Leveraging Imperfect Medical Data: A Manifold-Consistent Spatio-Temporal Network for Sensor-based Human Activity Recognition

2026-04-29

AI Summary

This work addresses the challenge of incomplete physiological signals in wearable healthcare IoT systems caused by sensor missingness, failures, and noise, proposing a Manifold-Consistent Spatio-Temporal Network (MCSTN) to enhance the robustness of human activity recognition under realistic imperfect sensing conditions. The method introduces a novel dual-level perturbation mechanism—combining physics-informed and diffusion-driven strategies—to simulate real-world sensor deficiencies, and leverages multi-view consistency constraints to learn perturbation-invariant semantic representations. Furthermore, it employs a decoupled dual-stream architecture that separately captures long-term temporal dynamics and inter-sensor spatial correlations. Extensive experiments on three benchmark datasets—PAMAP2, Opportunity, and WISDM—demonstrate that MCSTN significantly outperforms state-of-the-art approaches under imperfect sensing scenarios, exhibiting superior robustness and practical applicability.

Abstract

Sensor-based Human Activity Recognition (HAR) has attracted increasing attention in medical and healthcare monitoring, particularly with the growth of Internet of Medical Things (IoMT). However, in real-world wearable sensing scenarios, IoMT signals are often corrupted by missing measurements, sensor failures, and environmental noise, which significantly degrade the performance of conventional deep learning models that assume clean and complete inputs. To address this challenge, we propose a Manifold-Consistent Spatio-Temporal Network (MCSTN) for robust HAR under imperfect sensing conditions. The proposed framework introduces a dual-level corruption modeling mechanism that simulates realistic sensor imperfections through both physical-level corruption and diffusion-driven continuous corruption. By enforcing representation consistency across multiple corrupted views, the model learns stable and corruption-invariant semantic representations. Furthermore, we design a dual-stream spatio-temporal architecture that explicitly decouples temporal dynamics modeling and spatial correlation learning. The temporal stream captures long-term activity dynamics, while the spatial stream models inter-sensor relationships, enabling more effective spatio-temporal representation learning. Extensive experiments on three widely used HAR benchmark datasets, PAMAP2, Opportunity, and WISDM, demonstrate that the proposed MCSTN achieves competitive performance compared with existing state-of-the-art methods, particularly under imperfect sensing conditions. These results validate the effectiveness and robustness of the proposed framework for real-world wearable IoMT sensing applications.

Problem

• Human Activity Recognition
• Imperfect Medical Data
• Sensor-based HAR
• IoMT
• Robustness

Innovation

• manifold consistency
• spatio-temporal decoupling
• corruption-invariant representation
• dual-stream architecture
• sensor-based HAR