First congenital prosopagnosia case reported in China

Congenital prosopagnosia (CP) refers to the deficiency in recognizing faces without reportable brain damage. In this study we reported a CP patient (X.L.), the first CP case reported in China, and tested her with a battery of face recognition and visual cognition tasks to explore what aspects of face processing are impaired in CP individuals. Our results show that X.L. possesses a normal ability to detect the presence of faces yet her ability to discriminate between individual faces is significantly impaired compared to normal controls. In contrast, her ability to detect nonface objects is greatly preserved, suggesting that she suffers from a selective deficiency in face recognition. Further face tests reveal that she has intact feature-based analysis ability while configural and holistic processing are both impaired. We further evaluate her general capacity of holistic processing, and the results showed that her impairment in holistic processing is not restricted to faces, but to a more general level.     

First congenital prosopagnosia case reported in China

Congenital prosopagnosia (CP) refers to the deficiency in recognizing faces without reportable brain damage. In this study we reported a CP patient (X.L.), the first CP case reported in China, and tested her with a battery of face recognition and visual cognition tasks to explore what aspects of face processing are impaired in CP individuals. Our results show that X.L. possesses a normal ability to detect the presence of faces yet her ability to discriminate between individual faces is significantly impaired compared to normal controls. In contrast, her ability to detect nonface objects is greatly preserved, suggesting that she suffers from a selective deficiency in face recognition. Further face tests reveal that she has intact feature-based analysis ability while configural and holistic processing are both impaired. We further evaluate her general capacity of holistic processing, and the results showed that her impairment in holistic processing is not restricted to faces, but to a more general level.     

Neuropsychological impairment: the disturbed effect of self-processing in patients with major depressive disorder

Neuropsychological impairment has long been established as a fundamental characteristic of depression, but a generally accepted, specific pattern of neuropsycho- logical impairment has not been summarized. In this study, we examined the classic neuropsychological paradigm of self-face recognition, to explore whether the self was impaired in major depressive disorder （MDD）. Eighteen MDD patients and 20 healthy subjects were recruited to participate in this study. By using a face morphing tech- nique, we measured the size of processing bias in MDD patients during different face discrimination tasks relative to controls. Results of analysis of variance （ANOVA） showed a significant main effect of Group （F（1, 36） = 7.388, P = 0.01）. Subsequent independent t-tests further revealed that self bias （t = 2.636, P = 0.012） and self-recognition bias（t = 2.190, P = 0.035） observed in self-famous task and self-stranger task respectively for patients were significantly greater than that for controls. Both self-processing and self- recognition were impaired in patients with MDD, indicating that MDD individuals might exist level of self-abnormali- ties. These findings provide a new perspective for further study on the etiological and pathological mechanisms of MDD.     

基于小波分析的人脸识别算法

提出一种利用小波分析提取人脸特征的方法。对人脸图像做小波分解,用网格划分其子图像,在各子块上提取统计特征,用其训练多分类支持向量机模型,最后用训练好的支持向量机进行人脸识别。选择ORL人脸库对该算法进行实验,与PCA算法的比较结果证明了该算法在识别性能方面的优越性。     

彩色序列图像人脸检测系统

介绍了一个完备、新颖的彩色序列图像人脸检测系统。该系统利用颜色信息和运动信息发现人脸可能出现的区域，然后利用人脸的几何特征进行验证，并对正确检测到的人脸提取其面部特征进行精确定位。实验结果表明，该系统具有较强的抗光照强弱变化、人脸侧转、旋转等干扰因素的能力，并能根据输入图像的质量选择输出结果的定位精度，在提高定位精度的同时确保了系统的可信度。     

基于DTW的LPM在面部遮挡人脸识别中的应用

针对面部遮挡人脸识别问题中遮挡源变幻莫测、遮挡位置、大小及形状未知而导致从人脸图像中分割遮挡区域困难的问题,提出了基于动态时间规整(DTW)的局部分块匹配(LPM)算法。首先,将人脸图像划分成若干大小相等且互不重叠的局部小块;然后,借助于光栅扫描顺序将各个小块按照前额、眼睛、鼻子、嘴巴和下巴的顺序连接成一个单一序列;最后,计算查询人脸与注册人之间图像到类的距离,并利用动态时间规整的设计思想寻找查询序列与所有注册序列之间的最佳对齐方式。在两个公开人脸数据库FRGC2.0、AR及一个户外人脸数据库上的实验验证了所提算法的有效性及可靠性,实验结果表明,相比其它几种较为先进的人脸识别算法,所提算法取得了更高的识别率,此外,所提算法无需任何训练过程,计算成本低,更适合应用于现实面部遮挡人脸识别中。     

Recollection-like memory retrieval in rats is dependent on the hippocampus

Recognition memory may be supported by two independent types of retrieval, conscious recollection of a specific experience and a sense of familiarity gained from previous exposure to particular stimuli. In humans, signal detection techniques have been used to distinguish recollection and familiarity, respectively, in asymmetrical and curvilinear components of their receiver operating characteristic (ROC) curves, standard curves that represent item recognition across different levels of confidence or bias. To determine whether animals also employ multiple processes in recognition memory and to explore the anatomical basis of this distinction, we adapted these techniques to examine odour recognition memory in rats. Their ROC curve had asymmetrical and curvilinear components, indicating the existence of both recollection and familiarity in rats. Furthermore, following selective damage to the hippocampus the ROC curve became entirely symmetrical and remained curvilinear, supporting the view that the hippocampus specifically mediates the capacity for recollection.     

A study on neural mechanism of face processing based on fMRI

Recently, there were debates about the specificity of lateral middle fuisform in face processing. The debates focused on whether these areas were specialized in face processing or involved in processing of visual expertise and categorization at individual level. The present study aims to investigate the neural mechanism of face processing, using Chinese characters as comparison stimuli. Chinese characters are greatly similar to faces on a variety of dimensions, among which the most significant one is that both faces and Chinese characters not only are extremely familiar to literate Chinese adults but also are processed at individual level. In the present study, faces and Chinese characters activated bilateral middle fusiform with great correlation. Greater activities were observed in the right fusiform face area (FFA) for faces than for Chinese characters. These results demonstrate that FFA is specialized in face processing per se rather than the processing of visual expertise and categorization at individual level.     

基于广义典型相关分析融合和鲁棒概率协同表示的人脸指纹多模态识别

针对单模态生物特征识别容易受自身条件和环境变化的影响,鉴于人脸识别和指纹识别已经在生物识别系统中得到了广泛应用,提出了二者特征信息融合的多模态生物特征识别方法。该方法首先对人脸、指纹图像进行预处理,并对这两种模态均提取LBP和Gabor特征,然后将广义典型相关分析方法分别引入到人脸多特征融合和指纹多特征融合中,应用分块对角矩阵组合上述融合的人脸特征和指纹特征,最后用鲁棒概率协同表示分类器进行分类。在两个多模态数据库上的实验结果表明:与人脸或指纹单模态生物特征识别相比,基于人脸指纹的多模态生物特征识别具有更高的识别率和更好的稳定性;所提出的基于广义典型相关分析的特征融合方法优于传统的融合方法。     

A study on neural mechanism of face processing based on fMRI

Recently, there were debates about the specificity of lateral middle fusiform in face processing. The debates focused on whether these areas were specialized in face processing or involved in processing of visual expertise and categorization at individual level. The present study aims to investigate the neural mechanism of face processing, using Chinese characters as comparison stimuli. Chinese characters are greatly similar to faces on a variety of dimensions, among which the most significant one is that both faces and Chinese characters not only are extremely familiar to literate Chinese adults but also are processed at individual level. In the present study, faces and Chinese characters activated bilateral middle fusiform with great correlation. Greater activities were observed in the right fusiform face area （FFA） for faces than for Chinese characters. These results demonstrate that FFA is specialized in face processing per se rather than the processing of visual expertise and categorization at individual level.     

Norm-based face encoding by single neurons in the monkey inferotemporal cortex

The rich and immediate perception of a familiar face, including its identity, expression and even intent, is one of the most impressive shared faculties of human and non-human primate brains. Many visually responsive neurons in the inferotemporal cortex of macaque monkeys respond selectively to faces, sometimes to only one or a few individuals, while showing little sensitivity to scale and other details of the retinal image. Here we show that face-responsive neurons in the macaque monkey anterior inferotemporal cortex are tuned to a fundamental dimension of face perception. Using a norm-based caricaturization framework previously developed for human psychophysics, we varied the identity information present in photo-realistic human faces, and found that neurons of the anterior inferotemporal cortex were most often tuned around the average, identity-ambiguous face. These observations are consistent with face-selective responses in this area being shaped by a figural comparison, reflecting structural differences between an incoming face and an internal reference or norm. As such, these findings link the tuning of neurons in the inferotemporal cortex to psychological models of face identity perception.     

采用自适应加权扩展LBP的单样本人脸识别

提出了一种采用自适应加权扩展LBP(AWELBP,adaptively weighted extended local binarypattern)的单样本人脸描述方法,首先对单样本的人脸图像进行多尺度分块,对子块的图像进行扩展均匀LBP算子运算,同时同步生成图像局部熵图谱(LEM,local entropy map),计算每一子块对整体人脸图像纹理描述的贡献度图谱,根据贡献度图谱对每个子块的LBP直方图进行自适应加权,最后将各子块的LBP直方图进行连接形成人脸特征。本算法在ORL、Yale、Yale B人脸库上对部分遮挡、表情变化、光照变化等环境进行测试,并与传统算法以及与多种LBP改进算法进行比较,结果表明该算法对部分遮挡、表情变化和光照等环境下单样本人脸描述具有较好的效果。     

计算机人脸识别技术综述

概述了计算机人脸识别技术的历史及发展现状,讨论了在计算机人脸识别领域占有主流地位的Eigen脸方法(主元素分析方法)、最佳鉴别矢量集法(基于Fisher线性判别准则方法和基于Foley-Sammon变换方法)、Bayesian脸方法、基于傅里叶不变特征法和弹性图匹配法。指出了各个研究方向人脸识别方法, 给出了计算机人脸识别性能评价指标,包括识别率、计算时间、数据存储量和可扩展性等。根据这些性能评价指标,对当前的各种计算机人脸识别技术进行分析评价。讨论结果表明,基于Fisher线性判别准则的最佳鉴别矢量集法,Bayesian脸方法和基于傅里叶不变特征法都有较好的性能,具有一定的应用前景.     

基于特征脸的人脸识别系统

提出了一种快速有效的人脸识别系统．针对特征脸方法只能识别标准正面人脸的局限性，设计了前端处理模块，首先计算待识别图像中人脸的倾斜角度，进行相应角度的旋转，然后剪切出人脸图像，进行尺寸调整，经过此处理后再进行特征脸识别．利用ORL人脸库和自建人脸库进行仿真实验，实验结果表明，本识别系统速度快，误识率低，具有实用性．     

基于小波子带图像的人脸光照归一化方法

数字人像识别技术的研究成果有广泛的应用前景,但光照问题是影响人脸识别技术关键的因素之一。对未知光源参数且具有左右光照不均匀的正脸图像,根据小波不同子带图像上的频率特点,提出了一种小波子带图像上的人脸光照归一化方法。在Yale B人脸库上的实验表明,新方法能较有效的解决人脸识别中的左右不均匀光照恢复问题,使人脸识别率有显著提高。     

基于深度卷积神经网络与中心损失的人脸识别

传统人脸识别方法手工设计特征过程复杂、识别率较低,对于开集人脸识别通用深度学习分类模型特征判别能力较弱。针对这两方面的不足,提出了一种以分类损失与中心损失相结合作为模型训练监督信号的深度卷积神经网络。首先,利用构建的应用场景数据集优调从公共数据集获得初始化参数的深度人脸识别模型,解决训练数据过小和数据分布差异问题,同时提高模型训练速度;然后,以传统损失函数和新的中心损失作为迁移学习过程中的监督信号,使得类内聚合、类间分散,提高模型输出人脸特征的判别能力;最后,对人脸特征进行主成分分析,进一步去除冗余特征,降低特征复杂度,提高人脸识别准确率。实验结果表明,与传统人脸识别算法相比该算法可以自动进行特征提取,并且相对于通用深度学习分类模型该算法通过度量学习使特征表示更具判别力。在自建测试集和LFW、YouTube Faces标准测试集上都取得了较高的识别率。     

基于卷积神经网络(CNN)和CUDA加速的实时视频人脸识别

为了兼顾视频人脸识别中识别准确率和实时性,提出了基于卷积神经网络(CNN)和CUDA加速的实时视频人脸识别方法。构建了一个6层结构的CNN人脸识别网络,在视频帧中通过Adaboost算法检测到的人脸输入所构建的CNN中进行视频人脸识别,结合CUDA并行计算架构,对算法进行加速。此外为了更适用于实际视频监控情况,通过对CNN网络结构末尾Softmax分类器的分类结果进行多级判决引入了开集人脸识别功能。从多个角度对该方法进行了实验验证,结果证明,此方法可满足识别准确率和实时性要求,同时对于视频中人脸姿态变化、光照变化、距离远近等都具有良好的鲁棒性。     

神经网络和遗传算法在人脸定位中的应用

人脸是十分常见的复杂模式,在复杂景物图片中自动找出人脸是很困难的,但是却有着重要意义,因为这同时也是自动人脸识别系统的前提．提出了一个基于多层感知器（ＭＬＰ）的用遗传算法实现搜索的人脸定位系统,讨论了系统建立的理论基础,即ＭＬＰ直接感知图象和用遗传算法进行快速搜索．实验结果表明,它尤其适合在复杂背景下的低分辨图象中进行操作,鲁棒性好、定位快、适应面宽．     

基于RBF网络的人脸识别和表情分析方法研究

为解决人脸识别过程中出现的无法有效区分多姿态人脸的问题,进一步提高人脸表情识别率。本文在分析现有人脸表情识别方法的基础上,提出新的识别技术,即采用基于径向基函数(RBF)神经网络的方法,首先对图像皮肤和非皮肤像素进行分离,把人脸区域从检测到的皮肤区域中提取出来,然后以人脸表情数据库JAFFE为测试数据库,对人脸图像进行Gabor小波变换(GWT)和离散余弦变换(DCT),最后将该算法用于径向基函数神经网络的训练过程,建立相应优化模型,并将其应用到人脸表情的识别中,研究结果表明,具有收敛速度快、识别率高等优点,比文献中的方法提高了3%和8%的识别率。     

A mechanism for impaired fear recognition after amygdala damage

Ten years ago, we reported that SM, a patient with rare bilateral amygdala damage, showed an intriguing impairment in her ability to recognize fear from facial expressions. Since then, the importance of the amygdala in processing information about facial emotions has been borne out by a number of lesion and functional imaging studies. Yet the mechanism by which amygdala damage compromises fear recognition has not been identified. Returning to patient SM, we now show that her impairment stems from an inability to make normal use of information from the eye region of faces when judging emotions, a defect we trace to a lack of spontaneous fixations on the eyes during free viewing of faces. Although SM fails to look normally at the eye region in all facial expressions, her selective impairment in recognizing fear is explained by the fact that the eyes are the most important feature for identifying this emotion. Notably, SM's recognition of fearful faces became entirely normal when she was instructed explicitly to look at the eyes. This finding provides a mechanism to explain the amygdala's role in fear recognition, and points to new approaches for the possible rehabilitation of patients with defective emotion perception.