LLM-Guided Exemplar Selection for Few-Shot Wearable-Sensor Human Activity Recognition

2025-12-26

AI Summary

To address the bottlenecks in few-shot wearable-sensor-based human activity recognition (HAR)—namely, heavy reliance on large-scale labeled data and poor discriminability among semantically similar activities (e.g., walking, upstairs, downstairs)—this paper proposes a cognition-driven, semantic-enhanced exemplar selection method. It introduces, for the first time in HAR, semantic priors generated by large language models (LLMs), including feature importance, inter-class confusion scores, and sample budget adjustment factors, to guide exemplar selection by jointly leveraging semantic, structural, and geometric cues. The method integrates LLM knowledge distillation, margin-based validation, PageRank centrality, hubness-aware penalization, and facility-location optimization. Evaluated under the few-shot setting of UCI-HAR, it achieves a macro-F1 score of 88.78%, significantly outperforming baselines including random sampling, herding, and k-center selection. This work establishes a novel, interpretable, and highly discriminative paradigm for exemplar selection in low-resource HAR.

Technology Category

• Machine Learning: Active Learning
• Search and Optimization: Sampling/Simulation-based Search
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

In this paper, we propose an LLM-Guided Exemplar Selection framework to address a key limitation in state-of-the-art Human Activity Recognition (HAR) methods: their reliance on large labeled datasets and purely geometric exemplar selection, which often fail to distinguish similar weara-ble sensor activities such as walking, walking upstairs, and walking downstairs. Our method incorporates semantic reasoning via an LLM-generated knowledge prior that captures feature importance, inter-class confusability, and exemplar budget multipliers, and uses it to guide exemplar scoring and selection. These priors are combined with margin-based validation cues, PageRank centrality, hubness penalization, and facility-location optimization to obtain a compact and informative set of exemplars. Evaluated on the UCI-HAR dataset under strict few-shot conditions, the framework achieves a macro F1-score of 88.78%, outperforming classical approaches such as random sampling, herding, and $k$-center. The results show that LLM-derived semantic priors, when integrated with structural and geometric cues, provide a stronger foundation for selecting representative sensor exemplars in few-shot wearable-sensor HAR.

Problem

• Selects representative exemplars for few-shot wearable-sensor activity recognition
• Addresses reliance on large labeled datasets and geometric-only selection methods
• Distinguishes similar activities like walking upstairs and downstairs using semantic reasoning

Innovation

• LLM-guided semantic reasoning for exemplar selection
• Integration of margin-based validation with PageRank centrality
• Facility-location optimization for compact informative exemplars