Abstract
In this work, we present two new procedures for activity recognition that are based on the Fourier frequencies that are generated when the optical flow values of successive frames of video are processed simultaneously. In the first algorithm, we correlate these 2D Doppler Fourier spectra with the mean spectra of each activity class. These correlation vectors, which include only 30 features in number, are categorized using a reduced robust SVM classification model. This first procedure is of low computational cost for action recognition tasks for numerable activity classes. For large numbers of activity classes, we propose a new method of aggregated weighted spectra of optical flow values across the whole video. The above-mentioned Fourier spectra are concatenated with a short vector representing the distributions of the moving edges. These methods are insensitive to the presence of background as well as to the positions of the subjects and their shapes and can encode the information of a part or of the whole of a video into relatively short vectors. The results of the two procedures seem to be competitive to state-of-the-art action recognition methods when tested on the KTH Royal Institute Database and on the UCF101 Database for action recognition tasks.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsReferences
P. Dollar, V. Rabaud, G. Cottrel, S. Belongie, Behavior “recognition via spatiotemporal features”, in Proceedings 2nd Joint IEEE International Workshop on VS-PETS, Beijing (IEEE Computer Society Press, Los Alamitos, 2005), pp. 65–72
L. Wang et al., Fusion of static and dynamic body biometrics for gait recognition. IEEE Trans. Circ. Syst. Video Technol. 14(2), 149–158 (2004)
A.A. Efros, et al., Recognizing action at a distance, in ICCV, vol. 3 (2003)
J.C. Niebles, H. Wang, L. Fei-fei: Unsupervised learning of human action categories using spatial-temporal words, in BMVC (2006)
L.P. Wang, X.J. Fu, Data Mining with Computational Intelligence (Springer, Berlin, 2005)
X.J. Fu, L.P. Wang, Data dimensionality reduction with application to simplifying RBF network structure and improving classification performance. IEEE Trans. Syst. Man Cybern Part B Cybern 33(3), 399–409 (2003)
L.P. Wang, On competitive learning. IEEE Trans. Neural Networks 8(5), 1214–1217 (1997)
V. Luong, L.P. Wang, G. Xiao, Deep networks with trajectory for action recognition in videos, in The 18th Asia Pacific Symposium on Intelligent and Evolutionary Systems (IES 2014), Singapore, 10–12th Nov 2014
B.K.P. Horn, B.G. Schunck, Determining optical flow. Artif. Intell. 17(1–3), 185–203 (1981)
S. Danafar, N. Gheissari, Action recognition for surveillance applications using optic flow and SVM, in Asian Conference on Computer Vision (Springer, Berlin, 2007)
C. Schuldt, I. Laptev, B. Caputo, Recognizing human actions: a local SVM approach, in ICPR (2004), pp. 32–36
K. Soomro, A.R. Zamir and M. Shah, UCF101: a dataset of 101 human action classes from videos in the wild. CRCV-TR-12-01, November, 2012.
M. Hubert, P.J. Rousseeuw, K. Vanden Branden, ROBPCA: a new approach to robust principal components analysis. Technometrics 47, 64–79 (2005)
M. Hubert, S. Engelen, Fast cross-validation of high-breakdown resampling algorithms for PCA. Comput. Stat. Data Anal. 51, 5013–5024 (2007)
M. Hubert, P.J. Rousseeuw, T. Verdonck, Robust PCA for skewed data and its outlier map. Comput. Stat. Data Anal. 53, 2264–2274 (2009)
S. Serneels, T. Verdonck, Principal component analysis for data containing outliers and missing elements. Comput. Stat. Data Anal. 52, 1712–1727 (2008)
P. Ahammad, C. Yeo, S.S. Sastry, K. Ramchandran, Compressed domain real-time action recognition, MMSP, in Proceedings of 8th IEEE Workshop on Multimedia Signal Processing (IEEE Computer Society Press, Los Alamitos, 2006) pp. 33–36
K. Simonyan, A. Zisserman, Two-stream convolutional networks for action recognition in videos, in Advances in Neural Information Processing Systems (2014)
L. Wang, et al., Temporal segment networks: towards good practices for deep action recognition, in European Conference on Computer Vision (Springer International Publishing, 2016)
Acknowledgements
The KTH Royal Institute of Technology video database [11] that was used is publicly available for non-commercial use.
The UCF101, Human Activity Database [12] is freely available here: https://www.crcv.ucf.edu/data/UCF101.php.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2020 Springer Nature Singapore Pte Ltd.
About this paper
Cite this paper
Pavlidou, M., Zioutas, G. (2020). A New Use of Doppler Spectrum for Action Recognition with the Help of Optical Flow. In: Yang, XS., Sherratt, S., Dey, N., Joshi, A. (eds) Fourth International Congress on Information and Communication Technology. Advances in Intelligent Systems and Computing, vol 1041. Springer, Singapore. https://doi.org/10.1007/978-981-15-0637-6_35
Download citation
DOI: https://doi.org/10.1007/978-981-15-0637-6_35
Published:
Publisher Name: Springer, Singapore
Print ISBN: 978-981-15-0636-9
Online ISBN: 978-981-15-0637-6
eBook Packages: EngineeringEngineering (R0)