PaperarXiv:2606.13194
HARENA: A Living Benchmark for Sensor-Based Human Activity Recognition on Resource-Constrained Edge Devices
arXiv preprint (cs), v1 2026-06-11, v2 2026-11-09; 68 pages
TL;DR
A living benchmark with public leaderboard: 16 task-specific neural HAR architectures plus kNN / Random Forest / SVM on tsfresh features, across 29 public datasets, 2 s and 3 s windows, subject-group (≤10-fold, LOSO when ≤10 subjects) splits, one shared training config. Deployment cost is measured on a Galaxy S24 Ultra, a Galaxy Watch5 Pro and a Nordic nRF5340 (Cortex-M33, 128 MHz, 512 KB RAM, 1 MB flash) with INT8 TFLite Micro + CMSIS-NN, including energy via Power Profiler Kit II.
- Form factor: multi
Abstract
Sensor-based human activity recognition runs on edge devices, trading performance against latency, memory, and energy. Yet new architectures are usually compared with few, often outdated baselines on few datasets with study-specific data preparation and hardware, leaving open which perform best at what cost. We present HARENA, a living benchmark with a public leaderboard, to which new architectures and datasets can be added under one protocol. Its initial suite evaluates 16 neural architectures and three classical baselines on 29 datasets, measuring performance on unseen subjects and deployment cost on a smartphone, smartwatch, and microcontroller. CNN-HAR, the suite's oldest deep architecture, leads overall, but 8 architectures lie within 2.5 points of its mean test macro-F1, while their cost differs by orders of magnitude. On the microcontroller, Random Forest leads on latency and peak RAM, LITEWAY-Light on flash, and TinyHAR is most accurate within the real-time budget. Rankings shift with dataset and channel count, so general claims require many datasets, which a growing benchmark can provide. We find no sign of benchmark overfitting, but performance drops by a median of 15.0 points on unseen subjects, about three times the typical spread between architectures, making generalization and adaptation the most promising future directions.
Verbatim from arXiv (opens in a new tab) · arXiv metadata (abstracts are released under CC0)
Results on ChatPPG leaderboards
All leaderboardsNo results from this paper on ChatPPG leaderboards yet. Browse the leaderboards.
Tasks & signals
Tasks
Signals
Keywords
How it works
Cite this paper
Burzer, M., King, T., Riedel, T., Beigl, M., & Röddiger, T. (2026). HARENA: A Living Benchmark for Sensor-Based Human Activity Recognition on Resource-Constrained Edge Devices. arXiv preprint , v1 , v2. https://arxiv.org/abs/2606.13194
Burzer, Maximilian, Tobias King, Till Riedel, Michael Beigl, and Tobias Röddiger. 2026. "HARENA: A Living Benchmark for Sensor-Based Human Activity Recognition on Resource-Constrained Edge Devices." arXiv preprint , v1 , v2. https://arxiv.org/abs/2606.13194.
Burzer, M., King, T., Riedel, T., Beigl, M. and Röddiger, T. (2026) 'HARENA: A Living Benchmark for Sensor-Based Human Activity Recognition on Resource-Constrained Edge Devices', arXiv preprint , v1 , v2. Available at: https://arxiv.org/abs/2606.13194
Burzer, Maximilian, et al. "HARENA: A Living Benchmark for Sensor-Based Human Activity Recognition on Resource-Constrained Edge Devices." arXiv preprint , v1 , v2, 2026. https://arxiv.org/abs/2606.13194.
M. Burzer, T. King, T. Riedel, M. Beigl, and T. Röddiger, "HARENA: A Living Benchmark for Sensor-Based Human Activity Recognition on Resource-Constrained Edge Devices," arXiv preprint , v1 , v2, 2026, [Online]. Available: https://arxiv.org/abs/2606.13194.
@misc{burzer2026harena,
title = {{HARENA: A Living Benchmark for Sensor-Based Human Activity Recognition on Resource-Constrained Edge Devices}},
author = {Burzer, Maximilian and King, Tobias and Riedel, Till and Beigl, Michael and Röddiger, Tobias},
year = {2026},
howpublished = {arXiv preprint , v1 , v2},
eprint = {2606.13194},
archivePrefix = {arXiv}
}Added to the ChatPPG research hub
