TP-FER：基于优化卷积神经网络的三通道人脸表情识别方法

针对人脸五官在表达不同情绪时所起的作用不同,利用单一的卷积神经网络对人脸面部特征进行特征提取和表情识别可能会导致提取表情关键特征信息时聚焦性不够,而仅对眼部或者嘴部等重点部位进行特征提取,又有可能造成特征提取不够充分的问题,提出了一种基于优化卷积神经网络的三通道人脸表情识别方法TP-FER(tri-path networks for facial expression recognition).该方法基于构建的卷积神经网络训练,采用三个输入渠道,分别聚焦面部、眼部和嘴部区域进行特征提取和表情判别,最后采用基于决策层的融合技术将三个渠道的识别结果进行相对多数投票决策,获取整体最优识别率.将此方法应用于JCK+数据集和自建数据集上进行了实验判别分析,结果表明该方法在两个数据集上均提高了整体表情识别率.该方法既考虑了脸部整体特征的提取,又兼顾了某些表情主要聚焦在眼部、嘴部表达的特性,相互辅助,整体提高了表情的识别率;该方法也能对神经心理学研究提供数据支持.     

不同注意线索对积极情绪共情的影响

共情是知觉和理解他人情绪并合理反应的能力。通过行为实验要求被试对不同情绪、不同性别的面孔作相应的判断,考查不同注意线索对积极情绪共情的影响。实验发现,在情绪判断任务中,被试对于正性图片和中性图片的反应时存在显著差异,正性图片的反应时显著快于中性图片;而在性别判断任务中两类图片的反应时不存在显著差异。实验证实了不同注意线索对积极情绪的共情能够产生影响：当被试将注意集中情绪判断线索时,被试的积极情绪共情明显;当被试将注意集中在性别判断线索时,被试的积极情绪共情减弱。     

微型小说的艺术特点研究

经过将近30年的研究与探索,微型小说已发展为一种新兴小说文体,由于微型小说脱胎于传统的小说,它既具有小说的艺术特点,又有属于自己的独特艺术特色,文章阐述了微型小说主题的以小见大、构思的精致巧妙、抒情的感人至深的艺术特点。     

Effects of carbon content and solidification rate on thermal conductivity of grey cast iron

The thermal conductivity or diffusivity of pearlitic grey irons with various carbon contents is investigated by the laser flash method. The materials are cast in controlled thermal environments and produced in three dissimilar cooling rates. The cooling rate together with the carbon content largely influence the thermal conductivity of grey iron. Linear relationships exist between the thermal conductivity and the carbon content, the carbon equivalent and the fraction of former primary solidified austenite transformed into pearlite. The work shows that optimal thermal transport properties are obtained at medium cooling rates. Equations describing the thermal conductivity of pearlite,solidified as pre-eutectic austenite, and the eutectic of grey iron are derived. The thermal conductivity of pearlitic grey iron is modeled at both room temperature and elevated temperature with good accuracy.     

VALENCE: affective visualisation using EEG

VALENCE is an interactive visualisation controlled by live brainwave monitoring. We used a wireless EEG headset to monitor the player's alpha waves (an indicator of relaxation) and valence (an indicator of emotion or arousal). The game world is an emergent system of attractive and repulsive forces responding to EEG input.     

Dynamic evolution of collective emotions in social networks: a case study of Sina weibo

Emotion is a fundamental object of human existence and determined by a complex set of factors. With the rapid development of online social networks (OSNs), more and more people would like to express their emotion in OSNs, which provides wonderful opportunities to gain insight into how and why individual emotion is evolved in social network. In this paper, we focus on emotion dynamics in OSNs, and try to recognize the evolving process of collective emotions. As a basis of this research, we first construct a corpus and build an emotion classifier based on Bayes theory, and some effective strategies (entropy and salience) are introduced to improve the performance of our classifier, with which we can classify any Chinese tweet into a particular emotion with an accuracy as high as 82%. By analyzing the collective emotions in our sample networks in detail, we get some interesting findings, including a phenomenon of emotion synchronization between friends in OSNs, which offers good evidence for that human emotion can be spread from one person to another. Furthermore, we find that the number of friends has strong correlation with individual emotion. Based on those useful findings, we present a dynamic evolution model of collective emotions, in which both self-evolving process and mutual-evolving process are considered. To this end, extensive simulations on both real and artificial networks have been done to estimate the parameters of our emotion dynamic model, and we find that mutual-evolution plays a more important role than self-evolution in the distribution of collective emotions. As an application of our emotion dynamic model, we design an efficient strategy to control the collective emotions of the whole network by selecting seed users according to k-core rather than degree.     

游戏中图像内容情感语义的分析与研究

针对目前游戏中忽略情感因素的问题,在游戏场景中引入情感语义的概念,提出基于情感语义的游戏场景图像检索系统和情感计算研究情感信息的数字化处理方法,采用支持向量机算法建立场景图像内容和其所表达的情感语义之间的联系。来建立一个能感知识别人的情感,能对人的情感做出智能反应的游戏系统,从而使人机交互更加和谐化。     

浅谈产品设计中色彩的魅力

色彩是产品设计中重要的视觉传达要素,反映了人们心理审美视觉感受,赋予产品情感,增添产品的艺术魅力。本文通过对色彩的心理情感联想的阐述,根据不同年龄阶段的消费群体对色彩的不同喜好感受,结合实例论述了色彩在产品设计中的应用,最后通过色彩的创新运用,展望色彩在未来产品设计中的发展。     

一种基于脉冲耦合神经网络的语音情感识别新方法

针对汉语语音情感识别问题,提出了一种基于脉冲耦合神经网络(PCNN)的识别方法。该方法将语音转化为语谱图后输入到PCNN,得到输出图像的神经元点火序列及其熵序列作为语音情感的特征,利用其特征实现语音情感识别。实验结果表明,该方法可以有效地识别“高兴”与“平常”这两种不同的情感。该方法将PCNN引入到语音情感识别的应用研究中,开拓了语音和图像信号结合处理的新领域,同时对于PCNN的理论研究和实际应用具有重要的现实意义。     

融合语音、脑电和人脸表情的多模态情绪识别

本文提出了一种多模态情绪识别方法, 该方法融合语音、脑电及人脸的情绪识别结果来从多个角度综合判断人的情绪, 有效地解决了过去研究中准确率低、模型鲁棒性差的问题. 对于语音信号, 本文设计了一个轻量级全卷积神经网络, 该网络能够很好地学习语音情绪特征且在轻量级方面拥有绝对的优势. 对于脑电信号, 本文提出了一个树状LSTM模型, 可以全面学习每个阶段的情绪特征. 对于人脸信号, 本文使用GhostNet进行特征学习, 并改进了GhostNet的结构使其性能大幅提升. 此外, 我们设计了一个最优权重分布算法来搜寻各模态识别结果的可信度来进行决策级融合, 从而得到更全面、更准确的结果. 上述方法在EMO-DB与CK+数据集上分别达到了94.36%与98.27%的准确率, 且提出的融合方法在MAHNOB-HCI数据库的唤醒效价两个维度上分别得到了90.25%与89.33%的准确率. 我们的实验结果表明, 与使用单一模态以及传统的融合方式进行情绪识别相比, 本文提出的多模态情绪识别方法有效地提高了识别准确率.