北京pk10赛车开奖记录结果

服务热线:

13911387701

新闻资讯
联系我们
联系人:方经理
电 话:010-62631382
手机:13911387701
邮箱:2497477429@qq.com
地址:北京市大兴区西红门26号明珠庄园6号楼307
网址:wxxydh.com
基于人工智能的微表情识别技术
添加时间:2019-8-16 来源:网络收集
基于人工智能的微表情识别技术

作者:北京pk10赛车开奖记录结果 谢东亮, 徐宇翔(北京邮电大学网络与交换技术国家重点实验室,100876)人工智能开放创新平台(chinaopen.ai)联合学者

摘要:北京pk10赛车开奖记录结果 应对重大突发事件的能力是一个城市现代化程度的重要标志。自911 事件以后,各个国家更加迫切需要行之有效的社会安全风险预警。目前我国正进入“突发公共事件的高发期”和“社会高风险期”。如何利用科技手段应对“两高”,是我国政府的当务之急。随着人工智能技术的发展,机器智能可以利用海量的视频数据,结合模式识别、深度学习等先进算法,使视频分析精细化、可视化、自动化、智能化。本文介绍了一种新颖的基于人工智能的情绪分析技术,在非接触微表情研究、微表情与情绪关系的心理学研究理论基础上,介绍了基于微表情识别的灵敏、精准和鲁棒无感知情绪监测分析系统,并制定相应的预警策略,使其能够辅助人们理解和分析人员的动机,为社会安全风险控制提供预警与决策的潜在线索。该系统也可推广应用于金融评估、商业谈判、心理干预等,用于对人员的真实情绪进行分析,具有良好的潜在应用价值。

关键词: 人工智能,微表情,情绪

ZHENDUIRENMENZHENSHIQINGXUHEYITUDEYANJIUSHIZHONGSHIXINLIXUEHESHEHUIXUEZHUYAOYANJIUFANGXIANG。ZAI20SHIJIZAOQI,YOUXUEZHEKAISHIYISHENGLIZHIBIAOWEIJICHUDEQINGXUHUOZHEYITUYANJIU。1921NIANJIAZHOUDAXUEJohn LarsonFAMINGLECEHUANGYI,ZHIHOUYOUCHUXIANLEJIYURETUXIANG、NAODIANXINHAOYIJIGONGNENGCIGONGZHENCHENGXIANGDENGFANGFADESHENGLIJIANCEFANGFA。DANZHEIXIEFANGFADEXINXICAIJIXUYAOZHUANMENSHEBEI,ERQIEZHEIXIEPINGGUSHOUDUANJIBENGONGKAI,RENMENKEYIJINGGUOYIDINGDEXUNLIANLAIYINMANQIZHENSHIYITU。YUSHANGSHUSHENGLIXIANSUOXIANGBI,MIANBUBIAOQINGSHIZAIRENLEIJINHUAGUOCHENGZHONGXINGCHENGDE,SHIRENLEIZHIJIANCHUANDISHEHUIXINXIDEZHUYAOSHOUDUANHEZHIGUANSHOUDUAN。YOUYUMIANBUBIAOQINGTEZHENGBAOHANFENGFUZHIGUANDEQINGXUXINXI,BINGQIEKEYITONGGUOFEIJIECHUDECAIJIFANGSHIHUOQU,YINCIHUODERENMENGUANGFANGUANZHU。

北京pk10赛车开奖记录结果 BIAOQINGSHIQINGXUDEZHUGUANTIYANDEWAIBUBIAOXIANMOSHI,FENWEISHENGLIBIAOQING(ZHENSHIXINLIZHUANGTAI)、QINGXUBIAOQING(ZHENSHIXINLIZHUANGTAI+WEIZHUANGJUECE)HESHEJIAOBIAOQING(LIXINGJUECEHEKONGZHI)DENG。MEIGUOPaul EkmanJIAOSHOUJIANGRENLEIDEMIANBUBIAOQINGFENWEILIULEI:GAOXING、JINGYA、BEISHANG、FENNU、YANE、KONGJU。QIZHONG,XINLIXUEJIAHESHENJINGXUEJIAFAXIAN,QIPIANZHEHUITONGGUOQINGXUQIPIANSHITUYAYIMOUXIEFANYINGZHENSHIQINGXUDEXINHAO,DANQUEWUFAWANQUANYAYI,DAOZHIQIZHENSHIQINGXUXINHAOXIELU,ZHEIBIANCHUXIANLEWEIRUOQIEKUAISUDEMIANBUDONGZUO,JIWEIBIAOQING。WEIBIAOQINGZETEZHIRENLEISHITUYAYIHUOYINZANGZHENSHIQINGGANSHIXIELUDEFEICHANGDUANZANQIEBUNENGZIZHUKONGZHIDEMIANBUBIAOQING。MEIGUOZHUMINGXINLIXUEJIA,BIAOQINGHEWEIBIAOQINGDEDIANJIZHEEkmanJINGGUOYANJIURENWEI,WEIBIAOQINGJUYOUSANGETEDIAN:CHIXUSHIJIANBUCHAOGUO1/5MIAO,NENGFANYINGRENDEZHENSHIQINGGAN,ZAIQUANRENLEISHIPUBIANCUNZAIDE。

北京pk10赛车开奖记录结果 WEIBIAOQINGKENENGSHIPANDUANYIGERENZHENSHIQINGGANDEZUIYOULIDEXIANSUO。JINGGUOJISHINIANDELILUNFAZHANHESHIYANYANZHENG,WEIBIAOQINGZHUJIANBEIXUESHUJIEJIESHOUHERENKE,MEIGUOYIJINGZAIZHEIFANGMIANJINXINGLEJISHINIANDEYANJIUGONGZUO,YIBEIMEIGUOJIAOTONGYUNSHUANQUANBUYONGYUDUOGEJICHANGDEANJIANZHONG,CIWAI,ZAIMEIGUOSIFASHENXUN、LINCHUANGYIXUEDENGLINGYUYEJINXINGLEYINGYONGCESHI。DANGUONEIZAIWEIBIAOQINGDEYANJIUQIBUJIAOWAN,YANJIUCHENGGUOJIAOSHAO,ERYOUYUGAILINGYUDEYANJIUZAIHENDACHENGDUSHANGDUIYUGUOJIAANQUANHESIFASHIJIANJIAOWEIZHONGYAO,SUONENGHUODEDEGUOWAIZILIAOJIAOSHAO。ZHEIZHONGFENGSUOZAIYIDINGCHENGDUSHANGYESHUOMINGLEWEIBIAOQINGYANJIUDEZHONGYAOYIYIHEQIANZAIJIAZHI,YINCIYOUBIYAOJIAQIANGDUIWEIBIAOQINGDEYANJIU。

北京pk10赛车开奖记录结果 ZAISHIJIYINGYONGZHONG,RENMENWANGWANGXUYAOZHENDUIZHANGSHIPINZHONGDEMIANBUBIAOQINGJINXINGFENXI。YINCI,ZUOWEIYITAOWANZHENGSHIYONGXITONG,SHOUXIANXUYAOYANJIUWEIBIAOQINGHEHONGBIAOQINGLIANHEJIANCEJISHU,BINGDUIJIANCEDAODEMIANBUXULIEJINXINGJIUZHENG,RANHOUYIJIUZHENGGUODEMIANBUXULIEWEIJICHU,DUIQIZHONGBAOHANDEQINGXUJINXINGFENLEISHIBIE,JINERJIANLICONGJIANCEDAOSHIBIEDEXITONGTIXI。 ZHUYAOYANJIUNEIRONGRUXIA:

(1)基于长视频的宏表情与微表情检测研究

MUQIANDADUOWEIBIAOQINGYANJIURENGJIYUDUIYANGBENTUXIANGHEQUEDINGSHIPINZHENDESHIBIE,ERZHENSHIXITONGZEXUYAOCONGZHANGSHIPINZHONGJIANCEDAOWEIBIAOQINGDECHUXIANCAINENGJINYIBUDUIWEIBIAOQINGJINXINGFENXI,YOUCI,ZUOWEIWEIBIAOQINGYANJIUDEJISHUJICHU,SHOUXIANJIANGZAIWEIBIAOQINGYUHONGBIAOQINGJIANCEDEYANJIUJICHUSHANG,YANJIUBINGKEHUAHONGBIAOQINGYUWEIBIAOQINGZAISHIJIANHEKONGJIANSHANGDECHAYIXING,JIANGDIHONGBIAOQINGZAIWEIBIAOQINGJIANCESHIDEGANRAOYINGXIANG,BINGTONGGUODUIMIANBUYUNDONGQIANGDUHESHIKONGYUESHUDEFENXILAITANSUOSHISHIXINGHEKEKAOXINGDEZHIYUEGUANXI,JIANLIYOUHUAMOXINGJINXINGWENTIQIUJIE,JIEJUEWEIBIAOQINGHEHONGBIAOQINGBINGCUNDEJIANCENANTI,ZUIZHONGWEISHIXIANBIAOQINGBIANHUAFENXITIGONGLIANGHAODEJICHUBAOZHANG。

(2)基于人脸通用三维模型配准的正面视角表情图像合成

RENLIANZITAIDERENYIXINGKEGUANSHANGZAOCHENGLEBUTONGCHENGDUXINGBIANYASUODERENLIANXINGZHUANGHEZIZHEDANGDEBUKEJIANWENLI,ZHEIJIANGSHIBIAOQINGSHIBIEHEFENLEIZIXITONGXINGNENGJIJUEHUA。YINCI,RUHEGAOXIAOZHUNQUEDEDUISHURUTUXIANGJINXINGZITAIGUJISHITIGAOHECHENGTUXIANGZHUNQUELVDEGUANJIANWENTI。DUIYUJIDINGDESHURUTUXIANG,RUHEXIEDIAOJISUANFUZAXINGHEJIEGUOJINGQUEDUERZHEDEMAODUN,JINXINGGUANJIANQUYUDEBIYAOTEZHENGDIANBIAODING,SHIHECHENGZHENGMIANTUXIANGDEYOUYINANDIAN。

(3)基于多动态局部特征融合的微表情识别

北京pk10赛车开奖记录结果 WEIBIAOQINGSHIBIEDEKEKAOXINGSHIBAOZHANGWEIBIAOQINGFENXIDEJICHUHEGUANJIAN,MUQIANWEIBIAOQINGDESHIBIELVHESHISHIXINGDOUYUANYUANDABUDAOZHENSHIHUANJINGXIADEXINGNENGYAOQIU。RUHETONGGUODUIWEIBIAOQINGSHUJUDEFENXI,JIANSHAORONGYUZHENDEGANRAOHETIGAOWEIBIAOQINGDESHIBIESUDUSHISHIBIEXIAOLVDEGUANJIANXINGJISHU。WEICI,RUHETANSUOJIYUWENLITEZHENGHEJIYUYUNDONGTEZHENGDUIWEIBIAOQINGDEKEHUACHENGDU,TONGSHIKAOLVDAOWEIBIAOQINGZAIMIANBUJUBUXINGDESHIJUETIXIAN,JIANLIJIYUQUANZHONGCELVEDEJUBUWENLITEZHENGHEYUNDONGTEZHENGRONGHEDETEZHENGTIQUQIUJIEMOXING,ZUIZHONGSHIXIANSHISHIKEKAODEWEIBIAOQINGSHIBIESUANFA,WEIBIAOQINGBIANHUAFENXITIGONGZHICHENG。

(4)基于实时表情变化分析的行为预警
北京pk10赛车开奖记录结果 微表情识别的目的在于通过机器智能为人们提供预警参考。如何根据表情识别的结果,进行合理的表情变化预测分析,进而及时排查出可疑人员是预警系统的核心难题。研究多指标联合预警策略,保障预警的实时性和可靠性,辅助相关人员对特殊事件快速做出反应,是对情感分析所反映的潜在行为分析的有效途径。

北京pk10赛车开奖记录结果 WEIBIAOQINGZIDONGFENXIKEYIFENWEIJIANCEHESHIBIELIANGGEGUOCHENG。XIANGBIYUKEYIJIEJIANHONGBIAOQINGJIANCEJISHUDEWEIBIAOQINGJIANCE,WEIBIAOQINGDESHIBIEJISHUJUYOUGENGDADEYANJIUTIAOZHAN,ZHEIYESHIMUQIANWEIBIAOQINGLINGYUDEYANJIUZHONGDIAN。

YOUYUWEIBIAOQINGCHIXUSHIJIANDUANHEDONGZUOFUDUXIAOLIANGDASHIBIENANDIAN,MUQIANDESHIBIELVRENGYOUHENDADETISHENGKONGJIAN。CHUANTONGDADUOCAIYONGJIYUJULEIDEFANGFA,LIANHE3DGAOSILVBOQIHEKJUNZHISUANFA,LAICELIANGWEIBIAOQINGDEKAISHI、FENGZHIHEJIEWEIJIEDUAN,RANERZAIZHEIGEFANGFAZHONG,JULEIDESHULIANGHENNANJUEDING。LINGYIZHONGJIYUFENLEIDEFANGFAKEYILIYONGSHIKONGJUBUWENLIMIAOSHUQILAIBIAOSHITEZHENG,SUIHOUTONGGUOZHICHIXIANGLIANGJISVM(Support Vector Machine, SVM)FENLEIQILAIJINXINGFENLEI。ZHEIXIEGONGZUODADOUZHILIYUZAITEZHENGDECENGMIANSHANGGAIJINWEIBIAOQINGSHIBIEDEXINGNENG,QUDELEYIDINGDEXINGNENGGAIJIN,DANSHIRENGRANQIANQUEJISUANDEDAOTEZHENGDEKEJIESHIXING。WEICI,WOMENTICHUYIZHONGJIYUSHENDUXUEXIDEWEIBIAOQINGSHIBIEFANGFA、SHENDUXUEXIDEGAINIANYOUHintonDENGRENYU2006NIANTICHU,SHUYUJIQIXUEXIYANJIUZHONGDEYIGEXINDELINGYU,SHIYIZHONGSHITUSHIYONGBAOHANFUZAJIEGOUHUOYOUDUOZHONGFEIXIANXINGBIANHUANGOUCHENGDEDUOGECHULICENGDUISHUJUJINXINGGAOCENGCHOUXIANGDESUANFA。SUANFABENZHISHIDUISHUJUDEBIAOZHENGXUEXI,MUBIAOSHIXUNQIUGENGHAODEBIAOSHIFANGFABINGCHUANGJIANGENGHAODEMOXINGLAICONGDAGUIMOWEIBIAOJISHUJUZHONGXUEXIZHEIXIEBIAOSHIFANGFA。LIRU,ZHENDUIYIFUTUXIANG,GUANCEZHIKEYISHIYONGDUOZHONGFANGSHILAIBIAOSHI,RUMEIGEXIANGSUQIANGDUZHIDEXIANGLIANG,HUOZHEGENGCHOUXIANGDIBIAOSHICHENGYIXILIEBIAN、TEDINGXINGZHUANGDEQUYUDENG。ERSHIYONGMOUXIETEDINGDEBIAOSHIFANGFAGENGRONGYICONGSHILIZHONGXUEXIRENWU。ANZHAOXUNLIANYANGBENBIAOQIANDEYOUWU,SHENDUXUEXIKEYIFENWEIYOUJIANDUXUEXI(supervised learning)HEWUJIANDUXUEXI(unsupervised learning)。

北京pk10赛车开奖记录结果 SHENDUXUEXILILUNJICHUSHIJIQIXUEXIZHONGDEFENSANBIAOSHI(distributed representation)。FENSANBIAOSHIJIADINGGUANCEZHISHIYOUBUTONGYINZIXIANGHUZUOYONGSHENGCHENG。ZAITANJIUZHEIZHONGXIANGHUZUOYONGDEGUOCHENGZHONG,SHENDUXUEXISHOUDAORENLEISHIJUEYUANLIDEQIFA。RENLEIDESHIJUEYUANLIRUXIA:CONGYUANSHIXINHAOSHERUKAISHI(TONGKONGSHERUXIANGSU),JIEZHEZUOCHUBUCHULI(DANAOPICENGMOUXIEXIBAOFAXIANBIANYUANHEFANGXIANG),RANHOUCHOUXIANG(DANAOPANDING,YANQIANDEWUTIDEXINGZHUANG,SHIYUANXINGDE),RANHOUJINYIBUCHOUXIANG(DANAOJINYIBUPANDINGGAIWUTISHIZHIQIQIU)。YINER,SHENDUXUEXIYECAIYONGZHUCENGYICIJINXING,ZHUBUFANHUACHOUXIANGDEJIBENJIEGOU:JIADINGBUTONGYINZIXIANGHUZUOYONGDEGUOCHENGKEFENWEIDUOGECENGCI,DAIBIAODUIGUANCEZHIDEDUOCENGCHOUXIANG。BUTONGDECENGSHUHECENGDEGUIMOKEYONGYUBUTONGCHENGDUDECHOUXIANG。GENGGAOCENGCIDEGAINIANCONGDICENGCIDEGAINIANXUEXIDEDAO。ZHEIYIFENCENGJIEGOUCHANGCHANGSHIYONGTANLANSUANFAZHUCENGGOUJIANERCHENG,BINGCONGZHONGXUANQUYOUZHUYUJIQIXUEXIDEGENGYOUXIAODETEZHENG。

基于深度学习的微表情识别工作流程包括以下四个步骤:
1) 准备数据集:包含微表情的视频片段采集、视频图像归一化处理、训练/验证/测试集分割等;
2) 设计学习模型:选择基本模型框架为卷积神经网络CNN+循环神经网络RNN、调整网络层数、确定损失函数、设计学习率等超参数;
3) 训练模型:将模型输出误差通过BP算法反向传播,利用随机梯度下降SGD或Adam算法优化模型参数;
北京pk10赛车开奖记录结果 4) 验证模型:利用未训练的数据验证模型的泛化能力,如果预测结果不理想,则需要重新设计模型,进行新一轮的训练;

至今已有数种成熟的深度学习模型,包括深度神经网络DNN、卷积神经网络CNN和深度置信网络DBN和递归神经网络RNN等。在语音识别、机器视觉、自然语言处理、生物信息学等领域得到广泛应用、并且取得了显著效果。
微表情分析是目前极具前瞻性的研究领域,人工智能深度学习模型的引入,较大提升了微表情识别性能,也将加速该领域的应用进展。但是,由于深度学习的黑盒特性,难以对微表情识别的特征提取过程进行定性研究,为此,仍需要进一步加强对深度学习模型的可视化技术研究,提高学习模型的可靠性分析并在可解释性的基础上进一步提高微表情识别准确度。

  1. M. Shreve, S. Godavarthy, V. Manohar, D. Goldgof, S. Sarkar, Towards macro-andmicro-expression spotting in video using strain patterns, in: Applications of Computer Vision (WACV), 2009 Workshop on, IEEE, 2009, pp. 1–6.
  2. R. Huang, S. Zhang, T. Li, and R. He. Beyond face rotation: Global and local perception gan for photorealistic and identity preserving frontal view synthesis. arXiv:1704.04086, 2017. 2
  3. S. Polikovsky, Y. Kameda, Y. Ohta, Facial micro-expressions recognition using high speed camera and 3D-gradient descriptor, in: 3rd International Conference on Crime Detection and Prevention (ICDP 2009). IET, 2009, pp. 1–6.
  4. T. Pfister, X. Li, G. Zhao, et al., Recognising spontaneous facial microexpressions, in: 2011 IEEE International Conference on Computer Vision (ICCV).IEEE, 2011, pp. 1449–1456.
  5. Yan W J, Wu Q, Liu Y J, Wang S J, Fu X L. CASME database: a dataset of spontaneous micro-expressions collected from neutralized faces. In: Proceedings of the 10th IEEE International Conference and Workshops on Automatic Face and Gesture Recognition. Shanghai, China: IEEE, 2013. 1-7

ZHUANZAIZIRENGONGZHINENGKAIFANGCHUANGXINPINGTAI

SAOYISAO,GUANZHUWOMEN

pk10网站投注可靠的 北京pk10直播视频直播 北京赛pk10怎么看走势 极速pk10彩票官网网址 pk10官方开奖直播网站 北京赛pk10app下载 pk10赛车冠军计算公式 北京pk10走势图 pk10技巧图解 pk10一分赛车开奖记录 pk10官方开奖直播现场 北京赛pk10群 北京pk10彩票官网 pk10官网开奖直播现场 pk10手机投注网站 pk10赛车历史开奖记录彩 北京塞车pk10官网开奖号码 北京pk10直播开奖走势图 北京pk10赛车开奖记录结果查询 北京赛pk10车网站