Abstract
Human action recognition (HAR) technology is currently of significant interest. The traditional HAR methods depend on the time and space of the video stream generally. It requires a mass of training datasets and produces a long response time, failing to simultaneously meet the real-time interaction technical requirements-high accuracy, low delay, and low computational cost. For instance, the duration of a gymnastic action is as short as 0.2 s, from action capture to recognition, and then to the visualization of a three-dimensional character model. Only when the response time of the application system is short enough can it guide synchronous training and accurate evaluation. To reduce the dependence on the amount of video data and meet the HAR technical requirements, this paper proposes a three-stream long-short term memory (TS-CNN-LSTM) framework combining the CNN and LSTM networks. Firstly, human data of color, depth, and skeleton collected by Microsoft Kinect are used as input to reduce the sample sizes. Secondly, heterogeneous convolutional networks are established to reduce computing costs and elevate response time. The experiment results demonstrate the effectiveness of the proposed model on the NTU-RGB + D, reaching the best accuracy of 87.28% in the Cross-subject mode. Compared with the state-of-the-art methods, our method uses 75% of the training sample size, while the complexity of time and space only occupies 67.5% and 73.98% respectively. The response time of one set action recognition is improved by 0.90–1.61 s, which is especially valuable for timely action feedback. The proposed method provides an effective solution for real-time interactive applications which require timely human action recognition results and responses.
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13369-022-07236-z/MediaObjects/13369_2022_7236_Fig1_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13369-022-07236-z/MediaObjects/13369_2022_7236_Fig2_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13369-022-07236-z/MediaObjects/13369_2022_7236_Fig3_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13369-022-07236-z/MediaObjects/13369_2022_7236_Fig4_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13369-022-07236-z/MediaObjects/13369_2022_7236_Fig5_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13369-022-07236-z/MediaObjects/13369_2022_7236_Fig6_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13369-022-07236-z/MediaObjects/13369_2022_7236_Fig7_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13369-022-07236-z/MediaObjects/13369_2022_7236_Fig8_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13369-022-07236-z/MediaObjects/13369_2022_7236_Fig9_HTML.png)
Similar content being viewed by others
Code or Data Availability
The research in this paper used the NTU-RGB + D (or NTU-RGB + D 120) Action Recognition Dataset made available by the ROSE Lab at the Nanyang Technological University, Singapore. At the same time, thanks to the GitHub platform for releasing our results.
References
Aggarwal, J.K.; Ryoo, M.S.: Human activity analysis: A review. ACM Comput. Surv. 43(3), 1–43 (2011). https://doi.org/10.1145/1922649.1922653
Lee, J.; Ahn, B.: Real-time human action recognition with a low-cost RGB camera and mobile robot platform. Sens. (Basel, Switzerland). 20(10), 2886 (2020). https://doi.org/10.3390/s20102886
Johansson, G.: Visual motion perception. Sci. Am. 232(6), 76–88 (1975). https://doi.org/10.1038/scientificamerican0675-76
Dong, N.; Fang, F.; Xudong, M.: A human activity recognition method based on DBMM. Ind. Control Comput. 33(3), 12–14 (2020). https://doi.org/10.3969/j.issn.1001-182X.2020.03.005
Zhang, H.B.; Zhang, Y.X.; Zhong, B., et al.: A comprehensive survey of vision-based human action recognition methods. Sensors (Basel). (2020). https://doi.org/10.3390/s19051005
Dalal, N.; Triggs, B.: Histograms of oriented gradients for human detection. In: International Conference on Computer Vision & Pattern Recognition (CVPR ’05), Jun 2005, San Diego, United States, pp. 886–893. https://courses.cs.washington.edu/courses/cse576/12sp/notes/CVPR2005_HOG.pdf
Zhu, Y.; Zhang, Y.; Chen, J., et al.: An intelligent system based on human action control. China Sci. Technol. Inf. 1, 68–70 (2020). https://doi.org/10.3969/j.issn.1001-8972.2020.01.023
Pham, H.H.; Salmane, H.; Khoudour, L., et al.: A unified deep framework for joint 3D pose estimation and action recognition from a single RGB camera. Sensors (Basel) 20(7), 1825 (2020). https://doi.org/10.3390/s20071825
Dhiman, C.; Vishwakarma, D.K.: View-invariant deep architecture for human action recognition using two-stream motion and shape temporal dynamics. IEEE Trans. Image Process. 29, 3835–3844 (2020). https://doi.org/10.1109/TIP.2020.2965299
Kim, H.; Park, S.; Park, H., et al.: Enhanced action recognition using multiple stream deep learning with optical flow and weighted sum. Sens. (Basel). 20(14), 1 (2020). https://doi.org/10.3390/s20143894
Ali, S.; Shah, M., et al.: Human action recognition in videos using kinematic features and multiple instance learning. IEEE Trans. Pattern Anal. Mach. Intell. 32(2), 288–303 (2010). https://doi.org/10.1109/TPAMI.2008.284
Xue, F.; Ji, H.; Zhang, W.; Cao, Y.: Action recognition based on dense trajectories and human detection. In: 2018 IEEE International Conference on Automation, Electronics and Electrical Engineering (AUTEEE), pp. 340–343 (2018). https://doi.org/10.1109/AUTEEE.2018.8720753.
**-Ting, S.; Sheng, Y.; Yao, D., et al.: Human action recognition method based on deep learning. Comput. Eng. Des. 41(4), 304–307 (2020). https://doi.org/10.19734/j.issn.1001-3695.2018.05.0499
Panareda, B.P.; Iqbal, A.; Gall, J.: Open set domain adaptation for image and action recognition. IEEE Trans. Pattern Anal. Mach. Intell. 42(2), 413–429 (2020). https://doi.org/10.1109/tpami.2018.2880750
Ma, C.; Wang, Y.; Mao, Z.: Action recognition based on spatiotemporal dual flow fusion network and am softmax. Netw. Secur. Technol. Appl. 11, 47–50 (2019). https://doi.org/10.3969/j.issn.1009-6833.2019.11.027
Penghua, G.E.; Min, Z.; Hua, Y.U., et al.: Human action recognition based on two-stream independently recurrent neural network. Mod. Electron. Tech. 43(4), 137–141 (2020). https://doi.org/10.16652/j.issn.1004-373x.2020.04.035(InChinese)
Luvizon, D.; Picard, D.; Tabia, H.: Multi-task deep learning for real-time 3D human pose estimation and action recognition. IEEE Trans. Pattern Anal. Mach. Intell. 8(43), 27522764 (2020). https://doi.org/10.1109/TPAMI.2020.2976014
Yasin, H.; Hussain, M.; Weber, A.: Keys for action: an efficient keyframe-based approach for 3D action recognition using a deep neural network. Sensors (Basel). 20(8), 2226 (2020). https://doi.org/10.3390/s20082226
Chen, J.; Kong, J.; Sun, H., et al.: Spatiotemporal interaction residual networks with pseudo3D for video action recognition. Sensors (Basel). 20(11), 3126 (2020). https://doi.org/10.3390/s20113126
Meng, F.; Liu, H.; Liang, Y., et al.: Sample fusion network: an end-to-end data augmentation network for Skeleton-based human action recognition. IEEE Trans. Image Process. 28(11), 5281–5295 (2019). https://doi.org/10.1109/TIP.2019.2913544
Sun, Z.; Guo, X.; Li, W., et al.: Cooperative warp of two discriminative features for Skeleton based action recognition. J. Phys.: Conf. Ser. 1187, 42027 (2019). https://doi.org/10.1088/1742-6596/1187/4/042027
Ke, Q.; Bennamoun, M.; An, S., et al.: Learning clip representations for Skeleton-based 3D action recognition. IEEE Trans. Image Process. 27(6), 2842–2855 (2018). https://doi.org/10.1109/TIP.2018.2812099
Kim, D.; Kim, D.H.; Kwak, K.C.: Classification of K-Pop dance movements based on skeleton information obtained by a kinect sensor. Sens. (Basel). 17(6), 1261 (2017). https://doi.org/10.3390/s17061261
Xue-Chao, B.: Dance-specific action recognition based on spatial skeleton sequence diagram. Inf. Technol. 43(11), 16–19 (2019). https://doi.org/10.13274/j.cnki.hdzj.2019.11.004
Caetano, C.; Bremond, F.; Schwartz, W.R.: Skeleton image representation for 3D action recognition based on tree structure and reference joints. IEEE 1, 16–23 (2019). https://doi.org/10.1109/SIBGRAPI.2019.00011
Wen, Y.H.; Gao, L.; Fu, H., et al.: Graph CNNs with motif and variable temporal block for Skeleton-based action recognition. Proceedings of the AAAI Conference on Artificial Intelligence 33, 8989–8996 (2019). https://doi.org/10.1609/aaai.v33i01.33018989
Liu, J.; Shahroudy, A.; Xu, D., et al.: Skeleton-based action recognition using spatio-temporal LSTM network with trust gates. IEEE Trans. Pattern. Anal. Mach. Intell. 40(12), 3007–3021 (2018). https://doi.org/10.1109/TPAMI.2017.2771306
Min, S.; Lan, L.: Human movements recognition based on LSTM network model and front action view. J. Anqing Normal Univ. (Nat. Sci. Ed.) 26(1), 73–76 (2020). https://doi.org/10.13757/j.cnki.cn34-1328/n.2020.01.013
Donahue, J.; Hendricks, L.A.; Rohrbach, M., et al.: Long-term recurrent convolutional networks for visual recognition and description. IEEE Trans. Pattern Anal. Mach. Intell. 39(4), 677–691 (2017). https://doi.org/10.1109/TPAMI.2016.2599174
Wang, H.; Song, Z.; Li, W., et al.: A hybrid network for large-scale action recognition from RGB and depth modalities. Sensors (Basel). 20(11), 3305 (2020). https://doi.org/10.3390/s20113305
Wang, J.; Yu, L.C.; Lai, K.R., et al.: Tree-structured regional CNN-LSTM model for dimensional sentiment analysis. IEEE/ACM Trans. Audio Speech Language Process. 28, 581–591 (2019). https://doi.org/10.1109/TASKP.2019.2959251
Yenter, A.; Verma, A.: Deep CNN-LSTM with combined kernels from multiple branches for IMDb review sentiment analysis. 2017 IEEE 8th Annual Ubiquitous Computing, Electronics and Mobile Communication Conference (UEMCON), pp. 540–546 (2017).https://doi.org/10.1109/UEMCON.2017.8249013.
Yan, Z.; Chong-Chong, Y.U.; Han, L., et al.: Short-term traffic flow forecasting method based on CNN+LSTM. Comput. Eng. Des. 40(09), 1 (2019). https://doi.org/10.16208/j.issn1000-7024.2019.09.038
Yan, Z.; Yu, Z.; Han, L., et al.: Short term traffic flow prediction method based on CNN+LSTM. Comput. Eng. Des. 40(9), 2620–2624 (2019)
Mou, L.; Zhou, C.; Zhao, P., et al.: Driver stress detection via multimodal fusion using attention-based CNN-LSTM. Expert Syst. Appl. 173(12), 1193 (2021). https://doi.org/10.1016/j.eswa.2021.114693
Yu, T.; Chen, J.; Yan, N.; et al.: A Multi-Layer PaCrallel LSTM Network for Human Activity Recognition with Smartphone Sensors. In: 10th International Conference on Wireless Communications and Signal Processing (WCSP). IEEE. 1–6 (2018). https://doi.org/10.1109/WCSP.2018.8555945.
Gao, W.; Zhang, L.; Teng, Q., et al.: DanHAR: dual attention network for multimodal human activity recognition using wearable sensors. Appl. Soft Comput. (2021). https://doi.org/10.1016/j.asoc.2021.107728
Mutegeki, R.; Han, D.S.: A CNN-LSTM Approach to Human Activity Recognition. In: 2020 International Conference on Artificial Intelligence in Information and Communication (ICAIIC), pp. 362–366 (2020). https://doi.org/10.1109/ICAIIC48513.2020.9065078.
Zhu, A.; Qianyu, W.U.; Cui, R., et al.: Exploring a rich spatial-temporal dependent relational model for Skeleton-based action recognition by bidirectional LSTM-CNN. Neurocomputing 414(5), 90–100 (2020). https://doi.org/10.1016/j.neucom.2020.07.068
Chen, C.; Du, Z.; He, L., et al.: A novel gait pattern recognition method based on LSTM-CNN for lower limb exoskeleton. J. Bionic Eng. 18, 1059–1072 (2021). https://doi.org/10.1007/s42235-021-00083-y
Kim, T.; Kim, H.Y.; Hernandez Montoya, A.R.: Forecasting stock prices with a feature fusion LSTM-CNN model using different representations of the same data. PLoS ONE 14(2), e212320 (2019). https://doi.org/10.1371/journal.pone.0212320
Hadfield, S.; Lebeda, K.; Bowden, R.: Hollywood 3D: What are the best 3D features for action recognition. Int. J. Comput. Vis. 121(1), 95–110 (2017). https://doi.org/10.1007/s11263-016-0917-2
Shahroudy, A.; Liu, J.; Ng, T.T.; et al.: NTU RGB+D: a large scale dataset for 3D human activity analysis. In: 2016 IEEE Conference on Computer Vision and Pattern Recognition, pp. 1010–1019 (2016). https://doi.org/10.1109/CVPR.2016.115.
Liu, J.; Shahroudy, A.; Perez, M.; Wang, G.; Duan, L.-Y.; Kot, A.C.: NTU-RGB+D 120: a large-scale benchmark for 3D human activity understanding. IEEE Trans. Pattern Anal. Mach. Intell. (TPAMI) 42(10), 2684–2701 (2019). https://doi.org/10.1109/tpami.2019.2916873
Chan, W.; Tian, Z.; Wu, Y.: GAS-GCN: gated action-specific graph convolutional networks for skeleton-based action recognition. Sensors (Basel) 20(12), 3499 (2020). https://doi.org/10.3390/s20123499
Nie, Q.; Wang, J.; Wang, X., et al.: View-Invariant Human Action Recognition Based on a 3D Bio-Constrained Skeleton Model. IEEE Trans Image Process. 28(8), 3959–3972 (2019). https://doi.org/10.1109/TIP.2019.2907048
Christopher Olah, Understanding LSTM.http://colah.github.io/posts/2015-08-Understanding-LSTMs/Posted on August 27, 2015.
Feichtenhofer, C.; Pinz, A.; Zisserman, A.: Convolutional two-stream network fusion for video action recognition. Comput. Vis. Pattern Recognit. 1, 1933–1941 (2016). https://doi.org/10.1109/CVPR.2016.213
Liu, J.; Shahroudy, A.; Xu, D.; Wang, G.: Spatio-Temporal LSTM with Trust Gates for 3D Human Action Recognition. In: Leibe, B., Matas, J., Sebe, N., Welling, M. (eds) Computer Vision—ECCV 2016. Lecture Notes in Computer Science, Vol. 9907. Springer, Cham (2016). https://doi.org/10.1007/978-3-319-46487-9_50.
Shahroudy, A.; Liu, J.; Ng, T.T., et al.: NTU RGB+D: a large scale dataset for 3D human activity analysis. IEEE Comput. Soc. 1, 1010–1019 (2016). https://doi.org/10.1109/CVPR.2016.115
Liu, J.; Shahroudy, A.; Xu, D.; Wang, G.: Spatio-temporal lstm with trust gates for 3d human action recognition. European Conference on Computer Vision (ECCV) 1, 816–833 (2016). https://doi.org/10.1007/978-3-319-46487-9_50
Li, C.; Wang, P.; Wang, S.; Hou, Y.; Li, W.: Skeleton-based action recognition using lstm and CNN. IEEE International Conference on Multimedia & Expo Workshops 1, 585–590 (2017). https://doi.org/10.1109/ICMEW.2017.8026287
Liu, J.; Wang, G.; Hu, P.; Duan, L.; Kot, A.C.: Global context-aware attention LSTM networks for 3d action recognition. In: 2017 IEEE Conference on Computer Vision and pattern Recognition (CVPR), pp. 3671–3680 (2017). https://doi.org/10.1109/CVPR.2017.391.
Zhang, P.; Lan, C.; **ng, J.; Zeng, W.; Xue, J.; Zheng, N.: View adaptive recurrent neural networks for high-performance human action recognition from skeleton data. In IEEE international Conference on Computer Vision (ICCV), pp. 2136–2145 (2017). https://doi.org/10.1109/ICCV.2017.233.
Cui, R.; Zhu, A.; Zhang, S.; Gang, H.: Multi-source Learning for Skeleton -based Action Recognition Using Deep LSTM Networks, 2018 24th International Conference on Pattern Recognition (ICPR), 547–552,(2018). https://doi.org/10.1109/ICPR.2018.8545247
Zhang, S.; et al.: Fusing geometric features for skeleton-based action recognition using multilayer LSTM networks. In: IEEE Transactions on Multimedia, pp. 2330–2343 (2018).https://doi.org/10.1109/TMM.2018.2802648.
Zhu, A.; Wu, Q.; Cui, R.; Wang, T.; Hang, W.; Hua, G.; Snoussi, H.: Exploring a rich spatial-temporal dependent relational model for skeleton-based action recognition by bidirectional LSTM-CNN. Neurocomputing 414, 90–100 (2020). https://doi.org/10.1016/j.neucom.2020.07.068
Simonyan, K.; Zisserman, A.: ‘Two-stream convolutional networks for action recognition in videos’, Advances in Neural Information Processing Systems (NIPS). Montréal, Canada 1, 568–576 (2014). https://doi.org/10.1002/14651858.CD001941.pub3
Wang, H.; Wang, L.: Modeling temporal dynamics and spatial configurations of actions using two-stream recurrent neural networks. In: 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 3633–3642 (2017). https://doi.org/10.1109/CVPR.2017.387.
Li, C.; Hou, Y.; Wang, P.; Li, W.: Joint distance maps based action recognition with convolutional neural networks. IEEE Signal Process. Lett. 24(5), 624–628 (2017). https://doi.org/10.1109/LSP.2017.2678539
Wang, L.; Zhao, X.; Liu, Y.: Skeleton feature fusion based on multistream lstm for action recognition. IEEE Access 6, 50788–50800 (2018). https://doi.org/10.1109/ACCESS.2018.2869751
Caetano, C.; Brémond, F.; Schwartz, W. R.: Skeleton image representation for 3D action recognition based on tree structure and reference joints. In: 2019 32nd SIBGRAPI conference on graphics, patterns and images (SIBGRAPI), pp. 16–23 (2019). https://doi.org/10.1109/SIBGRAPI.2019.00011.
Ren, Z.; Zhang, Q.; Qiao, P., et al.: Joint learning of convolution neural networks for RGB-D-based human action recognition. Electron. Lett. (2020). https://doi.org/10.1049/el.2020.2148
Acknowledgements
This work was supported by the Research on Automatic Segmentation and Recognition of Teaching Scene with the Characteristics of Teaching Behavior of National Natural Science Foundation of China [61977034]; and the project supported by an open fund (NO.jykf20057) of the Key Laboratory of Intelligent Education Technology and Application of Zhejiang Province, Zhejiang Normal University, and Zhejiang Education Science Planning Project (N0. 2021SCG309), Zhejiang Province, China. At the same time, we thank anonymous reviewers and journal editors for their professional opinions and suggestions.
Author information
Authors and Affiliations
Contributions
The two authors contributed to the study conception and design. Investigation, data curation & writing, and original draft were performed by Dr. H.Z. Resources and supervision were performed by Pro. G.Z. and he is assigned as Corresponding Author.
Corresponding author
Ethics declarations
Conflict of interest
It does not contain potential conflicts of interest on financial interests or personal relationships including financial interests.
Ethics Approval
The Manuscript submitted compliance with Ethical Standards, The experiment does not include involving human participants and/or animals, and data collected from human subjects Compliance with theoretical approval. The individuals concerned Consent to participate and publication.
Rights and permissions
Springer Nature or its licensor holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Zan, H., Zhao, G. Human Action Recognition Research Based on Fusion TS-CNN and LSTM Networks. Arab J Sci Eng 48, 2331–2345 (2023). https://doi.org/10.1007/s13369-022-07236-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s13369-022-07236-z