A multi-task framework based on SDA-LSTM fusion network for gait phase recognition and gait cycle percentage progression prediction by IMU for forward walking

Abstract

Accurate recognition of gait phases and reliable prediction of gait cycles are critical for adaptive exoskeleton control. Although these two tasks are inherently coupled in both temporal and spatial domains, they have often been investigated in isolation. Such separation constrains the robustness and responsiveness of assistive devices in real-world applications. To bridge this gap, we propose a multi-task learning framework that integrates a Stacked Denoising Autoencoder with a Long Short-Term Memory network (SDA–LSTM). Inertial measurement unit (IMU) signals are employed for gait analysis, enabling both discrete gait phase classification and continuous estimation of gait cycle percentage during forward walking. The framework was evaluated against several established models, including Support Vector Machines (SVM), XGBoost, and standalone LSTM networks, under cross-subject validation. For phase classification, SDA–LSTM achieved 97.3% accuracy, with only a small reduction of 1.31% compared with its training performance. Robustness was further demonstrated under severe noise conditions (signal-to-noise ratio, SNR = 5 dB), where the model maintained 95.68% accuracy, surpassing all baseline methods. For cycle prediction, SDA–LSTM also showed strong stability. At SNR = 5 dB, error metrics increased only slightly, with RMSE and MAE rising by 0.083 and 0.075, while R² and PCC decreased marginally by 0.0173 and 0.0089. These results highlight the effectiveness of SDA–LSTM in capturing the spatiotemporal synergy of gait. The framework demonstrates high accuracy, robustness, and generalization. Its performance underscores strong potential for deployment in exoskeleton systems, paving the way toward reliable and adaptive human–robot interaction in daily locomotion.

Keywords

Gait phase recognition / Gait cycle percentage prediction / Exoskeleton / SDA–LSTM / Noise interference

CRediT authorship contribution statement

Zhi Pang: Writing – review & editing, Writing – original draft, Visualization, Validation, Software, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Wujing Cao: Writing – review & editing, Supervision, Methodology, Investigation, Data curation, Conceptualization. Ao Meng: Investigation, Data curation. Linna Tian: Investigation. Maolin Wang: Investigation. Mingxiang Luo: Investigation. Bingshan Hu: Writing – review & editing, Supervision, Project administration, Methodology, Funding acquisition, Data curation, Conceptualization. Xinyu Wu: Supervision, Methodology, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgment

This study is supported by the Research and Innovation Team Project awarded in 2024 (KYCX2403) from Shanghai Shidong hospital, China .

References

[1] X. Wu, S. Bai, L. O’Sullivan, Editorial for the special issue on wearable robots and intelligent device, Biomim. Intell. Robot. 3 (2023) 100102, https://doi.org/10.1016/j.birob.2023.100102.

[2] X. Wang, S. Guo, B. Qu, M. Song, P. Wang, D.-X. Liu, Design and experimental verification of a parallel hip exoskeleton system for full-gait-cycle rehabilitation, J. Mech. Robot.,, 14 (2022), https://doi.org/10.1115/1.4053735.

[3] X. Wang, Y. Ma, C. Chen, Z. Wang, S. Guo, K. Huat Low, X. Wu, Effective prediction of gait phase for assisted walking by means of gait-based adaptive oscillators, IEEE Trans. Autom. Sci. Eng. 22 (2025) 10205-10215, https://doi.org/10.1109/TASE.2024.3520148.

[4] E. Tricomi, N. Lotti, F. Missiroli, X. Zhang, M. Xiloyannis, T. Müller, S. Crea, E. Papp, J. Krzywinski, N. Vitiello, L. Masia, Underactuated soft hip exosuit based on adaptive oscillators to assist human locomotion, IEEE Robot. Autom. Lett. 7 (2022) 936-943, https://doi.org/10.1109/LRA.2021.3136240.

[5] Z. Wang, C. Chen, H. Chen, S. Wang, J. Zhang, X. Wang, X. Wu, A study on enhancing wearer adaptation through accurate gait phase prediction and gradual increase in assistive force magnitude in exosuits, IEEE Robot. Autom. Lett. 10 (2025) 7803-7810, https://doi.org/10.1109/LRA.2025.3578260.

[6] B. Su, E.M. Gutierrez-Farewik, Gait trajectory and gait phase prediction based on an LSTM network, Sensors, 20 (7127), (2020) , https://doi.org/10.3390/s20247127.

[7] T. Yan, A. Parri, V. Ruiz Garate, M. Cempini, R. Ronsse, N. Vitiello, An oscillator-based smooth real-time estimate of gait phase for wearable robotics, Auton Robot. 41 (2017) 759-774, https://doi.org/10.1007/s10514-016-9566-0.

[8] X. Ji, W. Zeng, Q. Dai, Y. Zhang, S. Du, B. Ji, Machine learning-based detection of cervical spondylotic myelopathy using multiple gait parameters, Biomim. Intell. Robot. 3 (2023) 100103, https://doi.org/10.1016/j.birob.2023.100103.

[9] J. Kim, M.-N. Bae, K.B. Lee, S.G. Hong, Gait event detection algorithm based on smart insoles, ETRI J. 42 (2020) 46-53, https://doi.org/10.4218/etrij.2018-0639.

[10] J.S. Park, C.M. Lee, S.-M. Koo, C.H. Kim, Gait phase detection using force sensing resistors, IEEE Sensors J. 20 (2020) 6516-6523, https://doi.org/10.1109/JSEN.2020.2975790.

[11] F. Attal, Y. Amirat, A. Chibani, S. Mohammed, Automatic recognition of gait phases using a multiple-regression hidden Markov model, IEEE/ASME Trans. Mechatronics 23 (2018) 1597-1607, https://doi.org/10.1109/TMECH.2018.2836934.

[12] X. Hu, Q. Duan, J. Tang, G. Chen, Z. Zhao, Z. Sun, C. Chen, X. Qu, A low-cost instrumented shoe system for gait phase detection based on foot plantar pressure data, IEEE J. Transl. Eng. Health Med. 12 (2024) 84-96, https://doi.org/10.1109/JTEHM.2023.3319576.

[13] E. Martini, T. Fiumalbi, F. Dell’Agnello, Z. Ivanić, M. Munih, N. Vitiello, S. Crea, Pressure-sensitive insoles for real-time gait-related applications, Sensors, 20(1448), (2020), https://doi.org/10.3390/s20051448.

[14] Y.C. Han, K.I. Wong, I. Murray, Gait phase detection for normal and abnormal gaits using IMU, IEEE Sensors J. 19 (2019) 3439-3448, https://doi.org/10.1109/JSEN.2019.2894143.

[15] S. Sahoo, M. Saboo, D.K. Pratihar, S. Mukhopadhyay, Real-time detection of actual and early gait events during level-ground and ramp walking, IEEE Sensors J. 20 (2020) 8128-8136, https://doi.org/10.1109/JSEN.2020.2980863.

[16] A.R. Anwary, H. Yu, M. Vassallo, Optimal foot location for placing wearable IMU sensors and automatic feature extraction for gait analysis, IEEE Sensors J. 18 (2018) 2555-2567, https://doi.org/10.1109/JSEN.2017.2786587.

[17] H.L. Bartlett, M. Goldfarb, A phase variable approach for IMU-based locomotion activity recognition, IEEE Trans. Biomed. Eng. 65 (2018) 1330-1338, https://doi.org/10.1109/TBME.2017.2750139.

[18] J.M. Mahoney, M.B. Rhudy, Methodology and validation for identifying gait type using machine learning on IMU data, J. Med. Eng. Technol. 43 (2019) 25-32, https://doi.org/10.1080/03091902.2019.1599073.

[19] F. Weigand, A. Höhl, J. Zeiss, U. Konigorski, M. Grimmer, Continuous locomotion mode recognition and gait phase estimation based on a shank-mounted IMU with artificial neural networks, 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS, 2022, pp. 12744-12751, https://doi.org/10.1109/IROS47612.2022.9981154.

[20] P. Wei, J. Zhang, F. Tian, J. Hong, A comparison of neural networks algorithms for EEG and sEMG features based gait phases recognition, Biomed. Signal Process. Control. 68 (2021) 102587,https://doi.org/10.1016/j.bspc.2021.102587.

[21] R. Luo, S. Sun, X. Zhang, Z. Tang, W. Wang, A low-cost end-to-end sEMG-based gait sub-phase recognition system, IEEE Trans. Neural Syst. Rehabil. Eng. 28 (2020) 267-276, https://doi.org/10.1109/TNSRE.2019.2950096.

[22] Pengna Wei, J. Zhang, Pingping Wei, B. Wang, J. Hong, Different sEMG and EEG features analysis for gait phase recognition, 2020 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society, EMBC, 2020, pp. 1002-1006, https://doi.org/10.1109/EMBC44109.2020.9175655.

[23] S. Cai, D. Chen, B. Fan, M. Du, G. Bao, G. Li, Gait phases recognition based on lower limb sEMG signals using LDA-PSO-LSTM algorithm, Biomed. Signal Process. Control. 80 (2023) 104272,https://doi.org/10.1016/j.bspc.2022.104272.

[24] D. Hollinger, M. Schall, H. Chen, S. Bass, M. Zabala, The influence of gait phase on predicting lower-limb joint angles, IEEE Trans. Med. Robot. Bionics 5 (2023) 343-352, https://doi.org/10.1109/TMRB.2023.3260261.

[25] N. Mijailovic, M. Gavrilovic, S. Rafajlovic, M. Ðuric-Jovicic, D. Popovic, Gait phases recognition from accelerations and ground reaction forces: Application of neural networks, Telfor J. 1 (2009) 34-36.

[26] J. Ryu, B.-H. Lee, J. Maeng, D.-H. Kim, sEMG-signal and IMU sensor-based gait sub-phase detection and prediction using a user-adaptive classifier, Med. Eng. Phys. 69 (2019) 50-57, https://doi.org/10.1016/j.medengphy.2019.05.006.

[27] X. Jiang, K.H.T. Chu, M. Khoshnam, C. Menon, A wearable gait phase detection system based on force myography techniques, Sensors, 18(1279), (2018), https://doi.org/10.3390/s18041279.

[28] Y. Qi, C.B. Soh, E. Gunawan, K.-S. Low, R. Thomas, Assessment of foot trajectory for human gait phase detection using wireless ultrasonic sensor network, IEEE Trans. Neural Syst. Rehabil. Eng. 24 (2016) 88-97, https://doi.org/10.1109/TNSRE.2015.2409123.

[29] J. Zhao, J. Yang, J. Wang, W. Wu, Spiking neural network regularization with fixed and adaptive drop-keep probabilities, IEEE Trans. Neural Networks Learn. Syst. 33 (2022) 4096-4109, https://doi.org/10.1109/TNNLS.2021.3055825.

[30] J. Camargo, A. Ramanathan, W. Flanagan, A. Young, A comprehensive, open-source dataset of lower limb biomechanics in multiple conditions of stairs, ramps, and level-ground ambulation and transitions, J. Biomech. 119 (2021) 110320,https://doi.org/10.1016/j.jbiomech.2021.110320.

[31] W. Cao, C. Li, L. Yang, M. Yin, C. Chen, W. Kobsiriphat, T. Utakapan, Y. Yang, H. Yu, X. Wu, A fusion network with stacked denoise autoencoder and meta learning for lateral walking gait phase recognition and multi-step-ahead prediction, IEEE J. Biomed. Health Inform. 29 (2025) 68-80, https://doi.org/10.1109/JBHI.2024.3380099.