RAG-HAR+: Towards Cost-Efficient LLM-Based Human Activity Recognition for Edge Deployment

2026-07-29

AI Summary

This work addresses the limited generalizability of traditional human activity recognition (HAR) methods, which rely heavily on dataset-specific designs and extensive labeled data, hindering adaptation to new sensors or activity classes. To overcome this, the authors propose a training-free, retrieval-augmented approach that constructs an offline library of statistical feature proxies and employs a retrieval-first strategy to guide large language model (LLM)-based classification. By integrating majority voting with an LLM-based fuzzy parsing mechanism, the method substantially reduces LLM invocation frequency, token consumption, and inference latency. Evaluated across six HAR benchmarks, it achieves competitive or superior performance while demonstrating practical feasibility for mobile deployment.

Abstract

Human Activity Recognition (HAR) from wearable sensors supports applications in healthcare, rehabilitation, fitness tracking, and smart environments. Yet, existing deep learning approaches require dataset-specific training, large labeled corpora, and repeated adaptation to new sensor settings or activity taxonomies. Retrieval-Augmented Generation for Human Activity Recognition (RAG-HAR) addresses this by framing HAR as a training-free, retrieval-augmented task, in which statistical descriptions of sensor windows are used to retrieve similar labeled examples that guide LLM-based classification. We introduce RAG-HAR+, a retrieval-first and cost-optimized extension that strengthens retrieval while reducing dependence on LLM-based inference. RAG-HAR+ uses an offline Retrieval Designer Agent to design dataset-specific feature groups from a diverse pool of motion descriptors, enabling sensor windows to be compared using features better aligned with dataset-specific activity patterns. During inference, RAG-HAR+ uses majority voting over retrieved neighbors for samples with strong retrieval evidence and defers only uncertain cases to an LLM-based Ambiguity Resolver Agent. Across six HAR benchmarks, RAG-HAR+ maintains competitive or improved performance while reducing LLM usage, token consumption, and inference time. We further extend the RAG-HAR mobile prototype to demonstrate the practical feasibility of retrieval-first, LLM-assisted HAR in mobile sensing scenarios.

Problem

• Human Activity Recognition
• Edge Deployment
• Retrieval-Augmented Generation
• LLM-based Inference
• Sensor Data

Innovation

• Retrieval-Augmented Generation
• Human Activity Recognition
• Edge Deployment
• Cost-Efficient LLM
• Training-Free Inference