Softmax regression based deep sparse autoencoder network for facial emotion recognition in human-robot interaction

被引:146
作者
Chen, Luefeng [1 ,2 ]
Zhou, Mengtian [1 ,2 ]
Su, Wanjuan [1 ,2 ]
Wu, Min [1 ,2 ]
She, Jinhua [1 ,2 ,3 ]
Hirota, Kaoru [4 ]
机构
[1] China Univ Geosci, Sch Automat, Wuhan 430074, Hubei, Peoples R China
[2] Hubei Key Lab Adv Control & Intelligent Automat C, Wuhan 430074, Hubei, Peoples R China
[3] Tokyo Univ Technol, Sch Engn, Tokyo 1920982, Japan
[4] Tokyo Inst Technol, Yokohama, Kanagawa 2268502, Japan
基金
中国国家自然科学基金;
关键词
Facial emotion recognition; Deep sparse autoencoder network; Softmax regression; Human-robot interaction; EXPRESSION RECOGNITION; COMMUNICATION ATMOSPHERE; MULTIROBOT BEHAVIOR; REPRESENTATION;
D O I
10.1016/j.ins.2017.10.044
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Deep neural network (DNN) has been used as a learning model for modeling the hierarchical architecture of human brain. However, DNN suffers from problems of learning efficiency and computational complexity. To address these problems, deep sparse autoencoder network (DSAN) is used for learning facial features, which considers the sparsity of hidden units for learning high-level structures. Meanwhile, Softmax regression (SR) is used to classify expression feature. In this paper, Softmax regression-based deep sparse autoencoder network (SRDSAN) is proposed to recognize facial emotion in human-robot interaction. It aims to handle large data in the output of deep learning by using SR, moreover, to overcome local extrema and gradient diffusion problems in the training process, the overall network weights are fine-tuned to reach the global optimum, which makes the entire depth of the neural network more robust, thereby enhancing the performance of facial emotion recognition. Results show that the average recognition accuracy of SRDSAN is higher than that of the SR and the convolutional neural network. The preliminarily application experiments are performed in the developing emotional social robot system (ESRS) with two mobile robots, where emotional social robot is able to recognize emotions such as happiness and angry. (C) 2017 Published by Elsevier Inc.
引用
收藏
页码:49 / 61
页数:13
相关论文
共 44 条
[1]   Deep learning approach for active classification of electrocardiogram signals [J].
Al Rahhal, M. M. ;
Bazi, Yakoub ;
AlHichri, Haikel ;
Alajlan, Naif ;
Melgani, Farid ;
Yager, R. R. .
INFORMATION SCIENCES, 2016, 345 :340-354
[2]   Towards an intelligent system for generating an adapted verbal and nonverbal combined behavior in human-robot interaction [J].
Aly, Amir ;
Tapus, Adriana .
AUTONOMOUS ROBOTS, 2016, 40 (02) :193-209
[3]  
An S., 2007, P IEEE C COMP VIS PA, DOI 10.1109/CVPR.2007.383105.
[4]  
[Anonymous], IEEE T CYBERN
[5]  
[Anonymous], INF SCI
[6]  
[Anonymous], 1998, JAPANESE FEMALE FACI
[7]   Emotion-Age-Gender-Nationality Based Intention Understanding in Human-Robot Interaction Using Two-Layer Fuzzy Support Vector Regression [J].
Chen, Lue-Feng ;
Liu, Zhen-Tao ;
Wu, Min ;
Ding, Min ;
Dong, Fang-Yan ;
Hirota, Kaoru .
INTERNATIONAL JOURNAL OF SOCIAL ROBOTICS, 2015, 7 (05) :709-729
[8]   Adapting Multi-Robot Behavior to Communication Atmosphere in Humans-Robots Interaction Using Fuzzy Production Rule Based Friend-Q Learning [J].
Chen, Lue-Feng ;
Liu, Zhen-Tao ;
Dong, Fang-Yan ;
Yamazaki, Yoichi ;
Wu, Min ;
Hirota, Kaoru .
JOURNAL OF ADVANCED COMPUTATIONAL INTELLIGENCE AND INTELLIGENT INFORMATICS, 2013, 17 (02) :291-301
[9]   Multi-robot behavior adaptation to local and global communication atmosphere in humans-robots interaction [J].
Chen, Lue-Feng ;
Liu, Zhen-Tao ;
Wu, Min ;
Dong, Fang-Yan ;
Yamazaki, Yoichi ;
Hirota, Kaoru .
JOURNAL ON MULTIMODAL USER INTERFACES, 2014, 8 (03) :289-303
[10]   Adaptive appearance model tracking for still-to-video face recognition [J].
Dewan, M. Ali Akber ;
Granger, E. ;
Marcialis, G. -L. ;
Sabourin, R. ;
Roli, F. .
PATTERN RECOGNITION, 2016, 49 :129-151