With the rapid development of intelligent sports analysis technology, automatic action recognition has become a key technology for performance evaluation and safety monitoring in winter sports. This study aims to improve the automatic recognition of complex behaviors in ice and snow sports and meet practical needs such as intelligent training and assisted judgment. The study proposes a behavior recognition method based on the Expectation -Maximization (EM) algorithm and the Re -parameterized Visual Geometry Group network (R epVGG). First, a multi -modal input feature channel is constructed by fusing residual maps and pose heatmaps extracted from video frames. This enhances the model ’s structural perception of the athlete and reduces background interference. Second, an Expectation -Maximization -based Refinement (EM -R) module is designed to perform soft clustering on frame -level features. By learning latent class distributions, this modu le introduces semantic priors to improve the classification capability of the subsequent backbo ne network. Finally, multi -scale convolution units and an EM -attention mechanism are integrated into the backbone to enhance the model’s response to local dynamics and key action areas. The RepVGG -based architecture is further re -parameterized to enable efficient inference and lightweight deployment. Experiments are conducted on a skiing -related subset built from the public University of Central Florida (UCF101) dataset, including representative ice and snow behaviors such as “Skiing”, “Ski Jumping”, and “Snowboarding”. Results show that the proposed method achieves 9 2.7% classification accuracy and a macro -average F1 -score of 90.4% on the test set, outperforming the original RepVGG by 3.4% and 3.3%, respectively. The inference speed reaches 110 Frames Per Second (FPS), and real -time performance of 45 FPS is achieved on Jetson NX devices, demonstrating strong deployment capability and platform adaptability. Ablation studies confirm the critical roles of the EM -R module, multi -scale structure, and pose channel in im proving overall performance .

[1] Abdelbaky A. and Aly S., “Human Action Recognition Using Three Orthogonal Planes with Unsupervised Deep Convolutional Neural Network,” Multimedia Tools and Applications , vol. 80, no. 13, pp. 20019 -20043, 2021. https://link.springer.com/article/10.1007/s11042 - 021 -10636 -2

[2] Alhakbani N., Alghamdi M., and Al -Nafjan A., “Design and Development of An Imitation Detection System for Human Action Recognition Using Deep Learning,” Sensors , vol. 23, no. 24, pp. 1 -16, 2023. DOI: 10.3390/s23249889

[3] Balakrishnan T. and Sengar S., “RepVGG - GELAN: Enhanced GELAN with VGG -Style Convnets for Brain Tumour Detection,” in Proceedings of the 9 th International Conference on Computer Vision and Image Processing , Chennai, pp. 417 -430, 2024. https://link.springer.com/chapter/10.1007/978 -3- 031 -93688 -3_30

[4] Chen D., Wu P., Chen M., Wu M., and et al., “LS - ViT: Vision Transformer for Action Recognition Based on Long and Short -Term Temporal Difference,” Frontiers in Neurorobotics , vol. 18, pp. 1 -12, 2024. https://doi.org/10.3389/fnbot.2024.1457843

[5] Cheng J., Ren Z., Zhang Q., Gao X., and Hao F., “Cross -Modality Compensation Convolutional Neural Networks for RGB -D Action Recognition,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 32, no. 3, pp. 1498 -1509, 2022. https://ieeexplore.ieee.org/document/9417202

[6] Ding G., Sener F., and Yao A., “Temporal Action Segmentation: An Analysis of Modern Techniques,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 46, no. 2, pp. 1011 -1030, 2023. DOI: 10.1109/TPAMI.2023.3327284

[7] Ding I., Zheng N., and Hsieh M., “Hand Gesture Intention -Based Identity Recognition Using Various Recognition Strategies Incorporated with VGG Convolution Neural Network -Extracted Deep Learning Features,” Journal of Intelligent and Fuzzy Systems , vol. 40, no. 4, pp. 7775 -7788, 2021. https://doi.org/10.3233/JIFS -189598

[8] Elpeltagy M. and Sallam H., “Automatic Prediction of COVID -19 from Chest Images Using Modified ResNet50,” Multimedia Tools and Applications , vol. 80, no. 17, pp. 26451 -26463, 2021. DOI: 10.1007/s11042 -021 -10783 -6

[9] Guo Y., Ju R., Li K., Lan Z., and et al., “A Smart Ski Pole for Skiing Pattern Recognition and Quantification Application,” Sensors , vol. 24, no. 16, pp. 1 -14, 2024. DOI: 10.3390/s24165291

[10] Han J. and Li J., “A Video Action Recognition Method via Dual -Stream Feature Fusion Neural Network with Attention,” International Journal of Uncertainty, Fuzziness and Knowledge -Based Systems , vol. 32, no. 4, pp. 673 -694, 2024. https://doi.org/10.1142/S0218488524400130

[11] Hu K., Jin J., Zheng F., Weng L., and Ding Y., “Overview of Behavior Recognition Based on Deep Learning,” Artificial Intelligence Review , vol. 56, no. 3, pp. 1833 -1865, 2023. https://doi.org/10.1007/s10462 -022 -10210 -8

[12] Kareem I., Ali S., Bilal M., and Hanif M., “Exploiting the Features of Deep Residual Network with SVM Classifier for Human Posture Recognition,” PLoS ONE , vol. 19, no. 12, pp. 1 - 26, 2024. DOI: 10.1371/journal.pone.0314959

[13] Korban M., Youngs P., and Acton S., “A Semantic and Motion -Aware Spatiotemporal Transformer Network for Action Detection,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 46, no. 9, pp. 6055 -6069, 2024. https://ieeexplore.ieee.org/document/10472872

[14] Li H. and Reddy M., “Remote Dance Action Correction Based on EM and Min -Min Algorithms,” Mobile Networks and Applications , vol. 30, pp. 1037 -1049, 2025. DOI: 10.1007/s11036 -024 -02427 -4

[15] Li T., Wang J., Wiltos K., and Wo zniak M., “An Intelligent Proofreading for Remote Skiing Actions Based on Variable Shape Basis,” Mobile Networks and Applications , vol, 30, pp. 1008 - 1021, 2026. DOI: 10.1007/s11036 -024 -02419 -4

[16] Lin C., Zheng Y., Xiao X., and Lin J., “CXR - RefineDet: Single -Shot Refinement Neural Network for Chest X -Ray Radiograph Based on Multiple Lesions Detection,” Journal of Healthcare Engineering , vol. 2022, no. 1, pp. 1 - 11, 2022. DOI: 10.1155/2022/4182191

[17] Liu J., Huang G., Hyyppä J., Li J., and et al., “A Survey on Location and Motion Tracking Technologies, Methodologies and Applications in Precision Sports,” Expert Systems with Applications , vol. 229, pp. 120492, 2023. https://doi.org/10.1016/j.eswa.2023.120492

[18] Liu Y., Zhang H., Li Y., He K., and Xu D., “Skeleton -Based Human Action Recognition via Large -Kernel Attention Graph Convolutional Network,” IEEE Transactions on Visualization and Computer Graphics , vol. 29, no. 5, pp. 2575 - 2585, 2023. DOI: 10.1109/TVCG.2023.3247075

[19] Lopez -Lozada E., Sossa H., Rubio -Espino E., and Montiel -Perez J., “Action Recognition in Videos Through a Transfer -Learning -Based Technique,” Mathematics , vol. 12, no. 20, pp. 1 -17, 2024. https://doi.org/10.3390/math12203245

[20] Ma L., Yang H., and Jin G., “A Spatio -Temporal Feature Representation of Multimodal Surveillance Images for Behavioral Recognition,” The International Arab Journal of Information Technology , vol. 22, no. 4, pp. 832 -843, 2025. https://doi.org/10.34028/iajit/22/4/15

[21] Muhammad K., Mustaqeem., Ullah A., Imran A., and et al., “Human Action Recognition Using Attention Based LSTM Network with Dilated CNN Features,” Future Generation Computer Systems , vol. 125, pp. 820 -830, 2021. https://doi.org/10.1016/j.future.2021.06.045

[22] Muttenthaler L., Dippel J., Linhardt L., Vandermeulen R., and Kornblith S., “Human Alignment of Neural Network Representations,” in Proceedings of the Machine Learning Group, Technische Universitat , Berlin, pp. 1 -38, 2023. https://iclr.cc/virtual/2023/poster/11553

[23] Nasaoui H., Bellamine I., and Silkan H., “Action Recognition Using Segmental Action Network,” Revue d’Intelligence Artificielle , vol. 37, no. 1, pp. 185 -190, 2023. DOI: 10.18280/ria.370123

[24] Patel S. and Kamdar D., “Survey on Sport Video Analysis and Event Detection,” International Journal of Autonomous and Adaptive Communications Systems , vol. 18, no. 1, pp. 67 - 98, 2025. DOI: 10.1504/IJAACS.2025.10059628

[25] Shi C., Han L., Zhang K., Xiang H., and et al., “Improved RepVGG Ground -Based Cloud Image Classification with Attention Convolution,” Atmospheric Measurement Techniques , vol. 17, no. 3, pp. 979 -997, 2024. https://amt.copernicus.org/articles/17/979/2024/

[26] Sowmya M., Balasubramanian M., and Vaidehi K., “Human Behavior Classification Using 2D Convolutional Neural Network, VGG16 and ResNet50,” Indian Journal of Science and Technology , vol. 16, no. 16, pp. 1221 -1229, 2023. DOI: 10.17485/IJST/v16i16.199

[27] Wang M., Xing J., Su J., Chen J., and Liu Y., “Learning Spatiotemporal and Motion Features in a Unified 2D Network for Action Recognition,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 45, no. 3, pp. 3347 - 3362, 2023. DOI: 10.1109/TPAMI.2022.3173658

[28] Wang Z. and Jing R., “Optimized Design of Flexible Sensors for Snow and Ice Sports Characterization and Intelligent Recognition of Training Technique Movements,” Journal of Radiation Research and Applied Sciences , vol. 17, no. 3, pp. 1 -9, 2024. https://doi.org/10.1016/j.jrras.2024.101028

[29] Wensel J., Ullah H., and Munir A., “ViT -RET: Vision and Recurrent Transformer Neural Networks for Human Activity Recognition in Videos,” IEEE Access , vol. 11, pp. 72227 -72249, 2023. DOI: 10.1109/ACCESS.2023.3293813

[30] Wu Q., Zhu A., Cui R., Wang T., and et al., “Pose - Guided Inflated 3D ConvNet for Action Recognition in Videos,” Signal Processing: Image Communication , vol. 91, pp. 116098, 2021. https://www.sciencedirect.com/science/article/abs /pii/S0923596520302186?via%3Dihub

[31] Yang J., Jia Q., Han S., Du Z., and Liu J., “An Efficient Multi -Scale Attention Two -Stream Inflated 3D Convnet Network for Cattle Behavior Recognition,” Computers and Electronics in Agriculture , vol. 232, pp. 110101, 2025. DOI: 10.1016/j.compag.2025.110101

[32] Ye X., Sakurai K., Nair N., and Wang K., “Machine Learning Techniques for Sensor -Based Human Activity Recognition with Data Heterogeneity -A Review,” Sensors , vol. 24, no. 24, pp. 1 -39, 2024. DOI: 10.3390/s24247975

[33] Zhao L., Lin Z., Sun R., and Wang A., “A Review of State -of-the -Art Methodologies and Applications in Action Recognition,” Electronics , vol. 13, no. 23, pp. 1 -39, 2024. DOI: 10.3390/electronics13234733

[34] Zhao X., “Research on Athlete Behavior Recognition Technology in Sports Teaching Video Based on Deep Neural Network,” Computational Intelligence and Neuroscience , vol. 2022, pp. 1 - 13, 2022. https://doi.org/10.1155/2022/7260894