Aug 2026· International Journal of Data Science and Analysis· Vol 22· 0 citations· 61 references
TL;DR
A novel frequency-domain feature learning framework named FIRE, which enhances domain generalization to improve HAR performance and validating its robustness against domain shifts is proposed.
A novel lightweight cross-domain few-shot sensor-based HAR network (CFSH-Net) is proposed for cross-domain activity recognition with limited labeled samples, which demonstrates strong cross-user generalization on PAMAP2 and USC-HAD, and stable cross-dataset transfer when trained on OPPORTUNITY and evaluated on four other datasets.
Hao Zheng, Hongji Xu, Fei Gao et al.· IEEE journal of biomedical a...· 0 citations
The proposed Joint Embedding Predictive Architecture framework designed to learn robust and generalizable representations from unlabeled datasets demonstrates superior generalization on minority, high variance transitional activities such as sit-to-stand and sit-to-lie where supervised learning tend to overfit due to limited support.
Mohd Halim Mohd Noor, AbdulRahman M. A. Baraka· arXiv.org· 0 citations
WiFi-based human activity recognition (HAR) models often perform well in a single environment but experience significant accuracy drops in new environments due to variations in spatial settings, human movement, and physical factors. This study introduces an unsupervised domain adaptation (UDA) method, feature alignment-based domain adversarial neural network (FA-DANN), to address these challenges. FA-DANN integrates feature alignment with domain adversarial neural networks (DANN) to improve cross-environment HAR performance. The method is evaluated on three public WiFi datasets—GJWiFi, OPERAnet, and SHARP-2—using 5-fold cross-validation. Without domain adaptation, the baseline CNN-ABiLSTM model achieves an average F1-score of only 15.65%, whereas FA-DANN improves this average F1-score to 92.67%, demonstrating substantial gains. In cross-domain evaluations, leave-one-subject-out cross-validation (LOSOCV) assesses generalization to unseen subjects and environments. FA-DANN outperforms existing DANN-based and Gaussian feature alignment methods, achieving a 19.98% F1-score improvement over state-of-the-art models. Ablation studies further analyze the impact of the decoder and domain classifier components on adaptation. By explicitly aligning target-domain features with source-domain distributions, FA-DANN enables WiFi-based HAR models to generalize across new environments without requiring labeled data, offering a scalable, cost-effective solution for real-world deployment.
Amany Elkelany, Robert Ross, Susan Mckeever· Journal of Ambient Intellige...· 0 citations
Human activity recognition (HAR) in smart environments plays a critical role in applications such as healthcare monitoring, intelligent transportation systems, and ambient assisted living; however, existing approaches are limited by their inability to effectively handle heterogeneous multimodal sensor data, capture long-range temporal dependencies, and generalize across diverse real-world environments under domain shifts. In this work, we present DMART-HAR, a Dynamic Multimodal Activity Recognition Transformer framework that unifies structured multimodal representation learning, transformer-based temporal modeling, cross-modal interaction, and adversarial domain adaptation within a single architecture. Specifically, the developed method introduces a sensor tokenization mechanism to encode heterogeneous IoT data into a unified representation space, followed by a transformer encoder to capture global contextual dependencies, while a cross-modal attention module enables deep interaction among sensor modalities and an adversarial domain adaptation strategy enhances robustness to unseen environments. Extensive experiments on benchmark datasets, including CASAS, PAMAP2, and Opportunity, demonstrate that DMART-HAR consistently outperforms both conventional baselines and recent state-of-the-art methods, achieving accuracy/F1-scores of 94.3%/92.8%, 96.2%/94.7%, and 89.8%/88.1%, respectively, and consistently outperforms the strongest competing approaches under cross-domain evaluation settings. These findings demonstrate the effectiveness of modeling temporal dynamics, multimodal relationships, and domain invariance simultaneously, establishing DMART-HAR as a scalable and robust solution for real-world HAR applications.
Self-Supervised Learning (SSL) has emerged as an effective paradigm for reducing the dependence of Human Activity Recognition (HAR) models on labeled data. To address the inadequate exploitation of IMU spatio-temporal correlations during pre-training and the limited generalization caused by simplistic fine-tuning strategies, a novel SSL framework for IMU-based HAR is proposed. The framework employs the Transformer and Depthwise Separable Convolution (DSC) to jointly capture global temporal dependencies and local spatial features, which are adaptively fused into discriminative spatio-temporal representations. These representations are subsequently enhanced through spatio-temporal feature extraction and multi-dimensional feature aggregation for downstream HAR. Furthermore, an IMU-based data acquisition platform was developed to construct the CQXY dataset. The proposed method was validated through comprehensive evaluations on four public datasets (UCI, Motion, HHAR, and Shoaib) and one self-collected dataset (CQXY). Experimental results show that, on the public datasets, the proposed method improves classification accuracy, F1-score, and Cohen’s kappa coefficient by an average of 13.11%, 14.24%, and 16.70%, respectively, compared with the baseline models. Similarly, on the self-collected dataset, the corresponding improvements reach 8.87%, 11.07%, and 10.81%. These results confirm the generalization of the proposed approach across datasets of different scales and domain.
Background: Distribution shift is one of the most pervasive data-related challenges for time-series classification. In sensor-based Human Activity Recognition (HAR), a model trained on one population of subjects often degrades substantially when deployed on another, even with identical hardware and activity labels. This paper addresses that setting specifically: cross-subject distribution shift. Methods: We propose FTDA (Frequency–Time Domain Alignment), an unsupervised domain-adaptation method that processes raw signals and their FFT-magnitude spectra through a dual-branch encoder, combines a Gradient-Reversal adversarial loss on the joint feature with a Multi-Kernel MMD on the frequency branch, and applies a symmetric-KL time–frequency consistency loss on unlabeled target data. Results: On UCI HAR, over four cross-subject transfer tasks and 10 random seeds (n=40 paired observations), FTDA reaches 95.01% target accuracy against 92.46% for Source-only, 92.29% for DANN and 93.32% for a RAINCOAT-style baseline. Because the tasks share subjects and two of them are mirror configurations, no pooled significance test is offered; inference is confined to within-task paired tests (10 seeds each, Holm-corrected), under which FTDA is ahead of every single-view baseline on all four tasks, while the difference from the two dual-view competitors (+1.69 pp against RAINCOAT-style, descriptive task-level interval [−0.40,+3.78]) is not established. Over a common set of seeds, controls widened to FTDA’s parameter and multiply-accumulate budgets shifted accuracy by at most +0.35 pp against DANN (intervals including zero), where FTDA gains +2.97 pp on the same runs, so added capacity alone does not appear to explain the improvement, and a target-only reference model, which is optimistic because it is evaluated on the windows it was trained on, sits 0.70 pp above FTDA. An ablation shows that the frequency branch is useful chiefly because it enables a coupled mechanism of cross-view consistency and auxiliary source supervision (+2.02 pp for that final step) rather than concatenated features (+1.53 pp); for the frequency MMD term, no measurable benefit was detected (−0.57 pp, 95% CI [−1.28,+0.13]), and it can be removed. On a second dataset, HHAR, the cross-subject advantage does not reproduce under cross-device shift: we detect no difference from time-only adversarial baselines (paired difference +0.48 pp against DANN, 95% CI [−1.20,+2.16]; exploratory, since all device pairs share the same subjects), so the benefit of the spectral view does not appear to carry across shift types. Conclusions: Under cross-subject shift, the operative ingredient is not the spectral view as an extra set of features but the coupled consistency-and-auxiliary mechanism that the second view makes possible; the alignment objectives themselves contribute little, and no measurable benefit was detected for the frequency MMD term. FTDA realizes this view at a modest 0.31 M/7.55 MMAC footprint with no OT solver in the training loop and no hyper-parameter that requires target labels. Transfer across devices and operating environments is not evaluated here and remains open for future work.
