Reducing Label Dependency in Human Activity Recognition with Wearables: From Supervised Learning to Novel Weakly Self-Supervised Approaches

TL;DR

This work investigates label efficiency in wearable HAR by spanning the supervision spectrum—from fully supervised to unsupervised—and introduces a novel weakly self-supervised framework that blends domain knowledge with minimal labeled data. By comparing supervised, unsupervised, weakly supervised (single- and multi-task), self-supervised, and the proposed weakly self-supervised methods across three benchmark HAR datasets, the authors show that weakly and weakly self-supervised approaches can achieve competitive or superior performance with far less labeling. Key contributions include a domain-informed self-supervised objective (temporal and feature consistency), a weakly supervised multi-task architecture leveraging activity and person signals, and a two-stage weakly self-supervised framework that attains up to 99.04% accuracy on PAMAP2 with only 10% labeled data. The findings offer practical guidance for deploying HAR systems under labeling constraints and highlight the complementary strengths of combined learning paradigms for robust, label-efficient wearable sensing.

Abstract

Human activity recognition (HAR) using wearable sensors has advanced through various machine learning paradigms, each with inherent trade-offs between performance and labeling requirements. While fully supervised techniques achieve high accuracy, they demand extensive labeled datasets that are costly to obtain. Conversely, unsupervised methods eliminate labeling needs but often deliver suboptimal performance. This paper presents a comprehensive investigation across the supervision spectrum for wearable-based HAR, with particular focus on novel approaches that minimize labeling requirements while maintaining competitive accuracy. We develop and empirically compare: (1) traditional fully supervised learning, (2) basic unsupervised learning, (3) a weakly supervised learning approach with constraints, (4) a multi-task learning approach with knowledge sharing, (5) a self-supervised approach based on domain expertise, and (6) a novel weakly self-supervised learning framework that leverages domain knowledge and minimal labeled data. Experiments across benchmark datasets demonstrate that: (i) our weakly supervised methods achieve performance comparable to fully supervised approaches while significantly reducing supervision requirements; (ii) the proposed multi-task framework enhances performance through knowledge sharing between related tasks; (iii) our weakly self-supervised approach demonstrates remarkable efficiency with just 10\% of labeled data. These results not only highlight the complementary strengths of different learning paradigms, offering insights into tailoring HAR solutions based on the availability of labeled data, but also establish that our novel weakly self-supervised framework offers a promising solution for practical HAR applications where labeled data are limited.

Figures (15)

• Figure S1: (a) ResNet, (b) Siamese architecture, and (c) TCN block.
• Figure S2: The TCN architecture used in the supervised approach.
• Figure S3: The residual autoencoder architecture used in the unsupervised approach.
• Figure S4: The Siamese architecture used in the weakly supervised single-task approach.
• Figure S5: The extended Siamese architecture used in the weakly supervised multi-task approach.