[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-327712-105":3,"detail-sidebar-cat-0-en-105":81,"doc-detail-327712-en":131},{"code":4,"msg":5,"data":6},0,"ok",{"site_id":7,"language":8,"slug":9,"title":10,"keywords":11,"description":12,"schema_data":13,"social_meta":74,"head_meta":76,"extra_data":78,"updated_unix":80},105,"en","pedestriandiffusion-multimodal-generative-denoising-and-dense-state-estimation-for-inertial-navigation","PedestrianDiffusion - Multimodal Generative Denoising and Dense State Estimation for Inertial Navigation","","Accuracy of consumer-grade inertial navigation is limited by stochastic MEMS noise, which can cause estimation jittering and loss of high-frequency kinematic fidelity in deterministic and MSE-optimized neural regressors. PedestrianDiffusion introduces a multimodal spectral-domain generative denoising framework that reformulates dense 6D state estimation as continuous conditional reverse diffusion, stabilizing trajectories via spectral covariance bounding. It also adds zero-shot semantic conditioning using vision-language embeddings and enables edge-feasible single-step probability flow ODE inference (T=1), validated across OxIOD, RIDI, RoNIN, and TLIO.",{"@graph":14,"@context":73},[15,34,56],{"@type":16,"itemListElement":17},"BreadcrumbList",[18,23,27,31],{"item":19,"name":20,"@type":21,"position":22},"https://docshare.wps.com","Home","ListItem",1,{"item":24,"name":25,"@type":21,"position":26},"https://docshare.wps.com/document/","Document",2,{"item":28,"name":29,"@type":21,"position":30},"https://docshare.wps.com/document/research-report/","Research & Report",3,{"item":32,"name":10,"@type":21,"position":33},"https://docshare.wps.com/document/pedestriandiffusion-multimodal-generative-denoising-and-dense-state-estimation-for-inertial-navigation/327712/",4,{"url":32,"name":10,"@type":35,"image":36,"author":41,"headline":10,"publisher":44,"fileFormat":47,"inLanguage":8,"description":12,"dateModified":48,"datePublished":49,"encodingFormat":47,"isAccessibleForFree":50,"interactionStatistic":51},"DigitalDocument",{"url":37,"@type":38,"width":39,"height":40},"https://docshare.wps.com/thumbnails/pedestriandiffusion-multimodal-generative-denoising-and-dense-state-estimation-for-inertial-navigation/327712.png","ImageObject",300,407,{"name":42,"@type":43},"Liam","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-11-04","2026-09-21",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",18,{"@type":57,"mainEntity":58},"FAQPage",[59,65,69],{"name":60,"@type":61,"acceptedAnswer":62},"What problem does PedestrianDiffusion target in consumer-grade inertial navigation?","Question",{"text":63,"@type":64},"It targets stochastic MEMS noise that makes traditional and even MSE-based neural approaches suffer estimation jittering and reduced high-frequency kinematic fidelity.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"How does PedestrianDiffusion reformulate dense 6D state estimation?",{"text":68,"@type":64},"It converts the task into a multimodal spectral-domain continuous conditional denoising process (reverse diffusion) operating in the frequency domain to better isolate periodic motion from noise.",{"name":70,"@type":61,"acceptedAnswer":71},"Why is the approach considered practical for edge hardware?",{"text":72,"@type":64},"It uses a single-step deterministic probability flow ODE solver with T=1, producing an amortized latency of 9.88ms per frame per inference, supporting computationally viable generative tracking for high-throughput buffering.","https://schema.org",{"og:url":32,"og:type":75,"og:title":10,"og:site_name":45,"og:description":12},"article",{"robots":77,"canonical":32},"index,follow",{"doc_id":79,"site_id":7},327712,1790216850,{"code":4,"msg":82,"data":83},"success",[84,88,92,96,101,106,111,115,120,123,127],{"id":22,"doc_module":4,"doc_module_name":25,"category_name":85,"show_sort_weight":86,"slug":87},"Story & Novel",90,"story-novel",{"id":26,"doc_module":4,"doc_module_name":25,"category_name":89,"show_sort_weight":90,"slug":91},"Literature",80,"literature",{"id":33,"doc_module":4,"doc_module_name":25,"category_name":93,"show_sort_weight":94,"slug":95},"Exam",70,"exam",{"id":97,"doc_module":4,"doc_module_name":25,"category_name":98,"show_sort_weight":99,"slug":100},5,"Comic",60,"comic",{"id":102,"doc_module":4,"doc_module_name":25,"category_name":103,"show_sort_weight":104,"slug":105},6,"Technology",50,"technology",{"id":107,"doc_module":4,"doc_module_name":25,"category_name":108,"show_sort_weight":109,"slug":110},7,"Healthcare",40,"healthcare",{"id":112,"doc_module":4,"doc_module_name":25,"category_name":29,"show_sort_weight":113,"slug":114},8,30,"research-report",{"id":116,"doc_module":4,"doc_module_name":25,"category_name":117,"show_sort_weight":118,"slug":119},9,"Religion & Spirituality",20,"religion-spirituality",{"id":118,"doc_module":4,"doc_module_name":25,"category_name":121,"show_sort_weight":118,"slug":122},"World Cup","world-cup",{"id":124,"doc_module":4,"doc_module_name":25,"category_name":125,"show_sort_weight":124,"slug":126},10,"Lifestyle","lifestyle",{"id":128,"doc_module":4,"doc_module_name":25,"category_name":129,"show_sort_weight":97,"slug":130},19,"General","general",{"code":4,"msg":82,"data":132},{"doc_id":79,"user_id":133,"nickname":42,"user_avatar":134,"doc_module":4,"category_id":112,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":55,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":140,"language":141,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":142,"faqs":143,"seo_title":144,"seo_description":12,"update_tm":145,"read_time":113,"preview_page_count":4},8796095461564,"https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d","PedestrianDiffusion: Multimodal Generative Denoising and Dense State Estimation for Inertial  \nNavigation  \nI-Hao Lu , Dongsoo Han  Senior Member, IEEE,  \narXiv :2607 .03349v 1 [ cs .LG] 3 Jul 2026  \nAbstract—The accuracy of consumer-grade inertial navigation is bottlenecked by the stochastic noise of Micro-ElectroMechanical Systems (MEMS). Traditional deterministic neural architectures often succumb to “estimation jittering,” sacrificing high-frequency kinematic fidelity for numerical stability. We propose PedestrianDiffusion, a multimodal spectral-domain generative framework reformulating dense 6D state estimation as a continuous conditional denoising process. By operating in the frequency domain, our formulation bounds the spectral covariance, acting as a mathematical preconditioner to stabilize the reverse diffusion trajectory. Furthermore, we introduce a zero-shot semantic conditioning mechanism leveraging visionlanguage embeddings as categorical priors to generalize across heterogeneous sensor noise profiles. To address the computational intractability of generative tracking, we deploy a single-step deterministic probability flow ODE solver (T = 1). This yields highcapacity asynchronous batch trajectory refinement, establishing the viability of generative architectures for asynchronous batch trajectory refinement on edge hardware. Extensive evaluations on the OxIOD, RIDI, RoNIN, and TLIO benchmarks demonstrate that PedestrianDiffusion achieves state-of-the-art performance, exhibiting unprecedented robustness to impulse perturbations and coupled 6D kinematic drift. This work provides a rigorous algorithmic blueprint for next-generation Neural Inertial Measurement Units (N-IMUs).  \nIndex Terms—IMUs, Navigation, diffusion, high-fidelity, spectrum, N-IMUs  \nI. INTRODUCTION  \nDespite the widespread integration of low-cost MEMS Inertial Measurement Units (IMUs), achieving navigation-grade accuracy remains elusive due to stochastic noise and bias instability [1] . Traditional filtering techniques (e.g., ZUPT [2], EKF [3]) depend on restrictive physical priors that frequently fail in unconstrained environments. While deep learning architectures such as RoNIN [4] and TLIO [5] provide robust alternatives, they predominantly formulate navigation as a regression task. However, this paradigm has encountered a“estimation jittering” caused by optimizing for Mean Squared Error (MSE), discriminative models inevitably overestimate the velocity and lose high-frequency kinematic details to minimize variance. Consequently, these models yield jittered, overestimated trajectories—statistically reflected by Total Length Ratios (TLR) consistently below unity—trading physical fidelity for numerical stability.  \nTo overcome this limitation, we propose a conditional generative denoising paradigm that fundamentally departs from  \nThe source code for the proposed model is available on GitHub at [https:](https:)//[github.com/jacklu333333/PedestrianDiffusion](github.com/jacklu333333/PedestrianDiffusion)  \ndiscriminative regression. We introduce PedestrianDiffusion (PD), a multimodal spectral-domain diffusion framework designed to actively reconstruct the clean motion manifold from noisy sensor inputs. By shifting from time-domain regression to spectral processing via the Short-Time Fourier Transform (STFT), our approach inherently isolates periodic pedestrian kinematics from stochastic sensor noise. This formulation enables PD to achieve optimal accuracy across the OxIOD [6], RIDI [7], RoNIN [4], and TLIO [5] benchmarks.  \nFurthermore, we challenge the prevailing consensus that generative models are computationally prohibitive for edge deployment. Our empirical analysis yields a critical insight: spectral diffusion exhibits rapid convergence, saturating state estimation accuracy in a single inference step (T = 1) . This convergence translates to an amortized latency of 9.88ms per frame per inference. While strict hard real-time execution currently nece","cbCaiimlG4bchke8","https://ap.wps.com/l/cbCaiimlG4bchke8","pdf",3626688,12,"English","# Abstract\n# I. Introduction\n# II. Related Work","[{\"question\":\"What problem does PedestrianDiffusion target in consumer-grade inertial navigation?\",\"answer\":\"It targets stochastic MEMS noise that makes traditional and even MSE-based neural approaches suffer estimation jittering and reduced high-frequency kinematic fidelity.\"},{\"question\":\"How does PedestrianDiffusion reformulate dense 6D state estimation?\",\"answer\":\"It converts the task into a multimodal spectral-domain continuous conditional denoising process (reverse diffusion) operating in the frequency domain to better isolate periodic motion from noise.\"},{\"question\":\"Why is the approach considered practical for edge hardware?\",\"answer\":\"It uses a single-step deterministic probability flow ODE solver with T=1, producing an amortized latency of 9.88ms per frame per inference, supporting computationally viable generative tracking for high-throughput buffering.\"}]","PedestrianDiffusion - Multimodal Generative Denoising and Dense State Estimation for Inertial Navigation | PDF",1789977886]