Research on human action recognition based on convolutional neural network

Peng Wang; Yuliang Yang; Wanchong Li; Linhao Zhang; Mengyuan Wang; Xiaobo Zhang; Mengyu Zhu

doi:10.1109/WOCC.2019.8770575

Research on human action recognition based on convolutional neural network

Peng Wang, Yuliang Yang^*, Wanchong Li, Linhao Zhang, Mengyuan Wang, Xiaobo Zhang, Mengyu Zhu

^*Corresponding author for this work

School of Medical and Technology

University of Science and Technology Beijing

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

3 Citations (Scopus)

Abstract

This paper proposes a human action recognition (HAR) algorithm based on convolutional neural network, which is used for human semaphore motion recognition. First, collecting datas in three scenarios and Deep Convolution Generative Adversarial Networks(DCGAN) is used to implement data enhancement to generate the dataset (DataSR). Then, the 1∗1 and 3∗3 convolution kernels are used to design the full convolution network and the model is further compressed using the group convolution to obtain the new model HARNET. Experiments show that the mAP of HARNET on the DataSR dataset is 94.36%, and the model size is 76M, which is 30% of the size of the YOLOv3 model.

Original language	English
Title of host publication	2019 28th Wireless and Optical Communications Conference, WOCC 2019 - Proceedings
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9781728106601
DOIs	https://doi.org/10.1109/WOCC.2019.8770575
Publication status	Published - May 2019
Event	28th Wireless and Optical Communications Conference, WOCC 2019 - Beijing, China Duration: 9 May 2019 → 10 May 2019

Publication series

Name	2019 28th Wireless and Optical Communications Conference, WOCC 2019 - Proceedings

Conference

Conference	28th Wireless and Optical Communications Conference, WOCC 2019
Country/Territory	China
City	Beijing
Period	9/05/19 → 10/05/19

Keywords

DCGAN
DataSR
Group convolution
Human action recognition

Access to Document

10.1109/WOCC.2019.8770575

Cite this

Wang, P., Yang, Y., Li, W., Zhang, L., Wang, M., Zhang, X., & Zhu, M. (2019). Research on human action recognition based on convolutional neural network. In 2019 28th Wireless and Optical Communications Conference, WOCC 2019 - Proceedings Article 8770575 (2019 28th Wireless and Optical Communications Conference, WOCC 2019 - Proceedings). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/WOCC.2019.8770575

@inproceedings{b5dc7de9cdc148e6a7c9bdca086d1b67,

title = "Research on human action recognition based on convolutional neural network",

abstract = "This paper proposes a human action recognition (HAR) algorithm based on convolutional neural network, which is used for human semaphore motion recognition. First, collecting datas in three scenarios and Deep Convolution Generative Adversarial Networks(DCGAN) is used to implement data enhancement to generate the dataset (DataSR). Then, the 1∗1 and 3∗3 convolution kernels are used to design the full convolution network and the model is further compressed using the group convolution to obtain the new model HARNET. Experiments show that the mAP of HARNET on the DataSR dataset is 94.36%, and the model size is 76M, which is 30% of the size of the YOLOv3 model.",

keywords = "DCGAN, DataSR, Group convolution, Human action recognition",

author = "Peng Wang and Yuliang Yang and Wanchong Li and Linhao Zhang and Mengyuan Wang and Xiaobo Zhang and Mengyu Zhu",

note = "Publisher Copyright: {\textcopyright} 2019 IEEE.; 28th Wireless and Optical Communications Conference, WOCC 2019 ; Conference date: 09-05-2019 Through 10-05-2019",

year = "2019",

month = may,

doi = "10.1109/WOCC.2019.8770575",

language = "English",

series = "2019 28th Wireless and Optical Communications Conference, WOCC 2019 - Proceedings",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

booktitle = "2019 28th Wireless and Optical Communications Conference, WOCC 2019 - Proceedings",

address = "United States",

}

Wang, P, Yang, Y, Li, W, Zhang, L, Wang, M, Zhang, X & Zhu, M 2019, Research on human action recognition based on convolutional neural network. in 2019 28th Wireless and Optical Communications Conference, WOCC 2019 - Proceedings., 8770575, 2019 28th Wireless and Optical Communications Conference, WOCC 2019 - Proceedings, Institute of Electrical and Electronics Engineers Inc., 28th Wireless and Optical Communications Conference, WOCC 2019, Beijing, China, 9/05/19. https://doi.org/10.1109/WOCC.2019.8770575

Research on human action recognition based on convolutional neural network. / Wang, Peng; Yang, Yuliang; Li, Wanchong et al.
2019 28th Wireless and Optical Communications Conference, WOCC 2019 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2019. 8770575 (2019 28th Wireless and Optical Communications Conference, WOCC 2019 - Proceedings).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Research on human action recognition based on convolutional neural network

AU - Wang, Peng

AU - Yang, Yuliang

AU - Li, Wanchong

AU - Zhang, Linhao

AU - Wang, Mengyuan

AU - Zhang, Xiaobo

AU - Zhu, Mengyu

PY - 2019/5

Y1 - 2019/5

N2 - This paper proposes a human action recognition (HAR) algorithm based on convolutional neural network, which is used for human semaphore motion recognition. First, collecting datas in three scenarios and Deep Convolution Generative Adversarial Networks(DCGAN) is used to implement data enhancement to generate the dataset (DataSR). Then, the 1∗1 and 3∗3 convolution kernels are used to design the full convolution network and the model is further compressed using the group convolution to obtain the new model HARNET. Experiments show that the mAP of HARNET on the DataSR dataset is 94.36%, and the model size is 76M, which is 30% of the size of the YOLOv3 model.

AB - This paper proposes a human action recognition (HAR) algorithm based on convolutional neural network, which is used for human semaphore motion recognition. First, collecting datas in three scenarios and Deep Convolution Generative Adversarial Networks(DCGAN) is used to implement data enhancement to generate the dataset (DataSR). Then, the 1∗1 and 3∗3 convolution kernels are used to design the full convolution network and the model is further compressed using the group convolution to obtain the new model HARNET. Experiments show that the mAP of HARNET on the DataSR dataset is 94.36%, and the model size is 76M, which is 30% of the size of the YOLOv3 model.

KW - DCGAN

KW - DataSR

KW - Group convolution

KW - Human action recognition

UR - http://www.scopus.com/inward/record.url?scp=85070362658&partnerID=8YFLogxK

U2 - 10.1109/WOCC.2019.8770575

DO - 10.1109/WOCC.2019.8770575

M3 - Conference contribution

AN - SCOPUS:85070362658

T3 - 2019 28th Wireless and Optical Communications Conference, WOCC 2019 - Proceedings

BT - 2019 28th Wireless and Optical Communications Conference, WOCC 2019 - Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 28th Wireless and Optical Communications Conference, WOCC 2019

Y2 - 9 May 2019 through 10 May 2019

ER -

Wang P, Yang Y, Li W, Zhang L, Wang M, Zhang X et al. Research on human action recognition based on convolutional neural network. In 2019 28th Wireless and Optical Communications Conference, WOCC 2019 - Proceedings. Institute of Electrical and Electronics Engineers Inc. 2019. 8770575. (2019 28th Wireless and Optical Communications Conference, WOCC 2019 - Proceedings). doi: 10.1109/WOCC.2019.8770575

Research on human action recognition based on convolutional neural network

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this