Every Step of the Way: Video-based Parkinsonian Turning Step Counting

2026-06-26

AI Summary

This work addresses the challenge of continuously monitoring turning-related gait in Parkinson’s disease patients, whose complex and atypical shuffling steps are poorly captured by existing wearable-based methods. The authors propose an unobtrusive video analysis framework that first estimates coarse step counts from 3D human mesh-derived foot trajectories, then models fine-grained gait dynamics by integrating optical flow with mesh features. A cross-attention mechanism fuses these coarse and fine motion cues, and multiple instance learning aggregates segment-level video embeddings to predict total step counts. This approach achieves high-accuracy, device-free turning step estimation for the first time, significantly outperforming state-of-the-art methods on real-world Parkinson’s datasets. Its core innovation lies in a coarse-to-fine multimodal motion representation and a cross-attention query architecture that effectively captures subtle gait abnormalities.

Abstract

As a prominent symptom of Parkinson's disease (PD), turning impairment is evaluated through parameters such as turning angle, duration, and particularly, the number of steps required to complete a turn, which directly reflects motor dysfunction. Accurate step counting is challenging due to variability in real-world turning movements and atypical shuffling patterns in parkinsonian gait. Existing methods are predominantly wearable-based, requiring users to wear and manage dedicated devices, which can be inconvenient for continuous daily use. To address this, we propose a passive, video-based framework that estimates step count in a coarse-to-fine manner using diverse motion representations. Specifically, an initial step count is estimated from foot movement signals derived from 3D human mesh recovery, providing high-level motion structures. To incorporate fine-grained motion details, a motion encoder learns complementary gait dynamics from mesh and optical flow to refine the initial estimate. In this process, coarse foot movement signals query the pixel-level motion cues via cross attention to capture subtle parkinsonian gait dynamics. To handle varying video lengths, we partition each video into clips and integrate clip-wise motion embeddings via multiple instance learning (MIL) for step count residual prediction. Extensive experiments show our method consistently outperforms existing step counting methods on real-world PD turning datasets.

Problem

• Parkinson's disease
• turning impairment
• step counting
• gait analysis
• video-based assessment

Innovation

• video-based step counting
• coarse-to-fine estimation
• 3D human mesh recovery
• cross-attention motion fusion
• multiple instance learning