A Branched Fusion Network for the Recognition of Continuous Multiterrain Lower-Limb Motion
IEEE Transactions on Instrumentation and Measurement · 2026 · 75 · 1-14
DOI: 10.1109/tim.2026.3677939 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Instrumentation and Measurement, 2026. Verfasst von Xiaohui Li und 8 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 40 Literaturangaben. Gefördert von: National Key Research and Development Program of China, National Natural Science Foundation of China, in part by Shenzhen Strategic Emerging Industry Support Plans.
Abstract (Originalsprache)
The real-time and accurate recognition of motion intention remains a significant challenge and serves as a fundamental prerequisite for achieving coordinated movement between lower-limb rehabilitation devices and the human body. Moreover, the scarcity of multiterrain lower-limb motion datasets critically constrains the performance and generalization capability of recognition models. In this study, the branched fusion network (BFNet) is proposed to achieve low-delay and high-accuracy recognition of continuous multiterrain lower-limb motions. The BFNet, which fuses the surface electromyography (sEMG) block and IMU block modules, takes as input the complete accelerometer and gyroscope data from the previous gait cycle, alongside the onset-window sEMG data from the current gait phase. A continuous multiterrain lower-limb motion dataset (CMLMD), comprising three scenarios (level ground, stairs, and ramps), five basic motion modes, and eight transitional modes, has been publicly released (DOI: 10.21227/cwgk-4a45). The dataset includes 35 participants and a total of 7000 gait cycles. An overall recognition accuracy of 95.41% was achieved on the publicly available CMLMD, with a corresponding time delay of only 114 ms. The accuracy reached 88.61% in the cross-subject evaluation. The optimal sensor configuration was determined to consist of a single IMU sensor and six sEMG sensors. Under this configuration, the recognition accuracy remained at 94.38%. The superiority of BFNet in achieving both high recognition accuracy and low delay is validated through comprehensive comparisons with existing methods. The proposed BFNet provides a promising solution for human–machine collaborative movement of the lower limbs. The CMLMD is intended to serve as a foundational benchmark for developing and validating algorithms for continuous lower-limb motion recognition.
Förderung
- National Key Research and Development Program of China
- National Natural Science Foundation of China
- in part by Shenzhen Strategic Emerging Industry Support Plans
- in part by Shenzhen Sustainable Development Sci-Tech Project
- Shenzhen Science and Technology Program
Zitieren
APA
Li, X., Lyu, X., Zou, H., Zhou, H., Yu, D., Wang, W., Li, J., Li, G., & Wang, L. (2026). A Branched Fusion Network for the Recognition of Continuous Multiterrain Lower-Limb Motion. IEEE Transactions on Instrumentation and Measurement, 75, 1-14. https://doi.org/10.1109/tim.2026.3677939
Vancouver
Li X, Lyu X, Zou H, Zhou H, Yu D, Wang W, et al. A Branched Fusion Network for the Recognition of Continuous Multiterrain Lower-Limb Motion. IEEE Transactions on Instrumentation and Measurement. 2026;75:1-14. doi:10.1109/tim.2026.3677939
BibTeX
@article{Li2026,
author = {Li, Xiaohui and Lyu, Xueyan and Zou, Hongfei and Zhou, Hao and Yu, Dezhi and Wang, Wenzhuo and Li, Jiaqi and Li, Guanglin and Wang, Lin},
title = {A Branched Fusion Network for the Recognition of Continuous Multiterrain Lower-Limb Motion},
journal = {IEEE Transactions on Instrumentation and Measurement},
year = {2026},
volume = {75},
pages = {1--14},
doi = {10.1109/tim.2026.3677939}
}
EUVIMED