Noisy Conditions Research Articles (Page 1)

Quantum communication has achieved significant performance gains compared to classical systems but remains sensitive to channel noise and decoherence. These limitations become especially critical in quantum image transmission, where high-dimensional visual data must be preserved with both structural fidelity and robustness. In this context, transform-based quantum encoding methods have emerged as promising approaches, yet their relative performance under noisy conditions has not been fully explored. This paper presents a comparative study of two such methods, the quantum Fourier transform (QFT) and the quantum Haar wavelet transform (QHWT), within an image transmission framework. The process begins with source coding (JPEG/HEIF), followed by channel coding to enhance error resilience. The bitstreams are then mapped into quantum states using variable qubit encoding and transformed using either QFT or QHWT prior to transmission over noisy quantum channels. At the receiver, the corresponding decoding operations are applied to reconstruct the images. Simulation results demonstrate that the QFT achieves superior performance under noisy conditions, consistently delivering higher Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), and Universal Quality Index (UQI) values across different qubit sizes and image formats compared to the QHWT. This advantage arises because QFT uniformly spreads information across all basis states, making it more resilient to noise. By contrast, QHWT generates localized coefficients that capture structural details effectively but become highly vulnerable when dominant coefficients are corrupted. Consequently, while QHWT emphasizes structural fidelity, QFT provides superior robustness, underscoring a fundamental trade-off in quantum image communication.

Drug repositioning accelerates therapeutic discovery, but existing computational methods are hampered by representation collapse, noisy supervision, and suboptimal negative sampling. To address these limitations, we introduce MGTAL-DR, a novel graph learning framework that integrates a dual-channel transformer architecture with adversarial contrastive learning and a purely negative sampling strategy. Its parallel graph encoders capture both multiscale similarity patterns and heterogeneous biological semantics. In one channel, structural neighborhoods are expanded via diffusion-based propagation to encode varying granularities of similarity. In the other, cross-entity relationships are contextualized through meta-path-guided attention over a unified drug-disease-protein graph. Adversarial perturbations enhance latent space robustness to prevent noise-induced collapse, while a low-rank decomposition strategy isolates informative hard negatives by disentangling global association trends from local residual signals. Together, these components sharpen the decision boundary under sparse and noisy conditions. Extensive experiments demonstrate that MGTAL-DR achieves state-of-the-art performance across three benchmark data sets. Furthermore, a case study on Alzheimer's disease highlights its practical utility by successfully identifying promising therapeutic candidates, thereby validating its real-world potential.

Ghost imaging, which relies on single-pixel detection, offers advantages in spectral flexibility and sensitivity compared to conventional imaging systems. However, its performance is inherently limited by noise and the characteristic square-root dependence of image quality on the number of measurements. This study investigates how compressive sensing methods—specifically L 1 and L 2 norm minimization—alter this dependence under noisy conditions, transforming it into a non-monotonic relationship. We analyze four illumination patterns (non-shifted random binary, subpixel-shifted, displacement, and color-noise patterns) in numerical simulation and an experimental setup, comparing their reconstruction via compressive sensing and second-order correlation methods. Our results provide practical guidelines for pattern and algorithm selection. These insights establish a framework for optimizing ghost imaging systems in noisy environments, including speckle noise from liquid crystal spatial light modulators, dynamic scattering, and source instabilities. The observed non-monotonic quality dependence underscores the need for careful measurement-number optimization in ghost imaging systems with compressive sensing.

This study introduces Itô-RMSProp, a novel extension of the RMSProp optimizer inspired by Itô stochastic calculus, which integrates adaptive Gaussian noise into the update rule to enhance exploration and mitigate overfitting during training. We embed this optimizer within Gated Recurrent Unit (GRU) networks for stock price forecasting, leveraging the GRU’s strength in modeling long-range temporal dependencies under nonstationary and noisy conditions. Extensive experiments on real-world financial datasets, including a detailed sensitivity analysis over a wide range of noise scaling parameters (ε), reveal that Itô-RMSProp-GRU consistently achieves superior convergence stability and predictive accuracy compared to classical RMSProp. Notably, the optimizer demonstrates remarkable robustness across all tested configurations, maintaining stable performance even under volatile market dynamics. These findings suggest that the synergy between stochastic differential equation frameworks and gated architectures provides a powerful paradigm for financial time series modeling. The paper also presents theoretical justifications and implementation details to facilitate reproducibility and future extensions.

Structural dynamics analysis is essential for predicting the behavior of engineering systems under dynamic forces. This study presents a hybrid framework that combines analytical modeling, machine learning, and optimization techniques to enhance the accuracy and efficiency of dynamic response predictions for Single-Degree-of-Freedom (SDOF) systems subjected to harmonic excitation. Utilizing a classical spring–mass–damper model, Fourier decomposition is applied to derive transient and steady-state responses, highlighting the effects of damping, resonance, and excitation frequency. To overcome the uncertainties and limitations of traditional models, Extended Kalman Filters (EKFs) and Physics-Informed Neural Networks (PINNs) are incorporated, enabling precise parameter estimation even with sparse and noisy measurements. This paper uses Adam followed by LBFGS to improve accuracy while limiting runtime. Numerical experiments using 1000 time samples with a 0.01 s sampling interval demonstrate that the proposed PINN model achieves a displacement MSE of 0.0328, while the Eurocode 8 response-spectrum estimation yields 0.047, illustrating improved predictive performance under noisy conditions and biased initial guesses. Although the present study focuses on a linear SDOF system under harmonic excitation, it establishes a conceptual foundation for adaptive dynamic modeling that can be extended to performance-based seismic design and to future calibration of Eurocode 8. The harmonic framework isolates the fundamental mechanisms of amplitude modulation and damping adaptation, providing a controlled environment for validating the proposed PINN–EKF approach before its application to transient seismic inputs. Controlled-variable analyses further demonstrate that key dynamic parameters can be estimated with relative errors below 1%—specifically 0.985% for damping, 0.391% for excitation amplitude, and 0.692% for excitation frequency—highlighting suitability for real-time diagnostics, vibration-sensitive infrastructure, and data-driven design optimization. This research deepens our understanding of vibratory behavior and supports future developments in smart monitoring, adaptive control, resilient design, and structural code modernization.

Interpretable epidemic state estimation via rule based modeling.

Evaluation of the Automatic Classifier AutoSense Sky OS for pediatric cochlear implant users using a virtual classroom environment.

Dynamic weighted multimodal fusion for fault diagnosis of marine rotating machinery under noisy and low-sample conditions

A novel multi-scale quadratic convolutional network for bearing fault diagnosis: Handling noisy conditions

Abstract Purpose In this study, a methodology for damage detection and localization in aeronautical structures based on Automatic Operational Modal Analysis and a strain-based damage index is presented. Methods The proposed approach enables the automatic extraction of strain mode shapes under operational conditions, relying solely on output measurements. Modal parameters are estimated using Stochastic Subspace Identification, with pole selection performed via Density-Based Spatial Clustering. The methodology is initially validated numerically on two different case studies: a stiffened cantilevered plate and a composite glider. In both cases, damage is simulated by locally reducing the stiffness of specific regions, and strain signals are collected from virtual sensors under spatially and temporally random excitation. The proposed approach is then applied to an experimental test on a manufactured composite glider model, instrumented with Fiber Bragg Grating sensors bonded to the wing surface. Results and Conclusion The numerical results demonstrate that the proposed methodology effectively detects and localizes damage across varying intensities, even under noisy conditions. Experimental results confirm the effectiveness of the methodology in real-world conditions, highlighting its potential for in-flight Structural Health Monitoring applications.

Advanced structural health monitoring (SHM) systems have emerged as a critical requirement for ensuring structural integrity and operational safety. However, only limited SHM data are utilized in traditional machine learning approaches for structural damage detection tasks, while complementary multisource data might be ignored. This study presents a bridge damage detection framework that integrates multi-attention mechanisms and cross-modal feature learning. The proposed method can address the shortcomings of traditional approaches in multisource heterogeneous data fusion and feature extraction. The framework enhanced feature extraction from multisource vibration signals and suppression of redundant information, which benefits from the parallel channel-spatial attention and residual compression excitation mechanisms. Furthermore, both global and local features were simultaneously captured from the fused multichannel time–frequency representations based on a multi-scale feature fusion strategy. The effectiveness of the proposed method was validated through the datasets from a laboratory-scale cable-stayed bridge under moving vehicle loading. The numerical results demonstrate the superior performance and robustness of the proposed method compared to traditional methods in identifying various damage patterns under noisy conditions.

In recent years, deep learning has been increasingly applied in the field of fault diagnosis, but it currently faces two challenges: (1) data privacy issues prevent the aggregation of data from different users to form a large training dataset; (2) the limited memory of edge devices or handheld detection devices restricts the application of some larger structural models. To address these issues, this article proposes a lightweight federated learning method with transformer network for intelligent fault diagnosis. A federated learning architecture is constructed to achieve distributed learning of different user data, which not only ensures the privacy and security of user data, but also enables feature learning of different user data. In addition, the lightweight transformer network is built locally for different users to achieve the applicability of the model on different devices. An experimental case was implemented to demonstrate the effectiveness of the proposed method, and the results showed that the proposed method can achieve effective fault diagnosis while preserving data privacy. Compared with other methods, the proposed diagnostic model requires less computing resources. In addition, even under noisy conditions, the method maintains significant robustness against acoustic interference.

Introduction: Deep learning-driven acoustic feature recognition significantly advances non-invasive disease screening. However, existing pneumonia detection methods based on cough sounds remain limited in terms of accuracy and robustness in complex acoustic environments. To address these challenges, we propose the Heterogeneous Cross-modal Topology Enhanced Network (HCTEN), which incorporates advanced attention mechanisms and convolutional strategies to capture diagnostically relevant, multi-scale acoustic events effectively. Methods: To address these limitations, we propose the Heterogeneous Cross-modal Topology Enhanced Network (HCTEN), which is introduced in the context of cough-based pneumonia classification. HCTEN integrates a Channel-Spatial Joint Attention Module (CSJAM) to emphasize diagnostic frequency bands and a Step Multi-scale Topological Convolution (SMTConv) to capture rich temporal-spectral representations. Results: Experiments conducted on 2,039 high-quality cough samples (cough score > 0.9) from the COUGHVID dataset demonstrate that HCTEN achieves 96.09% accuracy, outperforming state-of-the-art models with an average gain of 1.75% in accuracy, 5.42% in AUROC, 7.21% in F1-score, and 9.1% in sensitivity. Ablation studies further validate the contribution of each module. Discussion: Experimental results confirm that HCTEN benefits from the integration of CSJAM and SMT-Conv, which enhances the model’s ability to detect weak pathological acoustic features. These components significantly improve robustness under complex and noisy conditions, making the model suitable for real-world pneumonia screening scenarios. Conclusion: HCTEN provides a substantial advancement in pneumonia detection via acoustic cough analysis, reliably identifying pathological acoustic biomarkers. This framework holds significant potential for practical deployment in real-world intelligent diagnostic applications.

ABSTRACT Rolling element bearings are crucial in aero-engines, and accurate fault diagnostics can avert major aircraft accidents by enabling timely detection and maintenance. Machine learning (ML) excels in fault diagnosis but requires extensive time for model architecture tuning and hyperparameter selection, challenging operational staff and hindering implementation. Recently, Automated Machine Learning (AutoML) has emerged to tackle hyperparameter optimization in rotating machinery diagnostics. Traditional AutoML techniques, such as Bayesian optimization, grid search, and random search, automate model selection and tuning. Yet, they are underutilized in aero-engine bearing fault diagnosis, where research remains limited. To boost automation, this study proposes a meta-learning-driven AutoML framework. By incorporating health indicators, which reflecting rotating machinery fault states as meta-features, the framework precisely recommends optimal architectures and hyperparameters, expediting intelligent diagnostic system deployment. Experiments on aero-engine bearing faults show this method outperforms other ML approaches in classifying health indicators, time-domain waveforms, and frequency-domain vibration spectra. It exhibits remarkable robustness in noisy conditions and small-sample scenarios.

Emotion recognition from EEG signals has emerged as a pivotal area of research, driven by its transformative potential in healthcare, brain-computer interfaces, and affective computing systems. However, the intrinsic complexity, non-linearity, and susceptibility to noise in EEG data present significant challenges to accurate emotional state classification. This study proposes a robust and interpretable hybrid deep learning model for EEG-based emotion recognition. The architecture integrates Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) networks, and attention mechanisms, together with advanced signal processing techniques such as Continuous Wavelet Transform (CWT) and Power Spectral Density (PSD). This integrated approach facilitates the extraction of comprehensive spatial, temporal, and spectral features from EEG signals, enhancing the model’s ability to capture intricate patterns associated with emotional states. Experimental evaluations on the SEED-IV dataset, encompassing four emotional categories—Neutral, Happy, Sad, and Fear—demonstrated the model’s exceptional performance, achieving a macro-average F1-score of 93% and an area under the ROC curve (AUC) of 0.94. These results validate the model’s effectiveness in accurately distinguishing complex emotional patterns, even under noisy conditions and inter-class ambiguities. Overall, this research advances the domain of EEG-based emotion recognition by introducing a high-performing, interpretable framework suitable for real-world applications while laying the foundation for future developments in adaptive neurofeedback systems and emotion-aware brain-computer interfaces.

Abstract Rolling bearings, as one of the most vital components in rotating machinery, are frequently exposed to severe noise interference during operation, posing a significant challenge for accurate and rapid fault identification. To address this issue, this study proposes a novel fault diagnosis framework termed TKAN (Transformer-Kolmogorov Arnold Networks), which integrates the global feature extraction capability of the Transformer with the nonlinear noise suppression advantage of the KAN linear layer. In the proposed TKAN model, raw vibration signals are first segmented into structured samples to fully preserve temporal dynamics. A four-layer Transformer module is then employed to extract high-dimensional representations from the input data, leveraging multi-head self-attention to enhance discriminative feature learning across different subspaces. To improve robustness under noisy conditions, a KAN linear layer with B-spline activation is incorporated in place of traditional linear mappings, effectively smoothing the feature space and attenuating noise-induced fluctuations. Extensive experiments are conducted on two widely used benchmark datasets-CWRU and XJTU-to evaluate the performance of TKAN in both clean and noisy environments. Comparative results against five representative deep learning models (MLP, CNN, KAN, LSTM-KAN, and CNN-KAN) demonstrate that TKAN achieves superior performance across multiple evaluation metrics (Accuracy, Precision, Recall and F1-Score). Furthermore, under various levels of Gaussian, Uniform, and Impulse noise, TKAN consistently maintains high classification accuracy, underscoring its strong noise resilience and diagnostic robustness. This study provides a novel approach for fault diagnosis of bearings in noisy environments, offering significant practical and research value.

In surgical navigation-assisted pedicle screw fixation, cross-source pre- and intra-operative point clouds registration faces challenges like significant initial pose differences and low overlapping ratio. Classical algorithms based on feature descriptor have high computational complexity and are less robust to noise, leading to a decrease in accuracy and navigation performance. To address these problems, this paper proposes a coarse-to-fine registration framework. In the coarse registration stage, a Point Matching algorithm based on Curvature Feature Learning (CFL-PM) is proposed. Through CFL-PM and Farthest Point Sampling (FPS), the coarse registration of overlapping regions between the two point clouds is achieved. In the fine registration stage, the Iterative Closest Point (ICP) is used for further optimization. The proposed method effectively addresses the challenges of noise, initial pose and low overlapping ratio. In noise-free point cloud registration experiments, the average rotation and translation errors reached 0.34° and 0.27 mm. Under noisy conditions, the average rotation error of the coarse registration is 7.28°, and the average translation error is 9.08 mm. Experiments on pre- and intra-operative point cloud datasets demonstrate the proposed algorithm outperforms the compared algorithms in registration accuracy, speed, and robustness. Therefore, the proposed method can achieve the precise alignment of the surgical navigation-assisted pedicle screw fixation.

The transition toward Industry 4.0 and the emerging concept of Industry 5.0 demand intelligent tools that integrate efficiency, adaptability, and human-centered design. This paper presents a Computer Vision-based framework for automated motion classification in Methods-Time Measurement 2 (MTM-2), with the aim of supporting industrial time studies and ergonomic risk assessment. The system uses a Convolutional Neural Network (CNN) for pose estimation and derives angular kinematic features of key joints to characterize upper limb movements. A two-stage experimental design was conducted: first, three lightweight classifiers—K-Nearest Neighbors (KNN), Support Vector Machines (SVMs), and a Shallow Neural Network (SNN)—were compared, with KNN demonstrating the best trade-off between accuracy and efficiency; second, KNN was tested under noisy conditions to assess robustness. The results show near-perfect accuracy (≈100%) on 8919 motion instances, with an average inference time of 1 microsecond per sample, reducing the analysis time compared to manual transcription. Beyond efficiency, the framework addresses ergonomic risks such as wrist hyperextension, offering a scalable and cost-effective solution for Small and Medium-sized Enterprises. It also facilitates integration with Manufacturing Execution Systems and Digital Twins, and is therefore aligned with Industry 5.0 goals.

SUMMARY The precise picking of first arrivals in seismology is very important for characterizing and monitoring the earthquakes. Similarly, in exploration seismology, identifying the refraction phases is crucial for building accurate velocity models. However, it's identification poses significant challenges in the presence of noise, especially for smaller magnitude earthquakes. Here, we propose a method to identify the first arrival P-wave from local earthquake data, employing time-frequency mapping of raw seismograms using the Generalized S-Transform and then extracting optimal deep encoded features utilizing the convolutional neural network-based unsupervised deep learning approach without the need for labelling the data. The statistical and transformational metrics, generated from both the deep encoded features and the original waveform, are combined to create an enriched feature space. Quantum clustering is then applied to this combined feature space to identify patterns or clusters that distinguish useful waveform sections from noise. This waveform-level selective identification and segmentation facilitate the determination of first arrival times within the relevant sections. The effectiveness of this method is first validated on a suite of synthetic data contaminated with various level and types of noise, and then applied to the observed data from the STEAD global dataset and seismic stations from the Jammu and Kashmir Himalaya. The method demonstrates stable picking performance under noisy conditions when compared to STA/LTA, AIC Picker and the unsupervised deep learning with classic K-Means. It also shows a broadly similar trend to supervised models such as PhaseNet and EQTransformer, and is computationally efficient, even in low signal-to-noise ratio conditions.

Unexpected failures in rotating machinery can cause costly downtime and safety hazards in industrial systems, highlighting the need for accurate and robust fault diagnosis. However, fault-related signals are often weak and easily obscured by noise, making reliable diagnosis challenging in real-world environments. To address this, we propose Time-frequency and time-series dual-branch fusion network(TFDFNet), a novel dual-branch deep learning model designed to improve fault classification performance under noisy and complex conditions. The model combines two complementary types of information: time-frequency representations derived from continuous wavelet transform and raw time-sequence data captured through sliding-window sampling. A Swin Transformer is used to extract deep features from time-frequency images, while a specially designed module called Gated attention block(GABlock) learns key temporal patterns from the sequence data. These features are fused using a cross-attention mechanism to enhance fault-related information. Extensive experiments on two public bearing fault datasets (CWRU and Ottawa) show that TFDFNet achieves outstanding accuracy, even under severe noise interference. The model reaches up to 100% accuracy on CWRU and 99.44% on Ottawa, and consistently outperforms existing convolutional neural network (CNN) baselines. These results demonstrate the practical potential and robustness of TFDFNet for intelligent fault diagnosis in industrial applications.