PressTrack-HMR: Pressure-Based Top-Down Multi-Person Global Human Mesh Recovery

2025-11-12

AI Summary

To address signal aliasing in tactile pressure data caused by multiple individuals walking randomly on a pressure-sensitive mat, this paper proposes the first framework for multi-person global human mesh recovery from tactile pressure signals. Methodologically, it adopts a tracking-by-detection paradigm, integrating temporal pressure signal segmentation, unsupervised temporal clustering, and graph neural network–based optimization to achieve individual pressure source separation, cross-frame identity association, and monocular human mesh reconstruction. To support this work, we introduce MIP—the first large-scale, multi-person interactive pressure dataset. Experiments demonstrate robust performance under occlusion and low-light conditions, achieving an MPJPE of 89.2 mm and a WA-MPJPE₁₀₀ of 112.6 mm. These results validate the effectiveness and state-of-the-art capability of tactile floor mats for privacy-preserving group behavioral sensing.

Technology Category

• Humans and AI: Interaction Techniques and Devices
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Computer Vision: Motion & Tracking

Application Category

• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Multi-person global human mesh recovery (HMR) is crucial for understanding crowd dynamics and interactions. Traditional vision-based HMR methods sometimes face limitations in real-world scenarios due to mutual occlusions, insufficient lighting, and privacy concerns. Human-floor tactile interactions offer an occlusion-free and privacy-friendly alternative for capturing human motion. Existing research indicates that pressure signals acquired from tactile mats can effectively estimate human pose in single-person scenarios. However, when multiple individuals walk randomly on the mat simultaneously, how to distinguish intermingled pressure signals generated by different persons and subsequently acquire individual temporal pressure data remains a pending challenge for extending pressure-based HMR to the multi-person situation. In this paper, we present extbf{PressTrack-HMR}, a top-down pipeline that recovers multi-person global human meshes solely from pressure signals. This pipeline leverages a tracking-by-detection strategy to first identify and segment each individual's pressure signal from the raw pressure data, and subsequently performs HMR for each extracted individual signal. Furthermore, we build a multi-person interaction pressure dataset extbf{MIP}, which facilitates further research into pressure-based human motion analysis in multi-person scenarios. Experimental results demonstrate that our method excels in multi-person HMR using pressure data, with 89.2 $mm$ MPJPE and 112.6 $mm$ WA-MPJPE$_{100}$, and these showcase the potential of tactile mats for ubiquitous, privacy-preserving multi-person action recognition. Our dataset&code are available at https://github.com/Jiayue-Yuan/PressTrack-HMR.

Problem

• Distinguishing intermingled pressure signals from multiple individuals on tactile mats
• Acquiring individual temporal pressure data in multi-person walking scenarios
• Extending pressure-based human mesh recovery from single-person to multi-person situations

Innovation

• Top-down pipeline for multi-person mesh recovery
• Tracking-by-detection strategy to segment pressure signals
• Pressure-based human mesh recovery using tactile mats