|本期目录/Table of Contents|

[1]林智伟,朱文章,陈浩.差分特征融合改进的动态手势识别分类网络模型[J].厦门理工学院学报,2021,29(1):35-42.[doi:1019697/jcnki16734432202101006]
 LIN Zhiwei,ZHU Wenzhang,CHEN Hao.Network Modeling of Dynamic Gesture RecognitionClassification Improved on Difference Feature Fusion[J].Journal of JOURNAL OF XIAMEN,2021,29(1):35-42.[doi:1019697/jcnki16734432202101006]
点击复制

差分特征融合改进的动态手势识别分类网络模型(PDF/HTML)
分享到:

《厦门理工学院学报》[ISSN:1673-4432/CN:35-1289/Z]

卷:
29
期数:
2021年第1期
页码:
35-42
栏目:
光电与通信工程
出版日期:
2021-02-28

文章信息/Info

Title:
Network Modeling of Dynamic Gesture Recognition Classification Improved on Difference Feature Fusion
文章编号:
16734432(2021)01003508
作者:
林智伟朱文章陈浩
厦门理工学院光电与通信工程学院,福建 厦门 361024
Author(s):
LIN Zhiwei ZHU Wenzhang CHEN Hao
School of Optoelectronics and Communication Engineering, Xiamen University of Technology, Xiamen 361024, China
关键词:
动态手势识别分类网络模型差分特征融合卷积神经网络长短期记忆网络
Keywords:
dynamic gesture recognition classificationnetwork modeldifferential feature fusionconvolutional neural networklong and short term memory network
分类号:
TN91173TP3914
DOI:
1019697/jcnki16734432202101006
文献标志码:
A
摘要:
通过卷积神经网络和长短期记忆网络进行多模型结合,实现动态手势识别分类建模,并使用数据增强算法增加数据的多样性,通过差分特征融合改进网络。7种动态手势动作识别分类的实验结果显示,使用数据增强算法增加数据的多样性后,结合模型的识别率最佳可提升286%;通过差分算法改进网络,序列间差分特征融合模型识别率达到8381%,维度差分特征融合模型识别率达到8762%。表明多模型结合可解决单一模型的局限性,处理更加复杂的动态手势分类问题,两种不同形式的差分特征融合改进都可提升动态手势动作的识别率,从而验证了所设计的差分特征融合改进的动态手势识别分类网络模型的有效性和可行性。
Abstract:
Convolutional neural network and long short term memory network are combined to realize dynamic gesture recognition classification modeling.Data enhancement algorithm is used to increase the diversity of data,and differential feature fusion applied to improve the network.Experimental results of 7 kinds of dynamic gesture recognition classification show that:with the data diversity increase by data enhancement algorithm,recognition rate of the enhanced model is improved by 2.86%by using differential algorithm,the recognition rate of the sequence differential feature fusion model reaches 83.81%,and that of the dimensional differential feature fusion model reaches 87.62%.Multimodels can be applied to break the limits of single model to solve difficult dynamic gesture classification problems.Two different forms of differential feature fusion can both improve the recognition rate of dynamic gesture recognition,which proves the effectiveness and feasibility of the network modeling of dynamic gesture recognition classification.

参考文献/References:

[1] LIU H Y,WANG L H.Gesture recognition for human robot collaboration:a review[J].International Journal of Industrial Ergonomics,2018,68:355367. [2] 刘妍.基于视频的多目标运动人体动作识别[D].上海:东华大学,2017:1012. [3] ZHANG G,ZHANG D.Research on visionbased multiuser gesture recognition HumanComputer Interaction [C]∥IEEE. 7th International Conference on System Simulation and Scientific Computing.Beijing:IEEE,2008:1 4551 458. [4] 胡友树.手势识别技术综述[J].中国科技信息,2005(2):.4142. [5] GRIMES G J.Digital data entry glove interface device:US.4414537 [P/OL].19830110[20200901].http://www.freepatentsonline.com/4414537.html. [6] LEE H K,KIM J H.An HMMbased threshold model approach for gesture recognition[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2002,21(10):961973. [7] 李文生,姚琼,邓春健.粒子群优化神经网络在动态手势识别中的应用[J].计算机工程与科学,2011,33(5):7479. [8] 刘杨俊武.实时动态手势识别技术的研究与应用[D].南京:南京邮电大学,2017. [9] 操小文.基于卷积神经网络的手势识别研宄[J].微型机与应用,2016,35(9):5557. [10] 范长军,高飞.基于可穿戴传感器的普适化人体活动识别[J].传感技术学报,2018,31(7):1 1241 131. [11] 付天豪,于力革.改进卷积神经网络的动态手势识别[J].计算机系统应用,2020,29(9):225230. [12] JIAN C F,XIANG X Y,ZHANG M Y.Mobile terminal gesture recognition based on improved FAST corner detection[J].IET Image Processing,2019,13(6):991997. [13] HE K,ZHANG X,REN S,et al.Deep residual learning for image recognition[C]∥IEEE Conference on Computer Vision and Pattern Recognition.Las Vegas:CVPR,2015:4145. [14] PICKERING C A.The search for a safer driver interface:a review of gesture recognition human machine interface[J].Computing & Control Engineering Journal,2005,16(1):3440. [15] PIGOU L,VAN DEN ORAD A,DIELEMAN S,et al.Beyond temporal pooling:recurrence and temporal convolutions for geture recognition in ivdeo[J].International Journal of Computer Vision,2018,126(2/4):430439. [16] HUANG Z H,XU W,YU K.Bidirectional LSTMCRF models for sequence tagging[J/OL].Computer Science,2015:110[20200901].https:∥arxiv.org/abs/1508.01991. [17] IOFFE S,SZEGEDY C.Batch normalization:accelerating deep network training by reducing internal covariate shift[C]∥ICML’2015.Proceedings of the 32nd International Conference on International Conference on Machine Learning.lile,Frarce:ICML2015,2015,37:448456. [18] 莫伟珑.基于计算机视觉的手势识别方法研究[D].南宁:广西师范大学,2019:3133.

相似文献/References:

备注/Memo

备注/Memo:
收稿日期:20200930修回日期:20201029 基金项目:厦门市科技计划重大项目(3502ZCQ20191002) 通信作者:朱文章,男,教授,博士,研究方向为物联网、半导体照明、半导体器件,Email:wzzh@xmut.edu.cn。
更新日期/Last Update: