基于独立成分分析的面孔识别异族效应神经机制的研究

发布时间：2018-01-04 23:16

本文关键词：基于独立成分分析的面孔识别异族效应神经机制的研究　出处：《北京交通大学》2017年硕士论文　论文类型：学位论文

【摘要】：异族效应(Other-race effect,ORE)是一种面孔识别偏差现象,具体表现为识别本族面孔的要比识别外族面孔容易得多。异族效应形成机制的研究已经成为了认知神经科学等领域的研究热点之一。随着功能神经影像技术的迅速发展,利用功能磁共振(functional magnetic resonance imaging,fMRI)和脑电图(electroencephalographic,EEG)对异族效应的神经机制进行研究已经成为现实。已有研究表明本族面孔显著地激活了右侧梭状回人脸区(fusiform face area,FFA),同时,与本族面孔相比,外族人脸在杏仁核上的激活强度更大。随着对异族效应神经机制研究的深入,有学者发现成年人在识别本族和外族面孔时,在与面孔加工有关的皮层区域之间形成的有效网络存在显著差异。独立成分分析(independent components analysis,ICA)作为一种数据驱动方法,在近几年的功能神经影像学研究中占有越来越重要的地位。为了揭示人们识别不同种族面孔时大脑激活模式的差异,本文利用ICA数据处理方法和fNIRS成像技术,对异族效应的神经机制进行研究,具体研究过程如下:第一,采用fNIRS成像技术采集95名受试者的fNIRS数据,并对采集到的原始空间数据和光学数据进行预处理,得到配准后的空间数据和氧合血红蛋白浓度变化数据,为后续数据处理做准备。第二,对含氧血红蛋白浓度变化数据进行组ICA,并根据数据特点和实验设计研究出时间串联算法,从而实现在相同条件下分析不同种族面孔引起的大脑神经机制的差异。第三,对得到的本族面孔刺激对应的独立成分和外族面孔刺激对应的独立成分分别与本实验理想的血液动力学响应做相关分析,得到两组相关系数;然后对这两组相关系数做差异性分析,判断该成分是否有显著差异。得到显著差异的成分后,找到其对应的脑区分布及空间坐标。最后,总结本文研究发现:外族面孔刺激激活程度显著大于本族面孔刺激激活程度的脑区有:中央后回、距状裂周围皮层、眶内额上回、楔叶、枕上回。外族面孔刺激激活程度显著小于本族面孔刺激激活程度的脑区有:背外侧额上回、三角部额下回。本研究提出了 fNIRS数据的组ICA分析算法,通过对近红外脑功能数据的处理,揭示了异族效应的大脑激活模式,为以后的相关研究奠定了基础。本研究得到的结果与前人使用fMRI、EEG等方法得到的结果具有一致性。另外,本研究通过面孔识别的时空处理模式揭示了种族间差异。本研究不仅能帮助人们理解面孔加工对大脑视觉皮层功能组织的影响,也能为改善机器学习人脸识别算法提供仿生学依据。
[Abstract]:Heterosexual effect is a phenomenon of face recognition bias. It is much easier to recognize the faces of the native people than to recognize the faces of the foreigners. The study of the formation mechanism of the alien effect has become one of the hotspots in the field of cognitive neuroscience and so on. With the development of the functional neural imaging technology, it has become a hot topic in the field of cognitive neuroscience. Rapid development. Functional magnetic resonance imaging was used. FMRI) and electroencephalographic (EEG). The study of the neural mechanism of alien effect has become a reality. It has been shown that the facial region of the right fusiform face area is significantly activated by our faces. FFAA, at the same time, compared with the native faces, the Alien faces in the amygdala activation intensity is greater. With the in-depth study of the heterogeneous effect neural mechanism, some scholars have found that adults recognize their own and foreign faces. There were significant differences in the effective networks formed between the cortical regions associated with face processing. Independent component Analysis (ICA) was independent components analysis. ICA), as a data-driven method, plays an increasingly important role in the study of functional neuroimaging in recent years. In order to reveal the differences in brain activation patterns in recognition of different racial faces. In this paper, we use ICA data processing method and fNIRS imaging technology to study the neural mechanism of alien effect. The specific research process is as follows: first. The fNIRS data of 95 subjects were collected by fNIRS imaging technology, and the raw spatial data and optical data were preprocessed. The spatial data and oxygen hemoglobin concentration change data after registration were obtained to prepare for the subsequent data processing. Secondly, the changes of oxygen hemoglobin concentration data were carried out group ICA. And according to the characteristics of the data and experimental design to develop a time-series algorithm, so that under the same conditions to analyze the different ethnic faces caused by the difference of the neural mechanism. Third. The correlation analysis was made between the corresponding independent components of face stimulation and the ideal hemodynamic response of this experiment, and the correlation coefficients between the two groups were obtained. Then the difference between the two groups of correlation coefficients is analyzed to determine whether there is a significant difference between the two groups. After obtaining the significantly different components, the corresponding brain area distribution and spatial coordinates are found. Finally. Summing up the results of this study, we found that the activation degree of facial stimuli in foreign faces was significantly greater than that in their own faces: posterior central gyrus, paracral cortex, superior orbital frontal gyrus and cuneiform lobe. Supraoccipital gyrus. The brain regions with lower activation degree of foreign faces than their faces were dorsolateral superior frontal gyrus. In this study, we proposed a group ICA analysis algorithm for fNIRS data. By processing the near infrared brain function data, we revealed the brain activation mode of alien effect. The results obtained in this study are consistent with those obtained by previous methods such as fMRI EEG. This study reveals inter-ethnic differences through the spatio-temporal processing of face recognition. This study not only helps people understand the effects of face processing on the functional organization of the visual cortex of the brain. It can also provide bionic basis for improving the face recognition algorithm of machine learning.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;R445.2

【参考文献】