Garment Inertial Denoiser (GID): Endowing Accurate Motion Capture via Loose IMU Denoiser

TL;DR

We address accurate motion capture using loose-fitting garments by introducing Garment Inertial Denoiser (GID), a lightweight Transformer-based denoiser that factorizes the problem into denoising and pose estimation. GID employs location-specific experts with a spatio-temporal backbone and a cross-wear fusion module to map loose-wear IMU signals to tight-wear equivalents, enabling robust full-body MoCap with limited paired data. A new GarMoCap dataset combines public and newly collected garment data to evaluate cross-user, cross-motion, and cross-garment generalization; experiments show GID improves state-of-the-art methods in real-time denoising and pose accuracy, even when trained on a single garment. The work demonstrates strong data efficiency and cross-domain generalization, with a roadmap toward physics-informed denoising foundations and broader garment diversity.

Abstract

Wearable inertial motion capture (MoCap) provides a portable, occlusion-free, and privacy-preserving alternative to camera-based systems, but its accuracy depends on tightly attached sensors - an intrusive and uncomfortable requirement for daily use. Embedding IMUs into loose-fitting garments is a desirable alternative, yet sensor-body displacement introduces severe, structured, and location-dependent corruption that breaks standard inertial pipelines. We propose GID (Garment Inertial Denoiser), a lightweight, plug-and-play Transformer that factorizes loose-wear MoCap into three stages: (i) location-specific denoising, (ii) adaptive cross-wear fusion, and (iii) general pose prediction. GID uses a location-aware expert architecture, where a shared spatio-temporal backbone models global motion while per-IMU expert heads specialize in local garment dynamics, and a lightweight fusion module ensures cross-part consistency. This inductive bias enables stable training and effective learning from limited paired loose-tight IMU data. We also introduce GarMoCap, a combined public and newly collected dataset covering diverse users, motions, and garments. Experiments show that GID enables accurate, real-time denoising from single-user training and generalizes across unseen users, motions, and garment types, consistently improving state-of-the-art inertial MoCap methods when used as a drop-in module.

Figures (4)

• Figure 1: Garment Inertial Denoiser (GIP) can denoise real-time loose wearing IMU data, enabling robust full-body motion capture.
• Figure 2: Motivation. (a–d) Loose-wear garments introduce location-dependent disturbances to IMU signals, including fabric sliding, impact, and garment deformation. (e) PCA of rotation signals shows that tight-wear IMUs form clean, body-part–specific manifolds, whereas (f) PCA of acceleration signals reveals large spread and collapsed structure under loose-wear conditions. These observations highlight the need for our GID module to recover tight-wear signal characteristics from loose-wear measurements.
• Figure 3: Garment Inertial Denoiser pipeline. (a) We factorize $f$ into two sub-functions corresponding to signal denoising and pose reconstruction. (b) Each sensor reading $\mathbf{IMU}_{loose}$ will be assigned to a Temporal-Spatial(TS) transformer-based location-specific denoiser.
• Figure 4: Prototype of our full-body garment-based motion capture system. The loose-wear jacket and pants are embedded with six IMU sensors and two central circuit boards for data acquisition in total.