首页> 外国专利> METHOD AND APPARATUS FOR ESTIMATING HUMAN EMOTION BASED ON ADAPTIVE IMAGE RECOGNITION USING INCREMENTAL DEEP LEARNING

METHOD AND APPARATUS FOR ESTIMATING HUMAN EMOTION BASED ON ADAPTIVE IMAGE RECOGNITION USING INCREMENTAL DEEP LEARNING

机译:基于深度学习的自适应图像识别的人类情绪估计方法及装置

摘要

An adaptive image recognition-based emotion estimation apparatus using gradual deep learning learning according to an embodiment of the present invention includes: a face detection unit for detecting a face region of a person in an image including a person subject to emotion estimation; A body detection unit for detecting a person''s body region in an image including a person to be subjected to the emotion estimation; A facial expression recognition unit for recognizing a facial expression by using the detected image of the face region as an input; A behavior recognition unit for recognizing a person''s behavior by using the detected image of the body region as an input; And based on the recognized facial expressions and behaviors, including an emotion determination unit for judging a person''s sensibility, it is possible to quickly recover the reduced recognition rate according to changes in the environment with less effort, thereby recognizing the image robustly against the environment change You can do it, and gradually learn a person''s behavior and facial expressions with little data through a gradual deep learning method to accurately guess a person''s sensibility through a combination of the two results.
机译:根据本发明的实施例的使用渐进深度学习学习的基于自适应图像识别的情绪估计装置包括:面部检测单元,用于检测图像中的人的面部区域,该图像包括要进行情绪估计的人;以及身体检测单元,用于检测包括要进行情感估计的人的图像中的人的身体区域;面部表情识别单元,用于通过使用检测到的面部区域的图像作为输入来识别面部表情;行为识别单元,通过使用检测到的身体区域的图像作为输入来识别人的行为;并且基于识别的面部表情和行为,包括用于判断人的敏感度的情绪确定单元,可以根据环境的变化而迅速地恢复恢复降低的识别率,从而花费更少的精力,从而可靠地识别图像。环境变化您可以做到这一点,并通过渐进式深度学习方法以很少的数据逐渐了解一个人的行为和面部表情,以结合这两个结果准确地猜测一个人的敏感度。

著录项

  • 公开/公告号KR102132407B1

    专利类型

  • 公开/公告日2020-07-10

    原文格式PDF

  • 申请/专利权人 인하대학교 산학협력단;

    申请/专利号KR20180140407

  • 发明设计人 이필규;김진우;

    申请日2018-11-15

  • 分类号G06N3/08;G06N3/04;

  • 国家 KR

  • 入库时间 2022-08-21 11:04:18

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号