TransfHAR: Self-Supervised Wrist Representations for On-Demand Activity Recognition

2026-08-16

AI Summary

This study addresses the scarcity of fine-grained annotations and personalization challenges in wrist-based activity recognition by proposing a self-supervised pre-training framework for wrist IMUs that leverages coarse-grained unlabeled data. This approach effectively transfers motion priors to enable few-shot, on-demand recognition. Notably, this work is the first to validate the transferability of coarse-grained pre-training to unseen fine-grained tasks, implemented as a real-time smartwatch application supporting user-defined activities. Cross-dataset evaluations demonstrate that the proposed method achieves an average balanced accuracy exceeding fully supervised baselines by 6.2%. Furthermore, in a ten-subject experiment, five-shot accuracy reached 86.7%, improving to 90.4% after one minute of incremental updating, significantly outperforming existing state-of-the-art methods.

Abstract

Fine-grained wrist activity recognition can support applications such as procedural step guidance and context-aware assistance, yet acquiring labeled data for every new task, user, and activity granularity remains a bottleneck. We present TransfHAR, a self-supervised wrist IMU framework for on-demand, fine-grained activity recognition by learning transferable motion priors from global, unlabeled activities. We show that self-supervised pretraining on coarse wrist IMU activities (e.g., sitting, walking, exercise) learns motion structure rich enough to transfer to fine-grained manipulative, gestural, and procedural activities (e.g., snapping, stirring, waving) that are absent from pretraining. We implement TransfHAR as a real-time smartwatch application that lets users define and expand their own activity set for personalized recognition from only a few demonstrations. Across three offline cross-dataset evaluations, TransfHAR matches or exceeds fully supervised baselines that use complete label sets with equal or additional sensor channels, by 6.2 balanced-accuracy points on average. In an in-lab study with 10 participants each performing seven novel wrist activities, TransfHAR reaches 86.7% balanced accuracy across participants with five examples per class and 90.4% when updated from a single one-minute recording per class. These results indicate that broad self-supervised wrist pretraining provides an effective foundation for on-demand fine-grained activity recognition.

Problem

• Wrist Activity Recognition
• Labeled Data Bottleneck
• Fine-grained Recognition
• On-Demand Recognition

Innovation

• Self-Supervised Learning
• Wrist IMU
• Fine-grained Activity Recognition
• Few-shot Learning
• Transferable Motion Priors