Reinforcement Learning Driven Generalizable Feature Representation for Cross-User Activity Recognition

2025-08-31

AI Summary

To address degraded generalization in cross-user human activity recognition (HAR) caused by inter-subject variability in motion patterns, sensor placement, and physiological characteristics, this paper proposes a reinforcement learning–driven domain generalization framework. Methodologically, temporal feature extraction is formulated as a sequential decision-making process; autoregressive tokenization preserves intrinsic temporal structure, and a multi-objective reward function—designed without target-domain labels—jointly optimizes inter-class discriminability and user-invariance. The model integrates Transformer architectures with autoregressive generative mechanisms to learn robust, unsupervised representations. Evaluated on DSADS and PAMAP2 benchmarks, the approach significantly outperforms state-of-the-art methods in cross-user HAR accuracy, achieving high performance without subject-specific calibration. This enhances practicality and scalability in real-world wearable applications.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Activity and Plan Recognition

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Human Activity Recognition (HAR) using wearable sensors is crucial for healthcare, fitness tracking, and smart environments, yet cross-user variability -- stemming from diverse motion patterns, sensor placements, and physiological traits -- hampers generalization in real-world settings. Conventional supervised learning methods often overfit to user-specific patterns, leading to poor performance on unseen users. Existing domain generalization approaches, while promising, frequently overlook temporal dependencies or depend on impractical domain-specific labels. We propose Temporal-Preserving Reinforcement Learning Domain Generalization (TPRL-DG), a novel framework that redefines feature extraction as a sequential decision-making process driven by reinforcement learning. TPRL-DG leverages a Transformer-based autoregressive generator to produce temporal tokens that capture user-invariant activity dynamics, optimized via a multi-objective reward function balancing class discrimination and cross-user invariance. Key innovations include: (1) an RL-driven approach for domain generalization, (2) autoregressive tokenization to preserve temporal coherence, and (3) a label-free reward design eliminating the need for target user annotations. Evaluations on the DSADS and PAMAP2 datasets show that TPRL-DG surpasses state-of-the-art methods in cross-user generalization, achieving superior accuracy without per-user calibration. By learning robust, user-invariant temporal patterns, TPRL-DG enables scalable HAR systems, facilitating advancements in personalized healthcare, adaptive fitness tracking, and context-aware environments.

Problem

• Addressing cross-user variability in wearable sensor activity recognition
• Overcoming overfitting to user-specific patterns in supervised learning
• Eliminating dependency on domain labels for temporal generalization

Innovation

• Reinforcement learning driven domain generalization framework
• Transformer-based autoregressive tokenization for temporal coherence
• Multi-objective reward function balancing discrimination and invariance