Abstract—Sports video detection and analysis is a difficult
problem in the automatic acquisition of technical and tactical,
and also a key issue in whether the collection work can be
carried out quickly and efficiently. In this paper, a two-stream
deep convolution network model for racket motion recognition
and a set of algorithms for racket motion recognition are
proposed. The proposed model uses the basic network
structure combining Batch Normalization (BN) and Inception
network. The idea of batch normalization transformation is
used in the recognition of the racket motion, the mini-batch
normalization of the two-stream convolution network input is
realized. This operation can speed up the training of the
network model. The video detection and analysis algorithm
based on the above model is given. The algorithm uses RGB
image as the input of the spatial network, and the optical flow
field image is used as the input of the temporal network. The
SVM model is used to fuse the spatio-temporal network to
obtain the final detection result. Based on the UCF101
standard data set, both the table tennis and badminton two
action video were tested, and the accuracy of 66.92 was
obtained. Experiments were conducted on the table tennis and
badminton data of the UCF101 standard data set, achieving an
accuracy of 66.92%.
Index Terms—Racket motion recognition, batch normalization, convolution network, deep learning.
Huiqun Zhao was with Computer School, North China University of Technology, Beijing 100144 China (e-mail: email@example.com.
Wei Ye was with College of Computer, North China University of Technology, Beijing, 100144 China (e-mail: firstname.lastname@example.org).
Cite: Zhao Hui-Qun and Ye Wei, "Racket Motion Recognition Method Based on Improved Two-Stream Convolution Network," International Journal of Machine Learning and Computing vol. 10, no. 4, pp. 556-561, 2020.Copyright © 2020 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).