{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,11]],"date-time":"2026-05-11T10:56:36Z","timestamp":1778496996134,"version":"3.51.4"},"reference-count":58,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2025,8,13]],"date-time":"2025-08-13T00:00:00Z","timestamp":1755043200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100021856","name":"Ministero dell'Universit\u00e0 e della Ricerca","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100021856","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Comput. Sci."],"abstract":"<jats:sec><jats:title>Introduction<\/jats:title><jats:p>Wearable robotics for lower-limb assistance is increasingly investigated to enhance mobility in individuals with physical impairments and to augment performance in able-bodied users. A major challenge in this domain is the development of accurate and adaptive control systems that ensure seamless human-robot interaction across diverse terrains. While neural networks have recently shown promise in time-series analysis, no prior work has tackled the combined task of classifying ground conditions into five terrain classes and estimating high-level locomotion parameters such as ramp slope and stair height.<\/jats:p><\/jats:sec><jats:sec><jats:title>Methods<\/jats:title><jats:p>This study presents an experimental comparison of eight deep neural network architectures for terrain classification and locomotion parameter estimation. The models are trained on the publicly available CAMARGO 2021 dataset using inertial (IMU) and electromyographic (EMG) signals. Particular attention is given to evaluating the performance of IMU-only inputs versus combined IMU+EMG data, with an emphasis on cost-efficiency and sensor minimization. The tested architectures include LSTM, CNN, and hybrid CNN-LSTM models, among others. Model explainability is assessed via SHAP analysis to guide sensor selection.<\/jats:p><\/jats:sec><jats:sec><jats:title>Results<\/jats:title><jats:p>IMU-only configurations matched or outperformed those using both IMU and EMG, supporting a more efficient setup. The LSTM model, using only three IMU sensors, achieved high terrain classification accuracy (0.94 \u00b1 0.04) and reliably estimated ramp slopes (1.95 \u00b1 0.58\u00b0). The CNN-LSTM architecture demonstrated superior performance in stair height estimation, achieving a accuracy of 15.65 \u00b1 7.40 mm. SHAP analysis confirmed that sensor reduction did not compromise model accuracy.<\/jats:p><\/jats:sec><jats:sec><jats:title>Discussion<\/jats:title><jats:p>The results highlight the feasibility of using lightweight, IMU-only setups for real-time terrain classification and locomotion parameter estimation. The proposed system achieves an inference time of ~2 ms, making it suitable for real-time wearable robotics applications. This study paves the way for more accessible and deployable solutions in assistive and augmentative lower-limb robotic systems. Code and models are publicly available at: [<jats:ext-link>https:\/\/github.com\/cosbidev\/Human-Locomotion-Identification<\/jats:ext-link>].<\/jats:p><\/jats:sec>","DOI":"10.3389\/fcomp.2025.1597143","type":"journal-article","created":{"date-parts":[[2025,8,13]],"date-time":"2025-08-13T13:24:07Z","timestamp":1755091447000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":4,"title":["Deep learning for human locomotion analysis in lower-limb exoskeletons: a comparative study"],"prefix":"10.3389","volume":"7","author":[{"given":"Omar","family":"Coser","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Christian","family":"Tamantini","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Matteo","family":"Tortora","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Leonardo","family":"Furia","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Rosa","family":"Sicilia","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Loredana","family":"Zollo","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Paolo","family":"Soda","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"1965","published-online":{"date-parts":[[2025,8,13]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2406.04419","article-title":"TSCMAMBA: mamba meets multi-view learning for time series classification","author":"Ahamed","year":"2024","journal-title":"arXiv"},{"key":"B2","doi-asserted-by":"publisher","first-page":"350","DOI":"10.1504\/IJMR.2021.119633","article-title":"Human locomotion activity recognition using spectral analysis and convolutional neural networks","volume":"16","author":"Amer","year":"2021","journal-title":"Int. J. Manuf. Res"},{"key":"B3","doi-asserted-by":"publisher","first-page":"2249","DOI":"10.1109\/TCYB.2014.2368353","article-title":"Terrain classification from body-mounted cameras during human locomotion","volume":"45","author":"Anantrasirichai","year":"2014","journal-title":"IEEE Trans. Cybern"},{"key":"B4","doi-asserted-by":"publisher","first-page":"31314","DOI":"10.3390\/s151229858","article-title":"Physical human activity recognition using wearable sensors","volume":"15","author":"Attal","year":"2015","journal-title":"Sensors"},{"key":"B5","doi-asserted-by":"publisher","first-page":"1649","DOI":"10.1093\/comjnl\/bxt075","article-title":"Recognizing daily and sports activities in two open source machine learning environments using body-worn sensor units","volume":"57","author":"Barshan","year":"2014","journal-title":"Comput. J"},{"key":"B6","doi-asserted-by":"publisher","first-page":"1330","DOI":"10.1109\/TBME.2017.2750139","article-title":"A phase variable approach for IMU-based locomotion activity recognition","volume":"65","author":"Bartlett","year":"2017","journal-title":"IEEE Trans. Biomed. Eng"},{"key":"B7","doi-asserted-by":"publisher","first-page":"110320","DOI":"10.1016\/j.jbiomech.2021.110320","article-title":"A comprehensive, open-source dataset of lower limb biomechanics in multiple conditions of stairs, ramps, and level-ground ambulation and transitions","volume":"119","author":"Camargo","year":"2021","journal-title":"J. Biomech"},{"key":"B8","first-page":"212","article-title":"\u201cCase-based explanation of non-case-based learning methods,\u201d","volume-title":"Proceedings of the AMIA Symposium","author":"Caruana","year":"1999"},{"key":"B9","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1412.3555","article-title":"Empirical evaluation of gated recurrent neural networks on sequence modeling","author":"Chung","year":"2014","journal-title":"arXiv"},{"key":"B10","doi-asserted-by":"publisher","first-page":"1341580","DOI":"10.3389\/frobt.2024.1341580","article-title":"Ai-based methodologies for exoskeleton-assisted rehabilitation of the lower limb: a review","volume":"11","author":"Coser","year":"2024","journal-title":"Front. Robot. AI"},{"key":"B11","doi-asserted-by":"publisher","first-page":"1817","DOI":"10.1155\/2011\/759764","article-title":"Lower-limb robotic rehabilitation: literature review and challenges","volume":"2011","author":"D\u00edaz","year":"2011","journal-title":"J. Robot"},{"key":"B12","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1702.08608","article-title":"Towards a rigorous science of interpretable machine learning","author":"Doshi-Velez","year":"2017","journal-title":"arXiv"},{"key":"B13","doi-asserted-by":"publisher","first-page":"1223831","DOI":"10.3389\/fbioe.2023.1223831","article-title":"Autonomous motion and control of lower limb exoskeleton rehabilitation robot","volume":"11","author":"Gao","year":"2023","journal-title":"Front. Bioeng. Biotechnol"},{"key":"B14","doi-asserted-by":"publisher","first-page":"203","DOI":"10.1093\/biomet\/52.1-2.203","article-title":"A generalized Wilcoxon test for comparing arbitrarily singly-censored samples","volume":"52","author":"Gehan","year":"1965","journal-title":"Biometrika"},{"key":"B15","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2312.00752","article-title":"Mamba: linear-time sequence modeling with selective state spaces","author":"Gu","year":"2023","journal-title":"arXiv"},{"key":"B16","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3236009","article-title":"A survey of methods for explaining black box models","volume":"51","author":"Guidotti","year":"2018","journal-title":"ACM Comput. Surv"},{"key":"B17","doi-asserted-by":"publisher","first-page":"133","DOI":"10.1007\/s11062-019-09812-w","article-title":"semg interface design for locomotion identification","volume":"11","author":"Gupta","year":"2017","journal-title":"Int. J. Electr. Comput. Eng"},{"key":"B18","doi-asserted-by":"publisher","first-page":"14","DOI":"10.3389\/frobt.2018.00014","article-title":"Benchmark datasets for bilateral lower-limb neuromechanical signals from wearable sensors during unassisted locomotion in able-bodied individuals","volume":"5","author":"Hu","year":"2018","journal-title":"Front. Robot. AI"},{"key":"B19","doi-asserted-by":"publisher","first-page":"917","DOI":"10.1007\/s10618-019-00619-1","article-title":"Deep learning for time series classification: a review","volume":"33","author":"Ismail Fawaz","year":"2019","journal-title":"Data Min. Knowl. Discov"},{"key":"B20","doi-asserted-by":"publisher","first-page":"1035","DOI":"10.1109\/TNSRE.2021.3082551","article-title":"Open access dataset, toolbox and benchmark processing results of high-density surface electromyogram recordings","volume":"29","author":"Jiang","year":"2021","journal-title":"IEEE Trans. Neural Syst. Rehabil. Eng"},{"key":"B21","doi-asserted-by":"publisher","first-page":"686","DOI":"10.1109\/EMBC48229.2022.9871155","article-title":"\u201cAccurate and robust locomotion mode recognition using high-density emg recordings from a single muscle group,\u201d","author":"Jing","year":"2022","journal-title":"2022 44th Annual International Conference of the IEEE Engineering in Medicine"},{"key":"B22","doi-asserted-by":"publisher","first-page":"3234","DOI":"10.1109\/TBME.2022.3165547","article-title":"Subject-independent continuous locomotion mode classification for robotic hip exoskeleton applications","volume":"69","author":"Kang","year":"2022","journal-title":"IEEE Trans. Biomed. Eng"},{"key":"B23","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2209.12365","article-title":"Deep convolutional neural network and transfer learning for locomotion intent prediction","author":"Le","year":"2022","journal-title":"arXiv"},{"key":"B24","doi-asserted-by":"publisher","first-page":"207","DOI":"10.1016\/B978-0-12-814659-0.00011-4","article-title":"Lower limb exoskeleton systems\u2014overview","volume":"2020","author":"Lee","year":"2020","journal-title":"Wearable Robot"},{"key":"B25","doi-asserted-by":"publisher","first-page":"426","DOI":"10.31763\/ijrcs.v5i1.1588","article-title":"Comparative analysis of 1d-CNN, GRU, and LSTM for classifying step duration in elderly and adolescents using computer vision","volume":"5","author":"Lee","year":"2025","journal-title":"Int. J. Robot. Control Syst"},{"key":"B26","doi-asserted-by":"publisher","first-page":"105372","DOI":"10.1016\/j.bspc.2023.105372","article-title":"Deep-learning model for the prediction of lower-limb joint moments using single inertial measurement unit during different locomotive activities","volume":"86","author":"Liang","year":"2023","journal-title":"Biomed. Signal Process. Control"},{"key":"B27","first-page":"289","article-title":"\u201cA shapelet transform for time series classification,\u201d","volume-title":"Proceedings of the 18th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining","author":"Lines","year":"2012"},{"key":"B28","doi-asserted-by":"publisher","first-page":"31","DOI":"10.1145\/3236386.3241340","article-title":"The mythos of model interpretability: in machine learning, the concept of interpretability is both important and slippery","volume":"16","author":"Lipton","year":"2018","journal-title":"Queue"},{"key":"B29","first-page":"30","article-title":"\u201cA unified approach to interpreting model predictions,\u201d","volume-title":"Advances in Neural Information Processing Systems","author":"Lundberg","year":"2017"},{"key":"B30","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1007\/s13721-025-00522-4","article-title":"Multi-class classification of gait cycle phases using machine learning: a comprehensive study using two training methods","volume":"14","author":"Mekni","year":"","journal-title":"Netw. Model. Anal. Health Inform. Bioinform"},{"key":"B31","doi-asserted-by":"publisher","first-page":"89","DOI":"10.3390\/bdcc9040089","article-title":"Quinary classification of human gait phases using machine learning: investigating the potential of different training methods and scaling techniques","volume":"9","author":"Mekni","year":"","journal-title":"Big Data Cogn. Comput"},{"key":"B32","doi-asserted-by":"publisher","first-page":"eadi8852","DOI":"10.1126\/scirobotics.adi8852","article-title":"Estimating human joint moments unifies exoskeleton control, reducing user effort","volume":"9","author":"Molinaro","year":"2024","journal-title":"Sci. Robot"},{"key":"B33","volume-title":"Interpretable Machine Learning","author":"Molnar","year":"2020"},{"key":"B34","doi-asserted-by":"publisher","first-page":"742","DOI":"10.3390\/s24030742","article-title":"Imu-based fitness activity recognition using cnns for time series classification","volume":"24","author":"M\u00fcller","year":"2024","journal-title":"Sensors"},{"key":"B35","first-page":"40","author":"Napierala","year":"2012"},{"key":"B36","doi-asserted-by":"publisher","first-page":"1749","DOI":"10.1109\/JBHI.2021.3106110","article-title":"Real-time hierarchical classification of time series data for locomotion mode detection","volume":"26","author":"Narayan","year":"2021","journal-title":"IEEE J. Biomed. Health Inform"},{"key":"B37","doi-asserted-by":"publisher","first-page":"199","DOI":"10.1615\/CritRevBiomedEng.2020035013","article-title":"Human locomotion classification for different terrains using machine learning techniques","volume":"48","author":"Negi","year":"2020","journal-title":"Crit. Rev. Biomed. Eng"},{"key":"B38","doi-asserted-by":"publisher","first-page":"134","DOI":"10.1055\/s-0039-1687919","article-title":"Prosthetic rehabilitation in the lower limb","volume":"52","author":"O'Keeffe","year":"2019","journal-title":"Indian J. Plast. Surg"},{"key":"B39","doi-asserted-by":"publisher","first-page":"2727","DOI":"10.3390\/s21082727","article-title":"Wearable sensor-based real-time gait detection: a systematic review","volume":"21","author":"Prasanth","year":"2021","journal-title":"Sensors"},{"key":"B40","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1911.03803","article-title":"Xceptiontime: a novel deep architecture based on depthwise separable convolutions for hand gesture classification","author":"Rahimian","year":"2019","journal-title":"arXiv"},{"key":"B41","doi-asserted-by":"publisher","first-page":"107719","DOI":"10.1016\/j.cmpb.2023.107719","article-title":"Explainable artificial intelligence to predict and identify prostate cancer tissue by gene expression","volume":"240","author":"Ram\u00edrez-Mena","year":"2023","journal-title":"Comput. Methods Programs Biomed"},{"key":"B42","doi-asserted-by":"publisher","first-page":"1135","DOI":"10.18653\/v1\/N16-3020","author":"Ribeiro","year":"2016"},{"key":"B43","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1109\/ENBENG.2017.7889459","article-title":"\u201cInertial measurement units: a brief state of the art on gait analysis,\u201d","volume-title":"2017 IEEE 5th Portuguese Meeting on Bioengineering (ENBENG)","author":"Ribeiro","year":"2017"},{"key":"B44","doi-asserted-by":"publisher","first-page":"115582","DOI":"10.1016\/j.eswa.2021.115582","article-title":"Modeling spatiotemporal patterns of gait anomaly with a cnn-lstm deep neural network","volume":"185","author":"Sadeghzadehyazdi","year":"2021","journal-title":"Expert Syst. Appl"},{"key":"B45","first-page":"618","article-title":"\u201cGrad-cam: visual explanations from deep networks via gradient-based localization,\u201d","volume-title":"Proceedings of the IEEE International Conference on Computer Vision","author":"Selvaraju","year":"2017"},{"key":"B46","doi-asserted-by":"publisher","first-page":"016302","DOI":"10.1088\/1361-6501\/ad7f78","article-title":"Simple and efficient step detection algorithm for foot-mounted imu","volume":"36","author":"Shi","year":"2024","journal-title":"Meas. Sci. Technol"},{"key":"B47","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1312.6034","article-title":"Deep inside convolutional networks: visualising image classification models and saliency maps","author":"Simonyan","year":"2013","journal-title":"arXiv"},{"key":"B48","doi-asserted-by":"publisher","first-page":"1082","DOI":"10.3390\/bioengineering10091082","article-title":"Multivariate cnn model for human locomotion activity recognition with a wearable exoskeleton robot","volume":"10","author":"Son","year":"2023","journal-title":"Bioengineering"},{"key":"B49","doi-asserted-by":"crossref","first-page":"80","DOI":"10.1109\/WEARRACON.2019.8719626","article-title":"\u201cHuman locomotion activity and speed recognition using electromyography based features,\u201d","volume-title":"2019 Wearable Robotics Association Conference (WearRAcon)","author":"Sorkhabadi","year":"2019"},{"key":"B50","doi-asserted-by":"publisher","first-page":"7127","DOI":"10.3390\/s20247127","article-title":"Gait trajectory and gait phase prediction based on an lstm network","volume":"20","author":"Su","year":"2020","journal-title":"Sensors"},{"key":"B51","doi-asserted-by":"publisher","first-page":"47563","DOI":"10.1109\/ACCESS.2023.3275126","article-title":"Radiopathomics: multimodal learning in non-small cell lung cancer for adaptive radiotherapy","volume":"11","author":"Tortora","year":"2023","journal-title":"IEEE Access"},{"key":"B52","doi-asserted-by":"publisher","first-page":"e20","DOI":"10.1097\/JPO.0000000000000355","article-title":"How do we measure success? A review of performance evaluations for lower-limb neuroprosthetics","volume":"34","author":"Valgeirsd\u00f3ttir","year":"2022","journal-title":"J. Prosthet. Orthot"},{"key":"B53","first-page":"231","article-title":"\u201cLocomotion mode recognition method based on inertial measurement units in exoskeleton robot,\u201d","volume-title":"International Symposium on Robotics, Artificial Intelligence, and Information Engineering (RAIIE 2022), Vol. 12454","author":"Wang","year":"2022"},{"key":"B54","doi-asserted-by":"publisher","first-page":"2114","DOI":"10.1145\/3447548.3467401","article-title":"\u201cA transformer-based framework for multivariate time series representation learning,\u201d","author":"Zerveas","year":"2021","journal-title":"Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery"},{"key":"B55","doi-asserted-by":"publisher","first-page":"646","DOI":"10.1109\/TNSRE.2020.2966749","article-title":"Unsupervised cross-subject adaptation for predicting human locomotion intent","volume":"28","author":"Zhang","year":"2020","journal-title":"IEEE Trans. Neural Syst. Rehabil. Eng"},{"key":"B56","doi-asserted-by":"crossref","first-page":"2768","DOI":"10.1109\/CCDC55256.2022.10034195","article-title":"\u201cMulti-channel separated encoder based convolutional neural network for locomotion intention recognition,\u201d","volume-title":"2022 34th Chinese Control and Decision Conference (CCDC)","author":"Zhao","year":"2022"},{"key":"B57","article-title":"\u201cApplications of mems gyroscope for human gait analysis,\u201d","volume-title":"Gyroscopes-Principles and Applications","author":"Zhao","year":"2019"},{"key":"B58","doi-asserted-by":"publisher","first-page":"2250043","DOI":"10.1142\/S0219519422500439","article-title":"A cnn-svm model using imu for locomotion mode recognition in lower extremity exoskeleton","volume":"22","author":"Zheng","year":"2022","journal-title":"J. Mech. Med. Biol"}],"container-title":["Frontiers in Computer Science"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fcomp.2025.1597143\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,8,13]],"date-time":"2025-08-13T13:24:10Z","timestamp":1755091450000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fcomp.2025.1597143\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,8,13]]},"references-count":58,"alternative-id":["10.3389\/fcomp.2025.1597143"],"URL":"https:\/\/doi.org\/10.3389\/fcomp.2025.1597143","relation":{},"ISSN":["2624-9898"],"issn-type":[{"value":"2624-9898","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,8,13]]},"article-number":"1597143"}}