Rotation-Invariant Multi-IMU Activity Recognition under Independent Per-Location Orientation Shifts

2026-08-16

AI Summary

This study addresses orientation misalignment caused by device re-wearing in multi-IMU human activity recognition by proposing the TRI-HAR framework. The method internalizes rotational robustness as an intrinsic architectural property through an SO(3)-equivariant backbone, tri-axial vector reconstruction, and invariant projection, thereby eliminating reliance on data augmentation or calibration procedures. Experimental evaluations across four benchmark datasets demonstrate that TRI-HAR maintains stable macro-F1 scores and significantly outperforms traditional rotation-augmented baselines under target misalignment scenarios. Consequently, this framework effectively achieves robust multi-sensor feature fusion and classification without requiring additional preprocessing, offering a structurally principled solution to orientation variability in wearable sensing applications.

Abstract

Human Activity Recognition (HAR) with self-administered wearables, such as at-home rehabilitation and exercise monitoring, often requires reattaching inertial measurement units (IMUs) across sessions. In multi-IMU settings, this can induce independent orientation offsets across body locations, a deployment shift that conventional scalar HAR models do not structurally handle. Existing remedies rely on rotation augmentation, whose robustness depends on sampled transformations, or calibration and orientationnormalization pipelines requiring additional reference-frame assumptions or explicit procedures. We present Truly Rotation-Invariant HAR (TRI-HAR), a rotation-invariant framework that makes robustness to independent per-location IMU orientation offsets a structural model property. TRI-HAR reshapes accelerometer and gyroscope streams into triaxial vectors, applies a shared SO(3)-equivariant backbone and invariant projection to each IMU location, and fuses the resulting invariant features for activity classification. Across four multi-IMU benchmarks, TRI-HAR preserves macro-F1 under fixed independent per-location SO(3) rotations and outperforms rotation-augmented baselines under this target shift without requiring rotational augmentation.

Problem

• Human Activity Recognition
• Multi-IMU
• Rotation Invariance
• Orientation Shifts
• Deployment Shift

Innovation

• Rotation-Invariant HAR
• SO(3)-equivariant backbone
• Independent per-location orientation shifts
• Invariant projection
• Multi-IMU fusion