Real-time gait recognition system based on lightweight dual-modal pose estimation

Authors

  • Din Aodi Belarusian State University, 4 Niezaliezhnasci Avenue, Minsk 220030, Belarus
  • Ye Shiping Zhejiang Shuren University, 8 Shuren Street, Hangzhou 310015, China
  • Alexander M. Nedzved Belarusian State University, 4 Niezaliezhnasci Avenue, Minsk 220030, Belarus , United Institute of Informatics Problems, National Academy of Sciences of Belarus, 6 Surganava Street, Minsk 220012, Belarus
  • Victor S. Anosov 6th City Clinical Hospital, 5 Uralskaja Street, Minsk 220037, Belarus

Keywords:

gait analysis, 3D pose estimation, lightweight model, dual-modal recognition
Supporting Agencies
This work was carried out with the financial support of the China Scholarship Council and the Ministry of Human Resources and Social Security of China (grants No. H20250551 and H20240330).

Abstract

A lightweight approach for joint 2D and 3D gait analysis is proposed. A dual-modal gait recognition system is adopted, achieving high-precision real-time analysis through a 2D and 3D collaborative estimation architecture. The system employs MediaPipe for real-time 2D keypoint detection and gait metric extraction. An improved lightweight architecture MTSA-former with 2.5 mln parameters is used for 3D pose estimation. This model utilises cascaded processing incorporating temporal and spatial modules, along with an adjacency matrix based on skeletal topological structure to model physical constraints between joints. A local-global hybrid spatial modelling strategy fusing graph convolutional networks is implemented to achieve efficient 3D pose estimation, attaining high performance while maintaining lightweight characteristics. This approach ensures accurate and efficient 3D pose reconstruction in real time.

Author Biographies

  • Din Aodi, Belarusian State University, 4 Niezaliezhnasci Avenue, Minsk 220030, Belarus

    postgraduate student at the department of information management systems, faculty of applied mathematics and computer science

  • Ye Shiping, Zhejiang Shuren University, 8 Shuren Street, Hangzhou 310015, China

    vice-rector

  • Alexander M. Nedzved, Belarusian State University, 4 Niezaliezhnasci Avenue, Minsk 220030, Belarus, United Institute of Informatics Problems, National Academy of Sciences of Belarus, 6 Surganava Street, Minsk 220012, Belarus

    doctor of science (engineering), docent; head of the department of information management systems, faculty of applied mathematics and computer science, Belarusian State University, and chief researcher at the department of intellectual information systems, United Institute of Informatics Problems, National Academy of Sciences of Belarus

  • Victor S. Anosov, 6th City Clinical Hospital, 5 Uralskaja Street, Minsk 220037, Belarus

    chief physician

References

  1. Rastegari E, Azizian S, Ali H. Machine learning and similarity network approaches to support automatic classification of Parkinson’s diseases using accelerometer-based gait analysis. In: Bui TX, editor. Proceedings of the 52nd Hawaii international conference on system sciences; 2019 January 8–11; Grand Wailea, USA. Honolulu: ScholarSpace; 2019. p. 4231–4242. DOI: 10.24251/HICSS.2019.511.
  2. Ťupa O, Procházka A, Vyšata O, Schätz M, Mařík J, Vališ M, et al. Motion tracking and gait feature estimation for recognizing Parkinson’s disease using deep learning. BioMedical Engineering OnLine. 2015;14(1):97. DOI: 10.1186/s12938-015-0092-7.
  3. Lugaresi C, Tang J, Nash H, McClanahan C, Uboweja E, Hays M, et al. MediaPipe: a framework for building perception pipelines. arXiv:1906.08172 [Preprint]. 2019 [cited 2025 August 24]: [9 p.]. Available from: https://arxiv.org/pdf/1906.08172.
  4. Zeng W, Ouyang W, Luo P, Liu W, Wang X. 3D human mesh regression with dense correspondence. In: Institute of Electrical and Electronics Engineers. Proceedings of the 2020 IEEE/CVF conference on computer vision and pattern recognition (CVPR); 2020 June 13–19; Seattle, USA. [S. l.]: Institute of Electrical and Electronics Engineers; 2020. p. 7052–7061. DOI: 10.1109/CVPR42600.2020.00708.
  5. Bazarevsky V, Grishchenko I, Raveendran K, Zhu T, Zhang F, Grundmann M. BlazePose: on-device real-time body pose tracking. arXiv:2006.10204 [Preprint]. 2020 [cited 2025 May 22]: [4 p.]. Available from: https://arxiv.org/pdf/2006.10204.
  6. Grishchenko I, Ablavatski A, Kartynnik Y, Raveendran K, Grundmann M. Attention mesh: high-fidelity face mesh prediction in real-time. arXiv:2006.10962 [Preprint]. 2021 [cited 2025 May 1]: [4 p.]. Available from: https://arxiv.org/pdf/2006.10962.
  7. Ionescu C, Papava D, Olaru V, Sminchisescu C. Human3.6M: large scale datasets and predictive methods for 3D human sensing in natural environments. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2014;36(7):1325–1339. DOI: 10.1109/TPAMI.2013.248.
  8. Wandt B, Rosenhahn B. RepNet: weakly supervised training of an adversarial reprojection network for 3D human pose estimation. In: Institute of Electrical and Electronics Engineers. Proceedings of the 2019 IEEE/CVF conference on computer vision and pattern recognition (CVPR); 2019 June 15–20; Long Beach, USA. [S. l.]: Institute of Electrical and Electronics Engineers; 2019. p. 7774–7783. DOI: 10.1109/CVPR.2019.00797.
  9. Baker R, McGinley JL, Schwartz MH, Beynon S, Rozumalski A, Graham HK, et al. The gait profile score and movement analysis profile. Gait & Posture. 2009;30(3):265–269. DOI: 10.1016/j.gaitpost.2009.05.020.
  10. Kainz H, Carty ChP, Modenese L, Boyd RN, Lloyd DG. Estimation of the hip joint centre in human motion analysis: a systematic review. Clinical Biomechanics. 2020;30(4):319–329. DOI: 10.1016/j.clinbiomech.2015.02.005.
  11. Zheng C, Zhu S, Mendieta M, Yang T, Chen C, Ding Z. 3D human pose estimation with spatial and temporal transformers. In: Institute of Electrical and Electronics Engineers. Proceedings of the 2021 IEEE/CVF International conference on computer vision (ICCV); 2021 October 10–17; Montreal, Canada. [S. l.]: Institute of Electrical and Electronics Engineers; 2021. p. 11636–11645. DOI: 10.1109/ICCV48922.2021.01145.
  12. Li W, Liu H, Tang H, Wang P, Van Gool L. MHFormer: multi-hypothesis transformer for 3D human pose estimation. In: Institute of Electrical and Electronics Engineers. Proceedings of the 2022 IEEE/CVF conference on computer vision and pattern recognition (CVPR); 2022 June 18–24; New Orleans, USA. [S. l.]: Institute of Electrical and Electronics Engineers; 2022. p. 13147–13156.

Downloads

Additional Files

Published

2026-05-19

Issue

No. 1 (2026): Journal of the Belarusian State University. Mathematics and Informatics

Section

Data Science

License

Copyright (c) 2026 Journal of the Belarusian State University. Mathematics and Informatics

Creative Commons License

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

The authors who are published in this journal agree to the following:

  1. The authors retain copyright on the work and provide the journal with the right of first publication of the work on condition of license Creative Commons Attribution-NonCommercial. 4.0 International (CC BY-NC 4.0).
  2. The authors retain the right to enter into certain contractual agreements relating to the non-exclusive distribution of the published version of the work (e.g. post it on the institutional repository, publication in the book), with the reference to its original publication in this journal.
  3. The authors have the right to post their work on the Internet (e.g. on the institutional store or personal website) prior to and during the review process, conducted by the journal, as this may lead to a productive discussion and a large number of references to this work. (See The Effect of Open Access.)

How to Cite

[1]
Aodi, D. et al. 2026. Real-time gait recognition system based on lightweight dual-modal pose estimation. Journal of the Belarusian State University. Mathematics and Informatics. 1 (May 2026), 130–139.