Confusion hierarchy cuts costs in recognizing human activities with sensors

CHEAR: Confusion-Hierarchy Evidence-Aware Reasoning for Cost-Efficient Human Activity Recognition

Artificial Intelligence

Summary

Recognizing what people are doing from wearable sensor data can require processing many possible activities, which is slow and costly. The authors create a method called CHEAR that organizes activities into a hierarchy using past mistakes, so the system compares likely activities first and narrows down options efficiently. This saves a lot of computing time without hurting accuracy in identifying activities. CHEAR works well across multiple datasets and helps large language models recognize human motions more cheaply while keeping good performance.

What this means in practice

  • For wearable device developers: Design energy-efficient activity recognition systems that reduce computation by focusing on probable activities first and only use detailed knowledge when needed.
  • For health monitoring teams: Implement faster and cost-effective human activity classifiers in health tracking devices that maintain accuracy across different user groups.

Authors

Xiangqing Lu, Tao Zhu, Mingxing Nie, Liming Chen

Abstract

Large language models (LLMs) offer a new approach to wearable-sensor-based human activity recognition. Retrieval-augmented generation (RAG) further introduces similar motion examples and discriminative sensor knowledge, enabling decisions grounded in observations relevant to the current query. More evidence, however, does not necessarily mean better-organized evidence. When online prompts and LLM agents primarily orchestrate inference, decisions about which activities to compare and when to invoke discriminative knowledge remain concentrated at inference time. As the activity vocabulary grows, the model must repeatedly process more candidates and associated knowledge, whereas pruning candidates too early risks excluding the true activity. We propose confusion-hierarchy evidence-aware reasoning (CHEAR), which moves part of this knowledge organization offline. CHEAR constructs and validates an activity hierarchy from recurrent cross-subject retrieval errors and builds corresponding activity-pair knowledge. At inference time, the hierarchy orders candidate comparisons; the LLM retains candidates progressively, and local discriminative knowledge is accessed only for the remaining activities. Across eight datasets, CHEAR achieves 86.79\% mean cross-subject accuracy, a 75.91\% candidate reduction rate, and 96.44\% true-activity retention. Per-sample token use is 85.15\% lower on average than in a variant without the hierarchical structure. Mean accuracy across eight cross-dataset transfer directions is 88.81\%. These results show that organizing candidates and discriminative knowledge around observed activity confusions reduces online LLM input while maintaining recognition performance.