopted to analyse the ERP signals. The results indicate that the perception of MID and motion-in-plane (MIP) is different, and the MID ERPs are strongly affected by the motion direction, impending object’s size and moving speed. The perception of MID activates three main brain areas and showes five main temporal components. This paper discusses the significances and roles of these spatio-temporal dynamic responses on the perception of MID in detail. Basing on these achievements, we propose a simple multi-areas synchronous integrated model on interpreting how the visual information processed in the perception of MID.3. The study on integration of brain informationFrom the viewpoint of integration of brain information, this paper explores the visual information process in MID course on animal and human by participating in the study of different animal’s and human’s brain mechanism to propose more rational brain mechanism solution. The proposed spatio-temporal integrated model on animal’s detecting collision course and the multi-areas synchronous integrated model on human’s percepting MID course have definite physiological meanings and might be very possible visual brain mechanisms on interpreting the visual information process in the perception of MID. And they might be expected to apply in visual system on artificial intelligent robotics to achieve the perception of MID and avoid collision by bionics manner.更多还原