GenHAR: Generalizing Cross-domain Human Activity Recognition for Last-mile Delivery

2026-05-21

AI Summary

This study addresses the performance degradation in human activity recognition caused by distribution shifts across sensor domains. To enhance generalization to unseen target domains using only source-domain data, the authors propose GenHAR, a novel framework that learns domain-invariant representations without requiring target-domain samples during training. GenHAR innovatively tokenizes sensor signals, models inter-channel correlations in the frequency domain, and integrates selective masking with an efficient attention mechanism. Experimental results demonstrate that GenHAR outperforms state-of-the-art methods by 9.97% in accuracy on real-world datasets while reducing floating-point operations by 6.4×. The system has been deployed across four cities and has performed 2.15 billion real-time activity detections to date.

Abstract

Human Activity Recognition (HAR) has shown remarkable effectiveness in various applications, such as smart healthcare and intelligent manufacturing. However, a major challenge faced by HAR is the distribution shift across different sensor data domains, which often leads to decreased performance when deployed for real-world applications. To address this issue, this paper introduces GenHAR, a novel framework designed to mitigate the domain gap by learning domain-invariant sensor representations. GenHAR aims to enhance the generalization capabilities of HAR on target domains purely with data from the source domain. The key novelty of GenHAR lies in two aspects. Firstly, GenHAR tokenizes sensor data and learns correlations among frequency sensor channel dimensions to improve the robustness of HAR models. Secondly, GenHAR improves the efficiency via selective masking and an efficient attention mechanism. We conduct a systematic analysis of GenHAR by comparing it with state-of-the-art HAR methods on real-world human activity datasets. Results show that GenHAR outperforms state-of-the-art methods by 9.97% in accuracy, and reduces Floating Point Operations by 6.4 times. Moreover, we deploy GenHAR at a leading logistics company in 4 cities, and have detected 2.15 billion real-time activities. We release our code at: https://github.com/Sensor-FoundationModel/GenHAR.

Problem

• Human Activity Recognition
• domain generalization
• distribution shift
• cross-domain
• sensor data

Innovation

• domain generalization
• sensor representation learning
• efficient attention
• frequency-channel correlation
• human activity recognition