Automated dual-stream deep network design for activity recognition.

Structural mechanisms of cargo adaptors in membrane trafficking.

InGesture: An eight-class inertial sensor dataset for fluid intake and hand-gesture recognition.

CRISPR-based electrochemiluminescence biosensors: Principles, optimization strategies, and translational challenges – A review of recent progress

Microfluidic fabrication of peptide modified carrier-free self-assembled crizotinib-metal nanodrugs for NIR fluorescence imaging and dual-pathway therapy of non-small cell lung cancer.

Action-aware anchor-based frame selection strategy for action recognition

Understanding human actions from video streams remains a technically demanding endeavor, yet it is foundational to domains such as automated monitoring, behavioral analytics, and interactive system design. Consequently, research is steadily shifting toward automated frameworks capable of reliably interpreting video-based information. Prior studies have focused on separating spatial cues and temporal dependencies through diverse algorithmic strategies, often treating motion feature extraction as a standalone process outside the end-to-end learning pipeline. To resolve this constraint, a novel neural framework based on Gated Recurrent Units ( GRUs ) is developed to effectively integrate motion dynamics, spatial representations, and temporal evolution within a unified model. Skip connections are incorporated to counteract vanishing gradients, thereby ensuring smoother gradient flow and enhanced training stability. The proposed architecture is thoroughly evaluated on five established benchmark datasets, namely UCF101, HMDB51, Hollywood2, UCF50, and SS1, where it delivers state-of-the-art classification accuracies of 99.57%, 91.36%, 94.68%, 99.92%, and 80.37%, respectively. The experimental outcomes demonstrate the high precision, robustness, and versatility of the model across a variety of realistic application settings. Furthermore, we performed rigorous evaluations under various noisy conditions to assess the resilience of the proposed GRU architecture. The outcomes consistently demonstrated its strong resistance to noise and reinforced its reliability in challenging environments.

Fully integrated AI-enhanced flexible wearable sensor for real-time movement evaluation and table tennis training.

Unified adaptive open-set incremental learning for industrial human action recognition

DPBGFL: Debiased prototype bidirectional guided federated learning for human activity recognition

Role of prior in human activity recognition: A survey

Activity-dependent resolution adjustment for radar-based human activity recognition

Physical activity recognition based on electrocardiogram data only

ActivityCLIP: Enhancing group activity recognition by mining complementary information from text to supplement image modality

Real-time monitoring of tumor spatiotemporal dynamics and precise modulation of the tumor immune microenvironment are pivotal for effective cancer treatment. However, current platforms lack the ability to integrate quantitative molecular imaging with spatially confined immune modulation. Here, we report a nanotheranostic platform Pt2Te3:Ag2Te-αPD-L1 (PAPD) based on bandgap-engineered PtxTey:Ag2Te quantum dots. Platinum-induced electronic reconfiguration enables programmable bandgap modulation, yielding tunable near-infrared IIb emission spanning from 1550 to 2200 nm. Pt2Te3:Ag2Te quantum dots exhibit stable emission at 1730 nm and controllable mild photothermal effects. Conjugation with αPD-L1 antibodies endows the PAPD platform with dual tumor-targeting capability through EPR-mediated passive accumulation and PD-L1-specific active recognition. Crucially, near-infrared IIb molecular imaging enables quantitative monitoring of tumor targeting and therapeutic response, meanwhile mild photothermal modulation induces immune reprogramming that sensitizes tumors to immune checkpoint immunotherapy in vivo. This work establishes a generalizable quantum-engineered nanotheranostic platform, offering a scalable strategy for precision cancer treatment.

Dance motion recognition and correction present unique challenges due to the subtle distinctions in movement execution that differentiate correct performance from flawed attempts. This paper proposes an integrated system combining 3D motion capture technology with a novel spatiotemporal attention-based graph convolutional network for accurate dance action recognition and intelligent feedback generation. The proposed architecture features a dual-stream design incorporating adaptive graph topology learning that discovers task-relevant relationships between non-adjacent joints, alongside multi-scale temporal modeling to capture movement dynamics across varying time scales. A multi-dimensional correction feedback algorithm translates recognition outputs into prioritized, actionable guidance by comparing performer movements against professional reference templates through dynamic time warping alignment and joint-level deviation analysis. We constructed DanceMotion-86, a comprehensive dataset comprising 10,836 clips across 86 action categories spanning five dance genres. Experimental results demonstrate that the proposed method achieves 92.3% recognition accuracy, outperforming state-of-the-art baseline methods. User studies with 36 participants confirmed 87.4% error detection rate and showed significantly accelerated skill acquisition among feedback-enabled learners compared to control conditions. The system offers practical applications for intelligent dance instruction, cultural heritage preservation, and remote training platforms.

Among many issues, Robot vision experiences illumination challenges very frequently. Existing Human Action Recognition techniques perform excellently in state of the art, instead of scarce consideration on the vital issue of illumination. The illumination concern becomes highly sensitive when the Robot observes a medical-related action. The illumination severely affects the correct recognition of the action. Resultantly, misclassification of a medical action may lead to irreparable loss. To gauge the sensitivity of the concern, the current study proposes a deep learning-based model I2I (Illusion to Illumination). The model effectually identifies medical actions even in dark environments with sufficient accuracy. I2I model depth data has been selected from the NTU RGB+D dataset to judge the efficacy. The features are extracted from depth data using the Histogram of Depth (HoD) and provided to the I2I model to recognize actions. A threshold mechanism is applied to select depth data’s most prominent and valuable features. The efficacy and superiority of the I2I model are proven by comparing its performance with state-of-the-art research and provides 91.15% recognition accuracy.

Seven-step handwashing recognition based on multi-angle information fusion from dual millimeter-wave radars.

Motor imagery (MI)-based brain-computer interfaces (BCIs) enable users to control external devices using EEG signals, offering great potential in assistive and rehabilitation technologies. However, MI recognition remains challenging due to EEG's low signal-to-noise ratio (SNR), inter-subject variability, and complex spatiotemporal patterns. Existing approaches often suffer from limited accuracy, high computational cost, and poor interpretability. In response to these challenges, we present the first comprehensive benchmarking of the publicly available EEG-hand movement (EEG-HM) dataset. Our study aims to establish a standardized performance baseline, guide the selection of optimal models by jointly considering accuracy, prediction time, and explainability, and ultimately accelerate progress in MI-BCI development. We have proposed a two-stage optimization of machine learning models that employs both feature selection and hyperparameter tuning. We exploit five feature selection algorithms for selecting the best set of EEG electrodes and frequency bands, while Bayesian optimization is exploited for machine learning model optimization through hyperparameter tuning. Furthermore, to validate the neurophysiological basis of our model's decisions, we leverage explainable AI (XAI) algorithms-LIME and SHAP-quantifying the contributions of specific EEG electrodes and frequency bands to interpret its decision-making process. Through extensive simulations, the proposed two-stage optimization of the machine learning model demonstrates a superior performance in terms of accuracy, precision, and recall. This method outperforms the existing methods by 21.47% in accuracy with competitive prediction time. Its performance is further evaluated on the PhysioNet MI dataset, achieving a 4.67% accuracy improvement over state-of-the-art methods. Through LIME and SHAP, we provide the local and global explanations for no activity, left-hand, and right-hand imagery movements. Additionally, we analyze how various EEG frequency bands and electrode locations interact during the performance of different motor imagery hand movements.

ABSTRACT Population aging has increased the demand for intelligent smart home systems for elderly care. Although Internet of Things (IoT) sensors enable unobtrusive residential monitoring, existing human activity recognition methods often rely on centralized processing and have limited ability to model heterogeneous sensing sources and long‐range temporal dependencies. To address this issue, this paper proposes a Distributed Adaptive Multi‐Agent Transformer (DAMAT) framework for smart home activity recognition. DAMAT models heterogeneous sensing streams as collaborative agents, captures long‐range temporal and cross‐agent contextual dependencies through transformer‐based interaction, and employs adaptive coordination attention to regulate agent contributions under different activity contexts. Experiments on the CASAS and UCI HAR datasets show that DAMAT consistently outperforms representative deep learning baselines. In particular, the CASAS results directly support the effectiveness of the proposed framework in distributed smart home sensing environments, while the UCI HAR results provide auxiliary evidence that the temporal modeling and adaptive coordination mechanisms remain effective on wearable inertial sensor data.