Few-Shot Learning for Cross-Domain Human Activity Recognition Using Wearable Sensors.

Cross-Modal Feature Adapter for Few-Shot Human Activity Recognition.

Recent years have witnessed outstanding success of deep learning in sensor-based human activity recognition (HAR), spanning a wide range of real-world applications like healthcare management, fitness tracking, and fall detection. However, sensor data annotation scarcity still remains a main challenge unresolved, which requires human annotators to take a long-term and tedious observation to segment and timestamp sensor samples meticulously, hampering the wide use of deep learning models, especially in few-shot HAR scenario. To handle such issue, this paper introduces a cross-modal data augmentation, by exploiting activity label text as key words to search for activity-related images to construct an augmented dataset. On this basis, a new cross-modal feature adapter is designed, which fine-tunes a pre-trained CLIP image encoder to effectively align image-sensor pairs. Through a learnable residual ratio, it may adaptively blend the old knowledge inherited from the original zero-shot CLIP with the new knowledge adapted from few-shot training samples, which makes training converges faster whilst forming a streamlined time series sensor encoder. Extensive experiments and ablation studies are performed on three public HAR benchmarks. The experimental results demonstrate that the proposed method outperforms existing state-of-the-art HAR baselines under all few-shot scenarios. A practical on-device inference latency is provided.

A Semi-Supervised CPC-Transformer Approach for Human Activity Recognition Under Label Scarcity

Human Activity Recognition (HAR) in smart-home environments plays a vital role in applications such as ambient assisted living, elderly care, and health monitoring. Unlike vision-based HAR, smart-home systems rely on ambient sensors such as motion sensors, that capture sparse, asynchronous, and noisy data, making accurate recognition more challenging. However, the limited availability of labeled sequences in real-world homes poses a critical obstacle to traditional supervised learning methods. To address this limitation, we introduce a semi-supervised approach that integrates Contrastive Predictive Coding (CPC) with a hybrid BiGRU-Transformer architecture. CPC is utilized as a self-supervised pretraining stage to learn informative temporal representations from unlabeled sequences, which are then used by the downstream classifier. These representations are subsequently processed in parallel by Bi-GRU and Transformer components to model short-term and long-term temporal dependencies, respectively, thereby enabling comprehensive temporal modeling for human activity recognition. Experimental evaluations on two real-world environmental sensor datasets, CASAS Aruba and CASAS Milan, demonstrate that the proposed model outperforms several baseline architectures in semi-supervised settings, achieving improvements of 5.31 percentage points ( $\approx 6.4$ % relative) on Aruba and 9.37 percentage points ( $\approx 14.8$ % relative) on Milan.

Fire: frequency-domain integration for robust enhancement in domain-generalized human activity recognition

MST-HDQ: Multi-Scale temporal hash discriminative quantization for wearable sensor sequence recognition

Sensor-based human activity recognition (HAR) from wearable devices is fundamental to applications such as mobile health monitoring, smart sports training, rehabilitation assessment, and context-aware computing. Wearable sensor sequence recognition requires representations that are temporally expressive, compact, and robust to class imbalance. Existing sensing systems typically treat temporal modeling and compact representation learning as separate design problems: temporal architectures often terminate in dense classifier logits, whereas hash-based objectives operate on features that have already been formed. This decoupled design limits the discriminability of learned activity codes, particularly for minority classes in imbalanced datasets. We propose MST-HDQ, a compact sequence-code learning framework that co-designs a sensor-aware large-kernel temporal encoder, wearable-sequence hierarchical attentive aggregation, a hash-discriminative quantization head, and a class-robust training objective. The temporal encoder combines multi-scale local branches with depthwise large-kernel temporal blocks to capture both short transient motions and longer activity phases, while the aggregation module fuses attentive temporal pooling with multi-depth feature summaries to select informative temporal moments across encoder depths. The quantization head learns compact prototype-discriminative sequence codes through scalar codebook discretization and normalized prototype classification. Experiments on OPPORTUNITY, PAMAP2, WISDM, and UniMiB-SHAR under the Hash-MMDC protocol show consistent macro-F1 improvements over reproduced baselines, with paired 50-seed Wilcoxon tests showing p<0.001\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p < 0.001$$\end{document}. Controlled comparisons with additional re-implemented baselines, grouped ablations, hash-space diagnostics, sensitivity studies, and stricter segment-grouped and subject-independent splits further support the robustness of the gains. MST-HDQ preserves compact 64-dimensional codes while requiring only 0.622M–0.813M trainable parameters, making it suitable for efficient wearable sensing deployments. The results demonstrate that jointly optimizing temporal evidence extraction and code-space geometry yields more class-complete hash structures than treating these objectives independently.

Human activity recognition using CNN–BiLSTM with attention on hip-mounted wearable sensors

Human activity recognition (HAR) is the identification of daily human activities using wearable sensor data. In this study, we evaluate a deep learning–based HAR framework utilizing hip-mounted accelerometer and gyroscope signals from the USC-HAD dataset, which contains readings from healthy participants only. The proposed pipeline integrates convolutional feature extraction, bidirectional long short-term memory modeling, and an additive attention mechanism to capture temporal dependencies in the sensor data. The model is evaluated using performance matrices and leave-one-subject-out cross-validation (LOSO-CV) to assess subject-independent generalization. Performance is reported using accuracy, precision, recall, F1-score, and 95% confidence intervals, and statistical significance testing. Our experimental results show that under subject-exclusive splitting, the proposed model achieves 98% accuracy. Under strict LOSO-CV, the model achieves a performance of 78% ± 0.1130, providing a more realistic assessment of subject-independent generalization across unseen individuals. The dataset does not include clinical or patient populations. The findings are limited to non-clinical settings and should be interpreted within this scope. The results primarily contribute methodological insights into wearable-based HAR systems. The potential of this work for healthcare applications is discussed as a direction for future research, subject to validation on clinically representative datasets.

Joint-Embedding Predictive Architecture for Sensor-based Activity Recognition

Sensor-based human activity recognition (HAR) has achieved significant progressed in fully supervised learning settings. However, these supervised learning models rely on large amount of labeled data, which require labor-intensive collection and meticulous annotation. To address these challenges, this paper proposes a Joint Embedding Predictive Architecture framework tailored for sensor-based HAR, designed to learn robust and generalizable representations from unlabeled datasets. The proposed framework features an encoder designed to explicitly model both the fine-grained local temporal representations within individual window and the long-term temporal sequence of adjacent windows. Furthermore, we introduce an improved Variance-Invariance-Covariance Regularization (VICReg) objective function that incorporates computationally lightweight norm term to stabilize the JEPA pre-training phase. This term balances variance, invariance and covariance constraints to prevent representation collapse. The proposed HAR-JEPA framework is evaluated using two benchmark continuously performed activity datasets. The results show that high-quality representations are successfully learned by the proposed framework. Furthermore, the representations learned by HAR-JEPA 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.