Max-pooling convolutional neural network for Chinese digital gesture recognition

Zhao Qian, Li Yawei, Zhu Mengyu*, Yang Yuliang, Xiao Ling, Xu Chunyu, Li Lin

*此作品的通讯作者

科研成果: 书/报告/会议事项章节会议稿件同行评审

摘要

A pattern recognition approach is proposed for the Chinese digital gesture. We shot a group of digital gesture videos by a monocular camera. Then, the video was converted into frame format and turned into the gray image. We selected the gray image as our own dataset. The dataset was divided into six gesture classes and other meaningless gestures. We use the neural network (NN) combining convolution and Max-Pooling (MPCNN) for classification of digital gestures. The MPCNN presents some differences on the data preprocessing, the activation function and the network structure. The accuracy and the robustness have been verified by the simulation experiments with the dataset. The result shows that the MPCNN classifies six gesture classes with 99.98% accuracy using the Max-Pooling, the Relu activation function, and the binarization processing.

源语言英语
主期刊名Information Technology and Intelligent Transportation System - Volume 2, Proceedings of the International Conference on Information Technology and Intelligent Transportation Systems, ITITS 2015
编辑Lakhmi C. Jain, Xiangmo Zhao, Valentina Emilia Balas
出版商Springer Verlag
79-89
页数11
ISBN(印刷版)9783319387697
DOI
出版状态已出版 - 2017
活动International Conference on Information Technology and Intelligent Transportation Systems, ITITS 2015 - Xi’an, 中国
期限: 12 12月 201513 12月 2015

出版系列

姓名Advances in Intelligent Systems and Computing
455
ISSN(印刷版)2194-5357

会议

会议International Conference on Information Technology and Intelligent Transportation Systems, ITITS 2015
国家/地区中国
Xi’an
时期12/12/1513/12/15

指纹

探究 'Max-pooling convolutional neural network for Chinese digital gesture recognition' 的科研主题。它们共同构成独一无二的指纹。

引用此