摘 要：文章对步态识别的应用进行研究，基于深度学习技术研究开发了移动端步态识别系统。手机客户端主要完成步态数据的采集、上传以及结果显示，服务器端负责对步态数据进行轮廓提取、步态匹配与识别等功能。其中步态轮廓提取采用 DeepLabV3+ 语义分割模型，实现像素级别的轮廓分割；步态识别采用 GaitSet 模型，实现人体步态匹配。系统分别经CASIA-B 数据集和真实场景进行测试，显示系统能够获得较好的性能，准确率达到 77.5%。
关键词：步态识别；深度学习；轮廓提取；语义分割；手机摄像头
DOI:10.19850/j.cnki.2096-4706.2021.13.016
中图分类号：TP183 文献标识码：A 文章编号：2096-4706（2021）13-0063-06
Research on Gait Recognition Algorithm Based on Deep Learning and Its Application
ZHAO Xuhua 1 , HU Haigen2
(1. College of Information, Zhejiang Guangsha Vocational and Technical University of Construction, Dongyang 322100, China; 2. College of Computer Science and Technology, Zhejiang University of Technology, Hangzhou 310024, China)
Abstract: This paper studies the application of gait recognition, and develops a mobile terminal gait recognition system based on deep learning technology. The mobile client mainly completes the gait data collection, uploading and result displaying, while the server side is responsible for contour extraction, gait matching and recognition of gait data. The gait contour extraction adopts DeepLabV3+ semantic segmentation model to realize pixel level contour segmentation; Gaitset model is used for gait recognition to realize human gait matching. The system has been tested by CASIA-B data set and real scenarios respectively. It shows that the system can obtain good performance, and the accuracy rate can reach 77.5%.
Keywords: gait recognition; deep learning; contour extraction; semantic segmentation; phone camera
参考文献：
[1] CONNOR P，ROSS A. Biometric recognition by gait： A survey of modalities and features [J].Computer Vision and Image Understanding，2018，167：1-27.
[2] 朱应钊，李嫚 . 步态识别现状及发展趋势 [J]. 电信科学， 2020，36（8）：130-138.
[3] JU H，BIR B. Individual recognition using gait energy image [J].IEEE transactions on pattern analysis and machine intelligence， 2006，28（2）：316-322.
[4] 冯世灵，王修晖 . 结合非局部与分块特征的跨视角步态识 别 [J]. 模式识别与人工智能，2019，32（9）：821-827.
[5] 胡靖雯，李晓坤，陈虹旭，等 . 基于深度学习的步态识别 方法 [J]. 计算机应用，2020，40（S1）：69-73.
[6] CHAO H，HE Y，ZHANG J，et al. GaitSet：Regarding gait as a set for cross-view gait recognition [C]//Proceedings of the 33th AAAI Conference on Artificial Intelligence，2019：8126-8133.
[7] 贲晛烨，徐森，王科俊 . 行人步态的特征表达及识别综述 [J]. 模式识别与人工智能，2012，25（1）：71-81.
[8] 汪涛，汪泓章，夏懿，等.基于卷积神经网络与注意力模型的人体 步态识别 [J].传感技术学报，2019，32（7）：1027-1033.
[9] 张馨心，姚爱琴，孙运强，等 . 基于深度学习的步态识别 算法优化研究 [J]. 自动化仪表，2020，35（4）：70-74.
[10] 周志一，宋冰，段鹏松，等 . 基于 WiFi 信号的轻量级步 态识别模型 LWID [J]. 计算机科学，2020，47（11）：25-31.
[11] SHI C，LIU J，LIU H，et al. Smart user authentication through actuation of daily activities leveraging WiFi-enabled IoT [C]// Proceedings of the 18th ACM international symposium on Mobile Ad Hoc Networking and Computing. Chennai，India：Association for Computing Machinery，2017：1–10.
[12] WANG W，LIU A X L，SHAHZAD M. Gait recognition using WiFi signals [C]//Proceedings of the 18th ACM International Symposium on Mobile Ad Hoc Networking and Computing. Heidelberg Germany：Association for Computing Machinery，2016：363–373.
[13] ZENG Y，PATHAK P H，MOHAPATRA P. WiWho：wifibased person identification in smart spaces [C]//Proceedings of the 15th International Conference on Information Processing in Sensor Networks. Vienna Austria：IEEE Press，2016：1-12.
[14] ZHANG J，WEI B，HU W，et al. Wifi-id：human identifica-tion using wifi signal [C]//Proceedings of the International Conference on Distributed Computing in Sensor Systems. Washington， DC，USA：2016 International Conference on Distributed Computing in Sensor Systems （DCOSS），2016：75-82.
[15] CHEN L C，ZHU Y，PAPANDREOU G，et al. Encoderdecoder with atrous separable convolution for semantic image segmentation [C]//Proceedings of the European conference on computer vision （ECCV）. Munich，Germany：Computer Vision – ECCV 2018，2018：833-851.
[16] CHEN L C，PAPANDREOU G，SCHROFF F，et al. Rethinking atrous convolution for semantic image segmenta-tion [J/OL].arXiv. org，2017，3（2017-06-17）.https：//arxiv.org/abs/1706.05587v1.
[17] LIPTON A J，FUJIYOSHI H，PATIL R S. Moving target classification and tracking from real-time video [C]//Proceeding of Fourth IEEE Workshop on Applications of Computer Vision. Princeton， NJ，USA：IEEE，1998：8-14.
[18] 商磊，张宇，李平 . 基于密集光流的步态识别 [J]. 大连 理工大学学报，2016，56（2）：214-220.
作者简介：赵煦华（1973.12—），男，汉族，浙江东阳人， 中级工程师，硕士研究生，研究方向：嵌入式系统、大数据、人工 智能、机器学习。