面部表情分析进展和挑战

面部表情分析是计算机通过分析人脸信息尝试理解人类情感的一种技术,目前已成为计算机视觉领域的热点话题。其挑战在于数据标注困难、多人标签一致性差、自然环境下人脸姿态大以及遮挡等。为了推动面部表情分析发展,本文概述了面部表情分析的相关任务、进展、挑战和未来趋势。首先,简述了面部表情分析的几个常见任务、基本算法框架和数据库;其次,对人脸表情识别方法进行了综述,包括传统的特征设计方法以及深度学习方法;接着,对人脸表情识别存在的问题与挑战进行总结思考;最后,讨论了未来发展趋势。通过全面综述和讨论,总结以下观点：1）针对可靠人脸表情数据库规模小的问题,从人脸识别模型进行迁移学习以及利用无标签数据进行半监督学习是两个重要策略;2）受模糊表情、低质量图像以及标注者的主观性影响,非受控自然场景的人脸表情数据的标签库存在一定的不确定性,抑制这些因素可以使得深度网络学习真正的表情特征;3）针对人脸遮挡和大姿态问题,利用局部块进行融合的策略是一个有效的策略,另一个值得考虑的策略是先在大规模人脸识别数据库中学习一个对遮挡和姿态鲁棒的模型,再进行人脸表情识别迁移学习;4）由于基于深度学习的表情识别方法受很多超参数影响,导致当前人脸表情识别方法的可比性不强,不同的表情识别方法有必要在不同的简单基线方法上进行评测。目前,虽然非受控自然环境下的表情分析得到较快发展,但是上述问题和挑战仍然有待解决。人脸表情分析是一个比较实用的任务,未来发展除了要讨论方法的精度也要关注方法的耗时以及存储消耗,也可以考虑用非受控环境下高精度的人脸运动单元检测结果进行表情类别推断。     

结合像素模式和特征点模式的实时表情识别

目的 目前2D表情识别方法对于一些混淆性较高的表情识别率不高并且容易受到人脸姿态、光照变化的影响,利用RGBD摄像头Kinect获取人脸3D特征点数据,提出了一种结合像素2D特征和特征点3D特征的实时表情识别方法。方法 首先,利用3种经典的LBP（局部二值模式）、Gabor滤波器、HOG（方向梯度直方图）提取了人脸表情2D像素特征,由于2D像素特征对于人脸表情描述能力的局限性,进一步提取了人脸特征点之间的角度、距离、法向量3种3D表情特征,以对不同表情的变化情况进行更加细致地描述。为了提高算法对混淆性高的表情识别能力并增加鲁棒性,将2D像素特征和3D特征点特征分别训练了3组随机森林模型,通过对6组随机森林分类器的分类结果加权组合,得到最终的表情类别。结果 在3D表情数据集Face3D上验证算法对9种不同表情的识别效果,结果表明结合2D像素特征和3D特征点特征的方法有利于表情的识别,平均识别率达到了84.7%,高出近几年提出的最优方法4.5%,而且相比单独地2D、3D融合特征,平均识别率分别提高了3.0%和5.8%,同时对于混淆性较强的愤怒、悲伤、害怕等表情识别率均高于80%,实时性也达到了10~15帧/s。结论 该方法结合表情图像的2D像素特征和3D特征点特征,提高了算法对于人脸表情变化的描述能力,而且针对混淆性较强的表情分类,对多组随机森林分类器的分类结果加权平均,有效地降低了混淆性表情之间的干扰,提高了算法的鲁棒性。实验结果表明了该方法相比普通的2D特征、3D特征等对于表情的识别不仅具有一定的优越性,同时还能保证算法的实时性。     

多任务学习和对抗学习结合的自发与非自发表情识别

目的 如何提取与个体身份无关的面部特征以及建模面部行为的时空模式是自发与非自发表情识别的核心问题,然而现有的自发与非自发表情识别工作尚未同时兼顾两者。针对此,本文提出多任务学习和对抗学习结合的自发与非自发表情识别方法,通过多任务学习和对抗学习捕获面部行为的时空模式以及与学习身份无关的面部特征,实现有效的自发与非自发表情区分。方法 所提方法包括4部分：特征提取器、多任务学习器、身份判别器以及多任务判别器。特征提取器用来获取与自发和非自发表情相关的特征;身份判别器用来监督特征提取器学习到的特征,与身份标签无关;多任务学习器预测表情高峰帧相对于初始帧之间的特征点偏移量以及表情类别,并试图迷惑多任务判别器;多任务判别器辨别输入是真实的还是预测的人脸特征点偏移量与表情类别。通过多任务学习器和多任务判别器之间的对抗学习,捕获面部行为的时空模式。通过特征提取器、多任务学习器和身份判别器的协同学习,学习与面部行为有关而与个体身份无关的面部特征。结果 在MMI（M&M initiative）、NVIE（natural visible and infrared facial expression）和BioVid（biopotential and video）数据集上的实验结果表明本文方法可以学习出与个体身份相关性较低的特征,通过同时预测特征点偏移量和表情类别,有效捕获自发和非自发表情的时空模式,从而获得较好的自发与非自发表情识别效果。结论 实验表明本文所提出的基于对抗学习的网络不仅可以有效学习个体无关但表情相关的面部中特征,而且还可以捕捉面部行为中的空间模式,而这些信息可以很好地改善自发与非自发表情识别。     

融合表情和BVP生理信号的双模态视频情感识别

目的 针对当前视频情感判别方法大多仅依赖面部表情、而忽略了面部视频中潜藏的生理信号所包含的情感信息,本文提出一种基于面部表情和血容量脉冲（BVP）生理信号的双模态视频情感识别方法。方法 首先对视频进行预处理获取面部视频;然后对面部视频分别提取LBP-TOP和HOG-TOP两种时空表情特征,并利用视频颜色放大技术获取BVP生理信号,进而提取生理信号情感特征;接着将两种特征分别送入BP分类器训练分类模型;最后利用模糊积分进行决策层融合,得出情感识别结果。结果 在实验室自建面部视频情感库上进行实验,表情单模态和生理信号单模态的平均识别率分别为80%和63.75%,而融合后的情感识别结果为83.33%,高于融合前单一模态的情感识别精度,说明了本文融合双模态进行情感识别的有效性。结论 本文提出的双模态时空特征融合的情感识别方法更能充分地利用视频中的情感信息,有效增强了视频情感的分类性能,与类似的视频情感识别算法对比实验验证了本文方法的优越性。另外,基于模糊积分的决策层融合算法有效地降低了不可靠决策信息对融合的干扰,最终获得更优的识别精度。     

结合加权局部旋度模式的3维人脸表情识别

目的 3维人脸的表情信息不均匀地分布在五官及脸颊附近,对表情进行充分的描述和合理的权重分配是提升识别效果的重要途径。为提高3维人脸表情识别的准确率,提出了一种基于带权重局部旋度模式的3维人脸表情识别算法。方法 首先,为了提取具有较强表情分辨能力的特征,提出对3维人脸的旋度向量进行编码,获取局部旋度模式作为表情特征;然后,提出将ICNP（interactive closest normal points）算法与最小投影偏差算法结合,前者实现3维人脸子区域的不规则划分,划分得到的11个子区域保留了表情变化下面部五官和肌肉的完整性,后者根据各区域对表情识别的贡献大小为各区域的局部旋度模式特征分配权重;最后,带有权重的局部旋度模式特征被输入到分类器中实现表情识别。结果 基于BU-3DFE 3维人脸表情库对本文提出的局部旋度模式特征进行评估,结果表明其分辨能力较其他表情特征更强;基于BU-3DFE库进行表情识别实验,与其他3维人脸表情识别算法相比,本文算法取得了最高的平均识别率,达到89.67%,同时对易混淆的“悲伤”、“愤怒”和“厌恶”等表情的误判率也较低。结论 局部旋度模式特征对3维人脸的表情有较强的表征能力; ICNP算法与最小投影偏差算法的结合,能够实现区域的有效划分和权重的准确计算,有效提高特征对表情的识别能力。试验结果表明本文算法对3维人脸表情具有较高的识别率,并对易混淆的相似表情仍具有较好的识别效果。     

基于视频的人脸表情识别方法研究

近年来,表情识别逐渐成为计算机视觉和模式识别领域的研究热点之一。给出了一个包含人脸特征提取和表情识别的计算机视觉系统,通过对视频中人脸兼容运动特征的跟踪,提取人脸运动特征向量序列,与以往的方法不同,提取到的特征向量流被分割为两类,一类是表情特征向量流,另一类是视觉语音特征向量流。然后,利用基于CHMM（Couple Hidden Markov Model）的表情识别模型,进行人脸表情的识别,该模型允许两个向量流根据其各自的时域特征以异步方式进行处理,同时保持这两个向量流在时域上的自然关联。实验表明该方法优于传统的单通道处理方法。     

Exploring the effect of illumination on automatic expression recognition using the ICT-3DRFE database

One of the main challenges in facial expression recognition is illumination invariance. Our long-term goal is to develop a system for automatic facial expression recognition that is robust to light variations. In this paper, we introduce a novel 3D Relightable Facial Expression (ICT-3DRFE) database that enables experimentation in the fields of both computer graphics and computer vision. The database contains 3D models for 23 subjects and 15 expressions, as well as photometric information that allow for photorealistic rendering. It is also facial action units annotated, using FACS standards. Using the ICT-3DRFE database we create an image set of different expressions/illuminations to study the effect of illumination on automatic expression recognition. We compared the output scores from automatic recognition with expert FACS annotations and found that they agree when the illumination is uniform. Our results show that the output distribution of the automatic recognition can change significantly with light variations and sometimes causes the discrimination of two different expressions to be diminished. We propose a ratio-based light transfer method, to factor out unwanted illuminations from given images and show that it reduces the effect of illumination on expression recognition.     

基于特征区域自动分割的人脸表情识别?

针对目前三维人脸表情区域分割方法复杂、费时问题,提出一种人脸表情区域自动分割方法,通过投影、曲率计算的方法检测人脸的部分特征点,以上述特征点为基础进行人脸表情区域的自动分割。为得到更加丰富的表情特征,结合人脸表情识别编码规则对提取到的特征矩阵进行扩充,利用分类器进行人脸表情的识别。通过对三维人脸表情数据库部分样本的识别结果表明,该方法可以取得较高的识别率。     

Facial expression recognition experiments with data from television broadcasts and the World Wide Web

Facial expression recognition (FER) systems must ultimately work on real data in uncontrolled environments although most research studies have been conducted on lab-based data with posed or evoked facial expressions obtained in pre-set laboratory environments. It is very difficult to obtain data in real-world situations because privacy laws prevent unauthorized capture and use of video from events such as funerals, birthday parties, marriages etc. It is a challenge to acquire such data on a scale large enough for benchmarking algorithms. Although video obtained from TV or movies or postings on the World Wide Web may also contain ‘acted’ emotions and facial expressions, they may be more ‘realistic’ than lab-based data currently used by most researchers. Or is it? One way of testing this is to compare feature distributions and FER performance. This paper describes a database that has been collected from television broadcasts and the World Wide Web containing a range of environmental and facial variations expected in real conditions and uses it to answer this question. A fully automatic system that uses a fusion based approach for FER on such data is introduced for performance evaluation. Performance improvements arising from the fusion of point-based texture and geometry features, and the robustness to image scale variations are experimentally evaluated on this image and video dataset. Differences in FER performance between lab-based and realistic data, between different feature sets, and between different train-test data splits are investigated.     

A Facial Expression Emotion Recognition Based Human-robot Interaction System

A facial expression emotion recognition based human-robot interaction (FEER-HRI) system is proposed, for which a four-layer system framework is designed. The FEERHRI system enables the robots not only to recognize human emotions, but also to generate facial expression for adapting to human emotions. A facial emotion recognition method based on 2D-Gabor, uniform local binary pattern (LBP) operator, and multiclass extreme learning machine (ELM) classifier is presented, which is applied to real-time facial expression recognition for robots. Facial expressions of robots are represented by simple cartoon symbols and displayed by a LED screen equipped in the robots, which can be easily understood by human. Four scenarios, i.e., guiding, entertainment, home service and scene simulation are performed in the human-robot interaction experiment, in which smooth communication is realized by facial expression recognition of humans and facial expression generation of robots within 2 seconds. As a few prospective applications, the FEERHRI system can be applied in home service, smart home, safe driving, and so on.      

基于局部SVM分类器的表情识别方法

提出了一种新的视频人脸表情识别方法. 该方法将识别过程分成人脸表情特征提取和分类2个部分,首先采用基于点跟踪的活动形状模型(ASM)从视频人脸中提取人脸表情几何特征;然后,采用一种新的局部支撑向量机分类器对表情进行分类. 在Cohn2Kanade数据库上对KNN、SVM、KNN2SVM和LSVM 4种分类器的比较实验结果验证了所提出方法的有效性.     

Authentic facial expression analysis

There is a growing trend toward emotional intelligence in human–computer interaction paradigms. In order to react appropriately to a human, the computer would need to have some perception of the emotional state of the human. We assert that the most informative channel for machine perception of emotions is through facial expressions in video. One current difficulty in evaluating automatic emotion detection is that there are currently no international databases which are based on authentic emotions. The current facial expression databases contain facial expressions which are not naturally linked to the emotional state of the test subject. Our contributions in this work are twofold: first, we create the first authentic facial expression database where the test subjects are showing the natural facial expressions based upon their emotional state. Second, we evaluate the several promising machine learning algorithms for emotion detection which include techniques such as Bayesian networks, SVMs, and decision trees.     

Utilizing 3D flow of points for facial expression recognition

This paper presents an approach to recognize Facial Expressions of different intensities using 3D flow of facial points. 3D flow is the geometrical displacement (in 3D) of a facial point from its position in a neutral face to that in the expressive face. Experiments are performed on 3D face models from the BU-3DFE database. Four different intensities of expressions are used for analyzing the relevance of intensity of the expression for the task of FER. It was observed that high intensity expressions are easier to recognize and there is a need to develop algorithms for recognizing low intensity facial expressions. The proposed features outperform difference of facial distances and 2D optical flow. Performances of two classifiers, SVM and LDA are compared wherein SVM performs better. Feature selection did not prove useful.     

基于双相机捕获面部表情及人体姿态生成三维虚拟人动画

为了生成表情丰富、动作流畅的三维虚拟人动画,提出了一种基于双相机同步捕获面部表情及人体姿态生成三维虚拟人动画的方法。首先,采用传输控制协议（TCP）网络时间戳方法实现双相机时间同步,采用张正友标定法实现双相机空间同步。然后,利用双相机分别采集面部表情和人体姿态。采集面部表情时,提取图像的2D特征点,利用这些2D特征点回归计算得到面部行为编码系统（FACS）面部行为单元,为实现表情动画做准备;以标准头部3D坐标值为基准,根据相机内参,采用高效n点投影（EPnP）算法实现头部姿态估计;之后将面部表情信息和头部姿态估计信息进行匹配。采集人体姿态时,利用遮挡鲁棒姿势图（ORPM）方法计算人体姿态,输出每个骨骼点位置、旋转角度等数据。最后,在虚幻引擎4（UE4）中使用建立的虚拟人体三维模型来展示数据驱动动画的效果。实验结果表明,该方法能够同步捕获面部表情及人体姿态,而且在实验测试中的帧率达到20 fps,能实时生成自然真实的三维动画。     

Recognition of 3D facial expression dynamics

In this paper we propose a method that exploits 3D motion-based features between frames of 3D facial geometry sequences for dynamic facial expression recognition. An expressive sequence is modelled to contain an onset followed by an apex and an offset. Feature selection methods are applied in order to extract features for each of the onset and offset segments of the expression. These features are then used to train GentleBoost classifiers and build a Hidden Markov Model in order to model the full temporal dynamics of the expression. The proposed fully automatic system was employed on the BU-4DFE database for distinguishing between the six universal expressions: Happy, Sad, Angry, Disgust, Surprise and Fear. Comparisons with a similar 2D system based on the motion extracted from facial intensity images was also performed. The attained results suggest that the use of the 3D information does indeed improve the recognition accuracy when compared to the 2D data in a fully automatic manner.     

A video prediction approach for animating single face image

Generating dynamic 2D image-based facial expressions is a challenging task for facial animation. Much research work focused on performance-driven facial animation from given videos or images of a target face, while animating a single face image driven by emotion labels is a less explored problem. In this work, we treat the task of animating single face image from emotion labels as a conditional video prediction problem, and propose a novel framework by combining factored conditional restricted boltzmann machines (FCRBM) and reconstruction contractive auto-encoder (RCAE). A modified RCAE with an associated efficient training strategy is used to extract low dimensional features and reconstruct face images. FCRBM is used as animator to predict facial expression sequence in the feature space given discrete emotion labels and a frontal neutral face image as input. Both quantitative and qualitative evaluations on two facial expression databases, and comparison to state-of-the-art showed the effectiveness of our proposed framework for animating frontal neutral face image from given emotion labels.

     

Nonverbal social withdrawal in depression: Evidence from manual and automatic analyses

The relationship between nonverbal behavior and severity of depression was investigated by following depressed participants over the course of treatment and video recording a series of clinical interviews. Facial expressions and head pose were analyzed from video using manual and automatic systems. Both systems were highly consistent for FACS action units (AUs) and showed similar effects for change over time in depression severity. When symptom severity was high, participants made fewer affiliative facial expressions (AUs 12 and 15) and more non-affiliative facial expressions (AU 14). Participants also exhibited diminished head motion (i.e., amplitude and velocity) when symptom severity was high. These results are consistent with the Social Withdrawal hypothesis: that depressed individuals use nonverbal behavior to maintain or increase interpersonal distance. As individuals recover, they send more signals indicating a willingness to affiliate. The finding that automatic facial expression analysis was both consistent with manual coding and revealed the same pattern of findings suggests that automatic facial expression analysis may be ready to relieve the burden of manual coding in behavioral and clinical science.     

Semi‐supervised classification of facial expression using a mixture of multivariate t distributions

Abstract: This paper addresses the semi‐supervised classification of facial expression images using a mixture of multivariate t distributions. The facial expression features are first extracted into labelled graph vectors using the Gabor wavelet transformation. We then learn a mixture of multivariate t distributions by using the labelled graph vectors, and set correspondence between the component distributions and the basic facial emotions. According to this correspondence, the classification of a given testing image is implemented in a probabilistic way according to its fitted posterior probabilities of component memberships. Specifically, we perform hard classification of the testing image by assigning it into an emotional class that the corresponding mixture component has the highest posterior probability, or softly use the posterior probabilities as the estimates of the semantic ratings of expressions. The experimental results on the Japanese female facial expression database, Ekman's Pictures of Facial Affect database and the AR database demonstrate the effectiveness of the proposed method.     

基于二维主分量分析的面部表情识别

提出了一种直接基于图像矩阵的二维主分量分析(2DPCA)和多分类器联合的面部表情识别方法。首先利用2DPCA进行特征提取,然后用基于模糊积分的多分类器联合的方法对七种表情(生气、厌恶、恐惧、高兴、中性、悲伤、惊讶)进行识别。在JAFFE人脸表情静态图像库上进行实验,与传统主分量分析(PCA)相比,采用2DPCA进行特征提取,不仅识别率比较高,而且运算速度也有很大的提高。