深度人脸表情识别研究进展

随着人脸表情识别任务逐渐从实验室受控环境转移至具有挑战性的真实世界环境,在深度学习技术的迅猛发展下,深度神经网络能够学习出具有判别能力的特征,逐渐应用于自动人脸表情识别任务。目前的深度人脸表情识别系统致力于解决以下两个问题：1）由于缺乏足量训练数据导致的过拟合问题;2）真实世界环境下其他与表情无关因素变量（例如光照、头部姿态和身份特征）带来的干扰问题。本文首先对近十年深度人脸表情识别方法的研究现状以及相关人脸表情数据库的发展进行概括。然后,将目前基于深度学习的人脸表情识别方法分为两类：静态人脸表情识别和动态人脸表情识别,并对这两类方法分别进行介绍和综述。针对目前领域内先进的深度表情识别算法,对其在常见表情数据库上的性能进行了对比并详细分析了各类算法的优缺点。最后本文对该领域的未来研究方向和机遇挑战进行了总结和展望：考虑到表情本质上是面部肌肉运动的动态活动,基于动态序列的深度表情识别网络往往能够取得比静态表情识别网络更好的识别效果。此外,结合其他表情模型如面部动作单元模型以及其他多媒体模态,如音频模态和人体生理信息能够将表情识别拓展到更具有实际应用价值的场景。     

融合时空域特征的人脸表情识别

目的 人脸表情识别是计算机视觉的核心问题之一。一方面,表情的产生对应着面部肌肉的一个连续动态变化过程,另一方面,该运动过程中的表情峰值帧通常包含了能够识别该表情的完整信息。大部分已有的人脸表情识别算法要么基于表情视频序列,要么基于单幅表情峰值图像。为此,提出了一种融合时域和空域特征的深度神经网络来分析和理解视频序列中的表情信息,以提升表情识别的性能。方法 该网络包含两个特征提取模块,分别用于学习单幅表情峰值图像中的表情静态“空域特征”和视频序列中的表情动态“时域特征”。首先,提出了一种基于三元组的深度度量融合技术,通过在三元组损失函数中采用不同的阈值,从单幅表情峰值图像中学习得到多个不同的表情特征表示,并将它们组合在一起形成一个鲁棒的且更具辩识能力的表情“空域特征”;其次,为了有效利用人脸关键组件的先验知识,准确提取人脸表情在时域上的运动特征,提出了基于人脸关键点轨迹的卷积神经网络,通过分析视频序列中的面部关键点轨迹,学习得到表情的动态“时域特征”;最后,提出了一种微调融合策略,取得了最优的时域特征和空域特征融合效果。结果 该方法在3个基于视频序列的常用人脸表情数据集CK+(the e...     

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。     

An accurate recognition of facial expression by extended wavelet deep convolutional neural network

Facial expressions are essential in community based interactions and in the analysis of emotions behaviour. The automatic identification of face is a motivating topic for the researchers because of its numerous applications like health care, video conferencing, cognitive science etc. In the computer vision with the facial images, the automatic detection of facial expression is a very challenging issue to be resolved. An innovative methodology is introduced in the presented work for the recognition of facial expressions. The presented methodology is described in subsequent stages. At first, input image is taken from the facial expression database and pre-processed with high frequency emphasis (HFE) filtering and modified histogram equalization (MHE). After the process of image enhancement, Viola Jones (VJ) framework is utilized to detect the face in the images and also the face region is cropped by finding the face coordinates. Afterwards, different effective features such as shape information is extracted from enhanced histogram of gradient (EHOG feature), intensity variation is extracted with mean, standard deviation and skewness, facial movement variation is extracted with facial action coding (FAC),texture is extracted using weighted patch based local binary pattern (WLBP) and spatial information is extracted byentropy based Spatial feature. Subsequently, dimensionality of the features are reduced by attaining the most relevant features using Residual Network (ResNet). Finally, extended wavelet deep convolutional neural network (EWDCNN) classifier uses the extracted features and accurately detects the face expressions as sad, happy, anger, fear disgust, surprise and neutral classes. The implementation platform used in the work is PYTHON. The presented technique is tested with the three datasets such as JAFFE, CK+ and Oulu-CASIA.

     

An efficient deep learning technique for facial emotion recognition

Emotion recognition from facial images is considered as a challenging task due to the varying nature of facial expressions. The prior studies on emotion classification from facial images using deep learning models have focused on emotion recognition from facial images but face the issue of performance degradation due to poor selection of layers in the convolutional neural network model.To address this issue, we propose an efficient deep learning technique using a convolutional neural network model for classifying emotions from facial images and detecting age and gender from the facial expressions efficiently. Experimental results show that the proposed model outperformed baseline works by achieving an accuracy of 95.65% for emotion recognition, 98.5% for age recognition, and 99.14% for gender recognition.

     

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.     

基于多尺度分层双线性池化网络的细粒度表情识别模型

人脸表情细微的类间差异和显著的类内变化增加了人脸表情识别难度。构建一个基于多尺度双线性池化神经网络的识别模型。设计3种不同尺度网络提取人脸表情全局特征,并引入分层双线性池化层,集成多个同一网络及不同网络的多尺度跨层双线性特征以捕获不同层级间的部分特征关系,从而增强模型对面部表情细微特征的表征及判别能力。同时,使用逐层反卷积融合多层特征信息,解决神经网络通过多层卷积层、池化层提取特征时丢失部分关键特征的问题。实验结果表明,该模型在FER2013和CK+公开数据集上的识别率分别为73.725%、98.28%,优于SLPM、CL、JNS等人脸表情识别模型。     

Facial expression recognition based on dual-feature fusion and improved random forest classifier

Facial expression recognition (FER) is an important means for machines to understand the changes in the facial expression of human beings. Expression recognition using single-modal facial images, such as gray scale, may suffer from illumination changes and the lack of detailed expression-related information. In this study, multi-modal facial images, such as facial gray scale, depth, and local binary pattern (LBP), are used to recognize six basic facial expressions, namely, happiness, sadness, anger, disgust, fear, and surprise. Facial depth images are used for robust face detection initially. The deep geometric feature is represented by point displacement and angle variation in facial landmark points with the help of depth information. The local appearance feature, which is obtained by concatenating LBP histograms of expression-prominent patches, is utilized to recognize those expression changes that are difficult to capture by only the geometric changes. Thereafter, an improved random forest classifier based on feature selection is used to recognize different facial expressions. Results of comparative evaluations in benchmarking datasets show that the proposed method outperforms several state-of-the-art FER approaches that are based on hand-crafted features. The capability of the proposed method is comparable to that of the popular convolutional neural-network-based FER approach but with fewer demands for training data and a high-performance hardware platform.     

Adaptive increasing-margin adversarial neural iterative system based on facial expression recognition feature models

Face expression recognition has been proved as a challenging task in image processing. Many related works on facial expression recognition had done but they faced several challenges during the classification of data with the stored database. It carried out various workflows on improvisation of classifiers based on deep learning but they have been lagging in understanding facial expression mainly because of disastrous forgetting, time management, data mixing, and data overfitting, etc. Ignoring all these challenges would lead to inaccurate recognition of facial expressions. Hence to overcome all the above issues this work proposed a model named adaptive increasing-margin adversarial neural iterative model involves triple threat filtration techniques along with modified scaling density-based spatial clustering of applications with noise and dual feature model for obtaining a better quality featured image. Advance back propagation artificial neural network model is initiated to overcome catastrophic forgetting, underfitting of data, over fitting of data, etc. Thus, the proposed work achieves better efficiency as well as high accuracy in terms of facial expression recognition.

     

面向类内差距表情的深度学习识别

目的 为解决真实环境中由类内差距引起的面部表情识别率低及室内外复杂环境对类内差距较大的面部表情识别难度大等问题,提出一种利用生成对抗网络（generative adversarial network,GAN）识别面部表情的方法。方法 在GAN生成对抗的思想下,构建一种IC-GAN（intra-class gap GAN）网络结构,使用卷积组建编码器、解码器对自制混合表情图像进行更深层次的特征提取,使用基于动量的Adam（adaptive moment estimation）优化算法进行网络权重更新,重点针对真实环境面部表情识别过程中的类内差距较大的表情进行识别,使其更好地适应类内差异较大的任务。结果 基于Pytorch环境,在自制的面部表情数据集上进行训练,在面部表情验证集上进行测试,并与深度置信网络（deep belief network,DBN）和GoogLeNet网络进行对比实验,最终IC-GAN网络的识别结果比DBN网络和GoogLeNet网络分别提高11%和8.3%。结论 实验验证了IC-GAN在类内差距较大的面部表情识别中的精度,降低了面部表情在类内差距较大情况下的误识率,提高了系统鲁棒性,为面部表情的生成工作打下了坚实的基础。     

CFSM: a novel frame analyzing mechanism for real-time face recognition system on the embedded system

The development of web cameras and smart phones is mature, and more and more facial recognition-related applications are implemented on embedded systems. The demand for real-time face recognition on embedded systems is also increasing. In order to improve the accuracy of face recognition, most of the modern face recognition systems consist of multiple deep neural network models for recognition. However, in an embedded system, integrating these complex neural network models and execute simultaneously is not easy to achieve the goal of real-time recognition of human faces and their identities. In view of this, this study proposes a new frame analysis mechanism, continuous frames skipping mechanism (CFSM), which can analyze the frame in real time to determine whether it is necessary to perform face recognition on the current frame. Through the analysis of CFSM, the frames that do not need to be re-recognized for face are omitted. In this way, the workload of the face recognition system will be greatly reduced to achieve the goal of real-time face recognition in the embedded system. The experimental results show that the proposed CFSM mechanism can greatly increase the speed of face recognition in the video on the embedded system, achieving the goal of real-time face recognition.

     

构建并行卷积神经网络的表情识别算法

目的 表情识别在商业、安全、医学等领域有着广泛的应用前景,能够快速准确地识别出面部表情对其研究与应用具有重要意义。传统的机器学习方法需要手工提取特征且准确率难以保证。近年来,卷积神经网络因其良好的自学习和泛化能力得到广泛应用,但还存在表情特征提取困难、网络训练时间过长等问题,针对以上问题,提出一种基于并行卷积神经网络的表情识别方法。方法 首先对面部表情图像进行人脸定位、灰度统一以及角度调整等预处理,去除了复杂的背景、光照、角度等影响,得到了精确的人脸部分。然后针对表情图像设计一个具有两个并行卷积池化单元的卷积神经网络,可以提取细微的表情部分。该并行结构具有3个不同的通道,分别提取不同的图像特征并进行融合,最后送入SoftMax层进行分类。结果 实验使用提出的并行卷积神经网络在CK+、FER2013两个表情数据集上进行了10倍交叉验证,最终的结果取10次验证的平均值,在CK+及FER2013上取得了94.03%与65.6%的准确率。迭代一次的时间分别为0.185 s和0.101 s。结论 为卷积神经网络的设计提供了一种新思路,可以在控制深度的同时扩展广度,提取更多的表情特征。实验结果表明,针对数量、分辨率、大小等差异较大的表情数据集,该网络模型均能够获得较高的识别率并缩短训练时间。     

持续学习改进的人脸表情识别

目的 大量标注数据和深度学习方法极大地提升了图像识别性能。然而,表情识别的标注数据缺乏,训练出的深度模型极易过拟合,研究表明使用人脸识别的预训练网络可以缓解这一问题。但是预训练的人脸网络可能会保留大量身份信息,不利于表情识别。本文探究如何有效利用人脸识别的预训练网络来提升表情识别的性能。方法 本文引入持续学习的思想,利用人脸识别和表情识别之间的联系来指导表情识别。方法指出网络中对人脸识别整体损失函数的下降贡献最大的参数与捕获人脸公共特征相关,对表情识别来说为重要参数,能够帮助感知面部特征。该方法由两个阶段组成：首先训练一个人脸识别网络,同时计算并记录网络中每个参数的重要性;然后利用预训练的模型进行表情识别的训练,同时通过限制重要参数的变化来保留模型对于面部特征的强大感知能力,另外非重要参数能够以较大的幅度变化,从而学习更多表情特有的信息。这种方法称之为参数重要性正则。结果 该方法在RAF-DB（real-world affective faces database）,CK+（the extended Cohn-Kanade database）和Oulu-CASIA这3个数据集上进行了实验评估。在主流数据集RAF-DB上,该方法达到了88.04%的精度,相比于直接用预训练网络微调的方法提升了1.83%。其他数据集的实验结果也表明了该方法的有效性。结论 提出的参数重要性正则,通过利用人脸识别和表情识别之间的联系,充分发挥人脸识别预训练模型的作用,使得表情识别模型更加鲁棒。     

多信息融合的深度学习人脸表情识别算法研究

人脸表情识别作为人机交互系统的重要组成部分,在安防监控、人机交互等领域有广泛的应用,是计算机视觉的研究热点。传统的卷积神经网络方法一般提取单张人脸图像或者人脸标记点作为特征提取的输入数据,未能考虑到人脸全域的表情信息。提出了一种基于三通道多信息融合的深度学习人脸表情识别模型,以人脸图像表情平静到高峰时期标记点坐标的相对位移为输入,提取整个人脸表情图像特征信息,模型融合了稀疏自编码器以提高对边缘特征提取效率。该模型在CK+数据集上进行了训练和测试,实验结果表明,与该领域中的同类算法相比,该算法模型提高了表情识别的准确率。     

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。     

深度学习下融合不同模型的小样本表情识别

为了进一步提高人脸表情识别在小样本中的准确率,提出了一种深度学习下融合不同模型的小样本表情识别方法。该方法首先对单个卷积神经网络(CNN)模型进行比较,通过dropout层不同的节点保留概率p,筛选相对合适的CNN。之后采用尺度不变特征变换(SIFT)算法提取出特征,使用SIFT提取特征的目的是提高小数据的性能。为了减少误差,避免过拟合,将所有模型进行汇总,采用简单平均的模型融合方法得到CNN-SIFT-AVG模型。最后,只采用少量样本数据来训练模型即可。该模型已在FER2013、CK+和JAFFE数据集上进行了验证实验。实验结果表明,该模型可以很大程度上提高小样本表情识别的准确率,并在FER2013、CK+和JAFFE数据集上产生了较优异的结果,与其他表情识别方法相比,准确率最大提升约6%。     

结合改进卷积神经网络与通道加权的轻量级表情识别

目的 表情是人机交互过程中重要的信息传递方式,因此表情识别具有重要的研究意义。针对目前表情识别方法存在背景干扰大、网络模型参数复杂、泛化性差等问题,本文提出了一种结合改进卷积神经网络(convolutional neural network, CNN)与通道加权的轻量级表情识别方法。方法 首先,采用标准卷积和深度可分离卷积组合神经网络结构,再利用全局平均池化层作为输出层,简化网络的复杂程度,有效降低网络参数;其次,网络引入SE(squeeze-and-excitation)模块进行通道加权,通过在不同卷积层后设置不同的压缩率增强表情特征提取能力,提升网络模型精度;最后,用softmax分类函数实现各类表情的准确分类。结果 本文网络参数量为6 108 519,相较于识别性能较好的Xception神经网络参数减少了63%,并且通过对网络模型的实时性测试,平均识别速度可达128帧/s。在5个公开的表情数据集上验证网络模型对7种表情的识别效果,与7种卷积神经网络方法相比,在FER2013 (Facial Expression Recognition 2013)、CK+(the extended...     

面部表情的深度卷积级联森林识别

面部表情识别广泛应用于各种研究领域,针对面部表情识别使用深度神经网络方法结构复杂、可解释性差和传统机器学习方法特征提取缺乏多样性、识别率低的问题.提出了一种新的深度卷积级联森林(Deep Convolution Cascade Forest,DCCF)方法用于人脸面部表情识别,该方法通过卷积神经网络深度学习人脸面部显著特征,并采用基于随机森林的级联结构森林逐层学习识别出不同的面部表情特征,提高了人脸表情的识别准确率.DCCF在JAFFE、CK+和Fer2013 3个公开面部表情数据集进行了实验,并对面部表情提取的5种特征和7种分类方法进行了比较分析,结果显示DCCF在对比的算法中人脸表情识别性能最好,3个数据集的准确率分别达到91.4%,98.7%,71.6%.     

结合关键点与权重分配残差网络的表情识别

现有人脸表情识别算法易受图像背景、非表情内容等无关因素的影响。此外,部分人脸表情（例如害怕、生气、伤心等表情）的类间差异较小也制约着算法的性能。针对上述两个问题,提出了一种融合面部关键点和权重分配残差网络的表情识别算法。通过面部关键点获取最大的表情范围以消除图像背景和非表情内容的干扰,将预处理后的表情图像作为深度残差网络的输入,引入权重分配机制从通道和空间维度上进行注意权重推断,实现不同区域的权重分配,进而引导深度残差网络学习对表情具有鉴别力的局部特征。该算法分别在FER2013和CK+表情数据集上达到了74.14%和98.99%的识别准确率,有效改善了生气、伤心、害怕等类间差异较小的表情识别准确率。