LDA-FedHAR: Federated Human Activity Recognition for Wearable Devices Through Local HAR Data Alignment
Bibliographic record
Abstract
Wearable device-based Human Activity Recognition (HAR) has attracted considerable interest with the rapid development of the Internet of things (IoT), and Federated Learning (FL) has been widely adopted in this domain for its ability to collaboratively train models across decentralized devices while preserving privacy. However, its performance is hindered by data heterogeneity arising from variations in the placement of the wearable devices, user behaviors, and physiological characteristics. In this work, we present LDA-FedHAR, a federated HAR framework designed for wearable devices by capturing more common knowledge from aligned client HAR data. It performs Local HAR Data Alignment (LDA) on each client, which is an entirely on-device alignment method that operates independently on local HAR data. By computing the transformation matrix solely from local HAR data and applying it to the data itself, LDA projects heterogeneous client data into a unified space, thereby reducing inter-client discrepancies at the source. To further enhance efficiency and robustness, we propose two IMU-specific variants, LDA(S-IMU) and LDA(C-IMU), which explore intra-and inter-IMU correlations based on practical placements of wearable devices. Experiments are conducted on 4 public HAR datasets: HHAR, Shoaib2014, OPPORTUNITY++, and PAMAP2. The results show that LDA effectively reduces inter-client discrepancies, and LDA-FedHAR along with its variants consistently outperforms state-of-the-art FL methods. Moreover, the improvements achieved by integrating LDA into other FL methods highlight its applicability.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.002 | 0.001 |
| Bibliometrics | 0.001 | 0.002 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.003 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".