Generating Personalized Lower-Limb Kinematics Across Walking Speeds Using Subject-Conditioned Diffusion

2026-07-08

AI Summary

Personalized exoskeleton assistance typically requires extensive user-specific gait data across multiple walking speeds, which is costly to collect and particularly burdensome for clinical populations. To address this challenge, this work proposes a subject-conditioned residual diffusion framework that generates lower-limb kinematic trajectories at unseen walking speeds using only a single gait sequence from a known speed. This approach introduces conditional diffusion models to cross-speed personalized gait synthesis for the first time, integrating a Transformer-based denoising network with feature-wise linear modulation (FiLM) to jointly condition on subject identity and target speed—preserving individual gait characteristics without requiring fine-tuning on stroke patient data. Experiments demonstrate mean angular errors of 3.4° and 6.0° on healthy subjects and stroke survivors, respectively, surpassing supervised feedforward baselines by over 70% and achieving performance with single-speed input comparable to that of four-speed inputs.

Abstract

Personalizing exoskeleton assistance requires user-specific gait data across many locomotor tasks, yet collecting this data demands repeated motion capture sessions that are costly, time-intensive, and especially burdensome for clinical populations. This challenge is most acute across walking speeds, where gait changes substantially and deviates further in clinical gait. This work introduces a subject-conditioned residual diffusion framework that generates personalized lower-limb kinematics at unseen walking speeds from a subject's gait sequence at a single seen speed. Given sagittal-plane hip, knee, and ankle trajectories at a seen speed and a desired unseen speed, the model generates a residual that transforms the seen trajectory into the unseen one, using a transformer denoiser conditioned on the subject's gait and the two speeds through feature-wise linear modulation. Trained only on able-bodied data, the model achieved a mean absolute error (MAE) of 3.4° on held-out able-bodied subjects. Without any stroke-specific fine-tuning, it achieved a 6.0° MAE on out-of-training-distribution stroke subjects, retaining subject identity for clinical gait. The framework reduced the MAE by over 70% relative to supervised feed-forward baselines, and a single seen speed matched the accuracy of four speeds within 0.4°. These results demonstrate that subject-conditioned residual diffusion can synthesize personalized gait across speeds from minimal data, reducing the collection burden for downstream exoskeleton personalization.

Problem

• personalized gait
• walking speeds
• exoskeleton personalization
• motion capture
• lower-limb kinematics

Innovation

• subject-conditioned diffusion
• personalized gait synthesis
• residual diffusion model
• walking speed generalization
• exoskeleton personalization