Problem
Research questions and friction points this paper is trying to address.
pressure-based human pose monitoring
multi-device pressure images
3D human mesh estimation
limited monitoring range
unobtrusive sensing
PressureMesh: 3D Human Mesh Estimation from Multi-Device Pressure Images
🤖 AI Summary
This work addresses the limitation of existing pressure-image-based human pose monitoring methods, which are typically confined to single-device setups and thus hinder large-scale deployment. To overcome this, we propose MDP-Net, an end-to-end deep network that leverages a multimodal fusion mechanism inspired by Mixture of Experts (MoE) to effectively integrate temporal pressure images from multiple devices and directly reconstruct a 3D human body mesh. To facilitate training and evaluation, we introduce MDP, a high-quality multi-device pressure dataset. Experimental results demonstrate that our approach achieves a joint localization error of 12.6 cm on the MDP dataset, substantiating the effectiveness and potential of multi-device pressure fusion for everyday human pose monitoring.
📝 Abstract
Human pose monitoring is crucial in fields such as rehabilitation assessment and human-computer interaction. Due to its privacy-preserving nature, pressure-based human pose monitoring has become a primary approach for unobtrusive sensing. However, existing methods are generally limited to a single device, which restricts the effective monitoring range. To address this limitation, we propose MDP-Net, an end-to-end network capable of directly estimating human meshes from temporal pressure data across multiple devices. We introduce a multimodal fusion mechanism inspired by the Mixture of Experts (MoE) framework to achieve effective complementarity and enhancement of cross-device pressure information. To support the training and evaluation of MDP-Net, we constructed MDP, a high-quality multi-device temporal pressure dataset that includes various pose labels such as 2D/3D joints and human meshes. Experimental results demonstrate that MDP-Net achieves a joint position error of 12.6 cm on the MDP dataset. These results prove that fusing multi-device pressure information is an effective and promising new solution for daily human pose monitoring.
Innovation
Methods, ideas, or system contributions that make the work stand out.
multi-device pressure sensing
3D human mesh estimation
Mixture of Experts
temporal pressure data
MDP-Net
🔎 Similar Papers
No similar papers found.
