{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,2,27]],"date-time":"2026-02-27T19:14:49Z","timestamp":1772219689164,"version":"3.50.1"},"reference-count":35,"publisher":"Public Library of Science (PLoS)","issue":"2","license":[{"start":{"date-parts":[[2026,2,27]],"date-time":"2026-02-27T00:00:00Z","timestamp":1772150400000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100023674","name":"Deanship of Scientific Research, King Khalid University","doi-asserted-by":"publisher","award":["RGP2\/176\/46"],"award-info":[{"award-number":["RGP2\/176\/46"]}],"id":[{"id":"10.13039\/501100023674","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["www.plosone.org"],"crossmark-restriction":false},"short-container-title":["PLoS One"],"abstract":"<jats:p>Gait speed analysis is vital for applications in healthcare, rehabilitation, human-robot interaction, and autonomous systems, necessitating robust methods to address its complexity. This study introduces an advanced framework for gait speed classification through multimodal data fusion and deep learning, optimized using Layer-wise Relevance Propagation (LRP). Utilizing a publicly available 4 datasets of 50 injury-free adults walking at varying speeds, we integrate data from full-body motion capture, electromyography (EMG), and force plates to capture detailed gait dynamics. We propose two novel architectures: a hybrid Convolutional Neural Network with Long Short-Term Memory (CNN+LSTM) and a Multi-stream CNN, benchmarked against Temporal Convolutional Networks (TCNs), Transformer neural networks, Gated Recurrent Units (GRU) and statistical classifiers (e.g., Linear Discriminant Analysis, Quadratic Discriminant Analysis, Support Vector Machines). All models in this study were trained, validated and tested using three distinct strategies: experiment 1 (sample-based splitting \u2013 10-fold cross-validation), experiment 2 (10-fold cross-validation, subject-based splitting \u2013 5 subjects for testing), and experiment 3 (5-fold cross-validation, subject-based splitting \u2013 10 subjects for testing). Multi-stream Quads CNN is the best-performing model, achieving the highest F1 scores across all experiments (96.6%\u2009\u00b1\u20091.4% experiment 1, 96.2%\u2009\u00b1\u20091.6% experiment 2, 95.9%\u2009\u00b1\u20090.9% experiment 3 with top performance for all experiments 98%\u2009\u00b1\u20090.6%, outperforming traditional approaches. LRP is applied to evaluate feature relevance within the model, enabling the removal of critical features to verify robustness. Comprehensive assessments, including ROC curves, confusion matrices, t-test, and perturbation analyses, validate the model\u2019s enhanced performance and durability. By combining multimodal fusion to overcome single-sensor constraints with LRP-driven feature optimization, this work provides a highly accurate and resilient tool for gait analysis, with significant potentials.<\/jats:p>","DOI":"10.1371\/journal.pone.0341067","type":"journal-article","created":{"date-parts":[[2026,2,27]],"date-time":"2026-02-27T18:35:23Z","timestamp":1772217323000},"page":"e0341067","update-policy":"https:\/\/doi.org\/10.1371\/journal.pone.corrections_policy","source":"Crossref","is-referenced-by-count":0,"title":["Explainable AI for gait speed analysis from multimodal data fusion"],"prefix":"10.1371","volume":"21","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-6776-7503","authenticated-orcid":true,"given":"Abdullah","family":"Alharthi","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Abdulrahman","family":"Al Ayidh","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Ahmed","family":"Alqurashi","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Mohammed","family":"Alammar","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Shoaib Shahriar","family":"Mohammad","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5130-263X","authenticated-orcid":true,"given":"Houssem","family":"Bouchekara","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Yusuf Abubakar","family":"Sha\u2019aban","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Turki Essa","family":"Alharbi","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"340","published-online":{"date-parts":[[2026,2,27]]},"reference":[{"issue":"2","key":"pone.0341067.ref001","doi-asserted-by":"crossref","first-page":"3362","DOI":"10.3390\/s140203362","article-title":"Gait analysis methods: an overview of wearable and non-wearable systems, highlighting clinical applications","volume":"14","author":"A Muro-de-la-Herran","year":"2014","journal-title":"Sensors (Basel)"},{"key":"pone.0341067.ref002","doi-asserted-by":"crossref","first-page":"489","DOI":"10.1109\/UBMK.2018.8566368","article-title":"Human Identification Based on Gait Analysis: A survey","volume-title":"2018 3rd International Conference on Computer Science and Engineering (UBMK), Sarajevo, Bosnia and Herzegovina","author":"RG Birdal","year":"2018"},{"key":"pone.0341067.ref003","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1109\/ICORR.2013.6650387","article-title":"Human-robot interaction tests on a novel robot for gait assistance","volume-title":"2013 IEEE 13th International Conference on Rehabilitation Robotics (ICORR)","author":"NL Tagliamonte","year":"2013"},{"key":"pone.0341067.ref004","doi-asserted-by":"crossref","first-page":"193966","DOI":"10.1109\/ACCESS.2020.3032202","article-title":"Gait Analysis for Early Neurodegenerative Diseases Classification Through the Kinematic Theory of Rapid Human Movements","volume":"8","author":"V Dentamaro","year":"2020","journal-title":"IEEE Access"},{"issue":"5","key":"pone.0341067.ref005","doi-asserted-by":"crossref","first-page":"441","DOI":"10.1093\/ageing\/29.5.441","article-title":"Walking speed as a good predictor for the onset of functional dependence in a Japanese rural community population","volume":"29","author":"S Shinkai","year":"2000","journal-title":"Age Ageing"},{"issue":"4","key":"pone.0341067.ref006","doi-asserted-by":"crossref","first-page":"265","DOI":"10.1016\/j.clineuro.2009.12.015","article-title":"Falls and gait disorders in geriatric neurology","volume":"112","author":"H Axer","year":"2010","journal-title":"Clin Neurol Neurosurg"},{"issue":"2","key":"pone.0341067.ref007","doi-asserted-by":"crossref","first-page":"231","DOI":"10.1007\/s11357-017-9973-y","article-title":"Cognitive status, fast walking speed and walking speed reserve-the Gait and Alzheimer Interactions Tracking (GAIT) study","volume":"39","author":"ML Callisaya","year":"2017","journal-title":"Geroscience"},{"key":"pone.0341067.ref008","doi-asserted-by":"crossref","first-page":"197","DOI":"10.1016\/j.humov.2016.03.010","article-title":"Gait variability in people with neurological disorders: A systematic review and meta-analysis","volume":"47","author":"Y Moon","year":"2016","journal-title":"Hum Mov Sci"},{"issue":"1","key":"pone.0341067.ref009","doi-asserted-by":"crossref","first-page":"128","DOI":"10.1016\/j.gaitpost.2006.08.010","article-title":"Walking speed influences on gait cycle variability","volume":"26","author":"K Jordan","year":"2007","journal-title":"Gait Posture"},{"issue":"2","key":"pone.0341067.ref010","doi-asserted-by":"crossref","first-page":"164","DOI":"10.1016\/j.gaitpost.2004.01.013","article-title":"Interstride trunk acceleration variability but not step width variability can differentiate between fit and frail older adults","volume":"21","author":"R Moe-Nilssen","year":"2005","journal-title":"Gait Posture"},{"key":"pone.0341067.ref011","doi-asserted-by":"crossref","first-page":"4189","DOI":"10.1109\/TNSRE.2023.3325215","article-title":"A Systematic Review of Gait Analysis in the Context of Multimodal Sensing Fusion and AI","volume":"31","author":"R Katmah","year":"2023","journal-title":"IEEE Trans Neural Syst Rehabil Eng"},{"issue":"2","key":"pone.0341067.ref012","doi-asserted-by":"crossref","first-page":"849","DOI":"10.3390\/s23020849","article-title":"Novel Deep Learning Network for Gait Recognition Using Multimodal Inertial Sensors","volume":"23","author":"L-F Shi","year":"2023","journal-title":"Sensors (Basel)"},{"issue":"10","key":"pone.0341067.ref013","doi-asserted-by":"crossref","first-page":"986","DOI":"10.3390\/app7100986","article-title":"An Overview of Smart Shoes in the Internet of Health Things: Gait and Mobility Assessment in Health Promotion and Disease Monitoring","volume":"7","author":"B Eskofier","year":"2017","journal-title":"Applied Sciences"},{"issue":"2","key":"pone.0341067.ref014","doi-asserted-by":"crossref","first-page":"2255","DOI":"10.3390\/s120202255","article-title":"Gait analysis using wearable sensors","volume":"12","author":"W Tao","year":"2012","journal-title":"Sensors (Basel)"},{"issue":"1","key":"pone.0341067.ref015","doi-asserted-by":"crossref","DOI":"10.1038\/s41598-020-59810-w","article-title":"Corticomuscular control of walking in older people and people with Parkinson\u2019s disease","volume":"10","author":"L Roeder","year":"2020","journal-title":"Sci Rep"},{"key":"pone.0341067.ref016","doi-asserted-by":"crossref","first-page":"554","DOI":"10.1109\/CEC.2019.8790050","article-title":"Evolutionary Design of Recurrent Neural Network Architecture for Human Activity Recognition","volume-title":"2019 IEEE Congress on Evolutionary Com-putation (CEC)","author":"RA Viswambaran","year":"2019"},{"key":"pone.0341067.ref017","first-page":"238","article-title":"Deep Bidirectional GRU network for human activity recognition using Wearable inertial sensors","volume-title":"2022 3rd International Conference on Elec-tronic Communication and Artificial Intelligence (IWECAI)","author":"S Zhao","year":"2022"},{"key":"pone.0341067.ref018","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1109\/WCSP.2018.8555945","article-title":"A Multi-Layer Parallel LSTM Network for Human Activity Recognition with Smartphone Sensors","volume-title":"2018 10th International Conference on Wireless Communications and Signal Processing (WCSP)","author":"T Yu","year":"2018"},{"key":"pone.0341067.ref019","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1109\/TIM.2022.3232159","article-title":"Reduce System Redundancy and Optimize Sensor Disposition for EMG\u2013IMU Multimodal Fusion Human\u2013Machine Interfaces With XAI","volume":"72","author":"P Kang","year":"2023","journal-title":"IEEE Trans Instrum Meas"},{"key":"pone.0341067.ref020","article-title":"XAI-based gait analysis of patients walking with Knee-Ankle-Foot orthosis using video cameras","author":"A Mishra","year":"2024","journal-title":"ArXiv"},{"issue":"2","key":"pone.0341067.ref021","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1145\/3474121","article-title":"Explaining Machine Learning Models for Clinical Gait Analysis","volume":"3","author":"D Slijepcevic","year":"2021","journal-title":"ACM Trans Comput Healthcare"},{"issue":"3","key":"pone.0341067.ref022","doi-asserted-by":"crossref","first-page":"167","DOI":"10.3390\/bios12030167","article-title":"Explainable Artificial Intelligence and Wearable Sensor-Based Gait Analysis to Identify Patients with Osteopenia and Sarcopenia in Daily Life","volume":"12","author":"J-K Kim","year":"2022","journal-title":"Biosensors (Basel)"},{"issue":"1","key":"pone.0341067.ref023","doi-asserted-by":"crossref","first-page":"5998","DOI":"10.1038\/s41598-024-56656-4","article-title":"Identification and interpretation of gait analysis features and foot conditions by explainable AI","volume":"14","author":"ME \u00d6zate\u015f","year":"2024","journal-title":"Sci Rep"},{"key":"pone.0341067.ref024","doi-asserted-by":"crossref","first-page":"453","DOI":"10.1109\/IEMBS.2011.6090063","article-title":"Automatic classification of pathological gait patterns using ground reaction forces and machine learning algorithms","volume-title":"2011 Annual International Conference of the IEEE Engineering in Medicine and Biology Society","author":"M Alaqtash","year":"2011"},{"key":"pone.0341067.ref025","doi-asserted-by":"crossref","first-page":"65906","DOI":"10.1109\/ACCESS.2023.3289986","article-title":"Explainable Machine Learning in Human Gait Analysis: A Study on Children With Cerebral Palsy","volume":"11","author":"D Slijepcevic","year":"2023","journal-title":"IEEE Access"},{"issue":"1","key":"pone.0341067.ref026","doi-asserted-by":"crossref","first-page":"111","DOI":"10.1038\/s41597-019-0124-4","article-title":"A multimodal dataset of human gait at different walking speeds established on injury-free adult participants","volume":"6","author":"C Schreiber","year":"2019","journal-title":"Sci Data"},{"key":"pone.0341067.ref027","article-title":"A multimodal dataset of human gait at different walking speeds established on injury-free adult participants","author":"C Schreiber","year":"2019","journal-title":"figshare"},{"key":"pone.0341067.ref028","unstructured":"Mokka - Motion Kinematic & Kinetic Analyzer. [online] Available: http:\/\/biomechanical-toolkit.github.io\/mokka\/index.html"},{"issue":"15","key":"pone.0341067.ref029","doi-asserted-by":"crossref","first-page":"16904","DOI":"10.1109\/JSEN.2021.3078336","article-title":"Spatiotemporal Analysis by Deep Learning of Gait Signatures From Floor Sensors","volume":"21","author":"AS Alharthi","year":"2021","journal-title":"IEEE Sensors J"},{"key":"pone.0341067.ref030","article-title":"Attention is all you need","volume":"30","author":"A Vaswani","year":"2017","journal-title":"Advances in Neural Information Processing Systems"},{"key":"pone.0341067.ref031","article-title":"Transformers in Time Series: A Survey","author":"Q Wen","year":"2022","journal-title":"arXiv"},{"key":"pone.0341067.ref032","doi-asserted-by":"crossref","first-page":"211","DOI":"10.1016\/j.patcog.2016.11.008","article-title":"Explaining nonlinear classification decisions with deep Taylor decomposition","volume":"65","author":"G Montavon","year":"2017","journal-title":"Pattern Recognition"},{"issue":"11","key":"pone.0341067.ref033","doi-asserted-by":"crossref","first-page":"2660","DOI":"10.1109\/TNNLS.2016.2599820","article-title":"Evaluating the Visualization of What a Deep Neural Network Has Learned","volume":"28","author":"W Samek","year":"2017","journal-title":"IEEE Trans Neural Netw Learn Syst"},{"issue":"3","key":"pone.0341067.ref034","doi-asserted-by":"crossref","first-page":"247","DOI":"10.1109\/JPROC.2021.3060483","article-title":"Explaining Deep Neural Networks and Beyond: A Review of Methods and Applications","volume":"109","author":"W Samek","year":"2021","journal-title":"Proc IEEE"},{"key":"pone.0341067.ref035","unstructured":"Smilkov N, Thorat B, Kim F, Vi\u00e9gas M, Wattenberg M. Smooth-Grad: Removing noise by adding noise. 2017. doi: 10.48550\/arXiv.1706.03825"}],"container-title":["PLOS One"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pone.0341067","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,2,27]],"date-time":"2026-02-27T18:35:31Z","timestamp":1772217331000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pone.0341067"}},"subtitle":[],"editor":[{"given":"Anwar","family":"P.P. Abdul Majeed","sequence":"first","affiliation":[],"role":[{"role":"editor","vocabulary":"crossref"}]}],"short-title":[],"issued":{"date-parts":[[2026,2,27]]},"references-count":35,"journal-issue":{"issue":"2","published-online":{"date-parts":[[2026,2,27]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pone.0341067","relation":{},"ISSN":["1932-6203"],"issn-type":[{"value":"1932-6203","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,2,27]]}}}