Noise Robustness Research Articles

Noise robust damage detection of laminated composites using multichannel wavelet-enhanced deep learning model

This paper presents a noise-robust damage detection framework for composite structures via a commonly used vibration-based non-destructive testing (NDT) method. Recently, deep learning-based models have shown promising performance in the autonomous damage detection of laminated composites; however, the poor noise robustness of these models has plagued data-driven damage detection. Moreover, none of the existing studies on damage detection in laminated composites focus on noise-robust deep learning models with high generalization ability. Therefore, this study proposes a hybrid deep learning framework called a multi-channel convolutional autoencoder-support vector machine (MC-CAE-SVM) based on empirical mode decomposition (EMD) and correlation analysis for noise-robust damage detection. This framework aims to first decompose the vibrational signal from multiple health states into intrinsic mode functions (IMFs). Secondly, highly correlated IMFs were extracted using correlation analysis to remove noisy IMFs. Finally, these IMFs were transformed into a time-frequency representation using continuous wavelet transform (CWT) and input to the MC-CAE-SVM model for feature learning and damage detection. Additionally, the accuracy and sensitivity of the model to damage are enhanced by optimizing the MC-CAE-SVM model hyperparameters. Moreover, anti-noise analysis is performed to check the noise-robustness of the proposed model by incorporating noise at various levels. The results showed that the proposed model can provide better damage detection performance compared to conventional models with excellent noise robustness.

A Static Sign Language Recognition Method Enhanced with Self-Attention Mechanisms

For the current wearable devices in the application of cross-diversified user groups, it is common to face the technical difficulties of static sign language recognition accuracy attenuation, weak anti-noise ability, and insufficient system robustness due to the differences in the use of users. This paper proposes a novel static sign language recognition method enhanced by a self-attention mechanism. The key features of sign language gesture classification are highlighted by the weight function, and then the self-attention mechanism is combined to pay more attention to the key features, and the convolutional neural network is used to extract the features and classify them, which realizes the accurate recognition of different types of static sign language under standard gestures and non-standard gestures. Experimental results reveal that the proposed method achieves an average accuracy of 99.52% in the standard static sign language recognition task when tested against the standard 36 static gestures selected within the reference American Sign Language dataset. By imposing random angular bias conditions of ±(0°–9°] and ±(9°–18°], the average recognition rates in this range were 98.63% and 86.33%. These findings indicate that, compared to existing methods, the proposed method not only maintains a high recognition rate for standard static gestures but also exhibits superior noise resistance and robustness, rendering it suitable for static sign language recognition among diverse user populations.

Bearing fault diagnosis by sparse frequency spiral spectrum driven NAF-LDM under strong noise and small samples

Abstract The complexity of background noise and the scarcity of real fault samples seriously affect the diagnostic accuracy of the model. To address this, a noise-robust two-dimensional feature map, the sparse frequency spiral spectrum (SFSM), based on sparse representation theory, is proposed. A bridge penalty coefficient is applied to the sparse representation model to accurately select impact components, and the fast iterative shrinkage threshold algorithm is used to solve for sparse representation coefficients. Sparse reconstructed signals are obtained by convolving the impact patterns with these coefficients, leading to a sparse reconstruction algorithm with reduced computational complexity. Furthermore, the novel non-linear activation-free blocks (NAF Blocks) are embedded into the latent diffusion model to augment small samples, significantly improving image generation speed and quality. The integration of the Swin transformer for feature extraction and classification further enhances diagnostic performance. The superiority of this method is validated on the XJTU-SY dataset, a bearing experimental platform dataset, and enterprise engineering dataset. Experimental results demonstrate that the structural and generalization advantages of NAF Blocks are crucial for improving image quality and inference speed. The noise suppression capability of the proposed method, facilitated by the SFSM feature processing technique, is confirmed through ablation and noise robustness tests. Finally, the Swin transformer’s excellent feature extraction and classification capabilities for SFSM are verified. The proposed method achieves diagnostic accuracies of 99.10% and 98.7% on the XJTU-SY and experimental platform datasets, respectively.

Robust Real-Time Cancer Tracking via Dual-Panel X-Ray Images for Precision Radiotherapy

Respiratory-induced tumor motion presents a critical challenge in lung cancer radiotherapy, potentially impacting treatment precision and efficacy. This study introduces an innovative, deep learning-based approach for real-time, markerless lung tumor tracking utilizing orthogonal X-ray projection images. It incorporates three key components: (1) a sophisticated data augmentation technique combining a hybrid deformable model with 3D thin-plate spline transformation, (2) a state-of-the-art Transformer-based segmentation network for precise tumor boundary delineation, and (3) a CNN regression network for accurate 3D tumor position estimation. We rigorously evaluated this approach using both patient data from The Cancer Imaging Archive and dynamic thorax phantom data, assessing performance across various noise levels and comparing it with current leading algorithms. For TCIA patient data, the average DSC and HD95 values were 0.9789 and 1.8423 mm, respectively, with an average centroid localization deviation of 0.5441 mm. On CIRS phantoms, DSCs were 0.9671 (large tumor) and 0.9438 (small tumor) with corresponding HD95 values of 1.8178 mm and 1.9679 mm. The 3D centroid localization accuracy was consistently below 0.33 mm. The processing time averaged 90 ms/frame. Even under high noise conditions (S2 = 25), errors for all data remained within 1 mm with tracking success rates mostly at 100%. In conclusion, the proposed markerless tracking method demonstrates superior accuracy, noise robustness, and real-time performance for lung tumor localization during radiotherapy. Its potential to enhance treatment precision, especially for small tumors, represents a significant step toward improving radiotherapy efficacy and personalizing cancer treatment.

Audiovisual Speech Recognition Method Based on Connectionism

Audio-visual speech recognition technology has greatly improved the performance of pure speech recognition by combining visual speech information and acoustic speech information, but there are problems such as large data demand, audio and video data alignment, and noise robustness. Scholars have proposed many solutions to these problems. Among them, deep learning algorithms, as representatives of connectionist artificial intelligence technology, have good generalization ability and portability, and are easier to migrate to different tasks and fields. They are becoming one of the mainstream technologies for audio-visual speech recognition. This paper mainly studies and analyzes the application of deep learning technology in the field of audio-visual speech recognition, especially the audio-visual speech recognition model of the end-to-end framework. Through experimental comparative analysis, relevant data sets and evaluation methods are summarized, and finally hot issues that need to be further studied and solved are proposed.

Fast Hyperspectral Image Classification with Strong Noise Robustness Based on Minimum Noise Fraction

A fast hyperspectral image classification algorithm with strong noise robustness is proposed in this paper, aiming at the hyperspectral image classification problems under noise interference. Based on the Fast 3D Convolutional Neural Network (Fast-3DCNN), this algorithm enables the classification model to have good tolerance for various types of noise by using a Minimum Noise Fraction (MNF) as dimensionality reduction module for hyperspectral image input data. In addition, by introducing lightweight hybrid attention modules with the spatial and the channel information, the deep features extracted by the Convolutional Neural Network are further refined, ensuring that the model has high classification accuracy. Public dataset experiments have shown that compared to traditional methods, the MNF in this algorithm reduces the dimensionality of input spectral data, preserves information with higher signal-to-noise ratio(SNR) in the spectral bands, and aggregates spectral features into class feature vectors, greatly improving the noise robustness of the model. At the same time, based on a lightweight spectral–spatial hybrid attention mechanism, combined with fewer spectral dimensions, the model effectively avoids overfitting. With less loss in model training speed, it achieved better classification accuracy in small-scale training sample experiments, fully demonstrating the good generalization ability of this algorithm.

Ground-penetrating radar subsurface attenuation analysis based on a sparsity-promoting time-frequency transform

The attenuation characteristics of a medium involve the absorption of electromagnetic wave energy during its propagation and represent a crucial parameter for describing the properties of subsurface media. Due to the complex nature of signal attenuation, demanding consideration of numerous parameters, quantitative analysis is challenging. Time-frequency (TF) analysis is a promising technique that simultaneously analyzes nonstationary signals in the time and frequency domains. Although the quantitative characterization of attenuation factors using TF tools can be challenging, TF analysis retains substantial potential for qualitatively estimating attenuation from ground-penetrating radar (GPR) data. The sparsity-promoting TF (SPTF) transform method with [Formula: see text] and [Formula: see text] norms joint constraints has a significantly higher TF resolution than the linear TF transform in the time and frequency domains. We develop a method for high-resolution qualitative estimation of GPR attenuation based on the SPTF method. Synthetic experiments indicate that SPTF surpasses other methods regarding TF resolution, noise robustness, and fidelity to dominant frequencies. Subsequent qualitative attenuation results using synthetic and field data agree with synthetic model outcomes and actual leakage ranges. This alignment substantiates the superior resolution and accuracy of our approach compared with alternative methods, affirming its efficacy and superiority in GPR attenuation depiction.

Robust and stochastic sparse subspace clustering

Sparse subspace clustering (SSC) has been widely employed in machine learning and pattern recognition, but it still faces scalability challenges when dealing with large-scale datasets. Recently, stochastic SSC (SSSC) has emerged as an effective solution by leveraging the dropout technique. However, SSSC cannot robustly handle noise, especially non-Gaussian noise, leading to unsatisfactory clustering performance. To address the above issues, we propose a novel robust and stochastic method called stochastic sparse subspace clustering with the Huber function (S3CH). The key idea is to introduce the Huber surrogate to measure the loss of the stochastic self-expression framework, thus S3CH inherits the advantage of the stochastic framework for large-scale problems while mitigating sensitivity to non-Gaussian noise. In algorithms, an efficient proximal alternating minimization (PAM)-based optimization scheme is developed. In theory, the convergence of the generated sequence is rigorously proved. Extensive numerical experiments on synthetic and six real datasets validate the advantages of the proposed method in clustering accuracy, noise robustness, parameter sensitivity, post-hoc analysis, and model stability.

Enhancing accuracy and robustness in FBS decoupling through Tikhonov regularization and unit balancing

Frequency-based Substructuring (FBS) enables the prediction of transfer functions for an assembled system based on those of a single-unit system. Conversely, FBS decoupling derives the frequency response function (FRF) of a single-unit system from the assembled system. This approach is valuable when obtaining the transfer functions of a system is challenging due to boundary conditions and deformation conditions. FBS decoupling can enhance accuracy by using additional indicator sensors. However, this often results in significant noise due to the ill-posed FRF matrix inversion. In this paper, Tikhonov regularization is employed to reduce noise-induced errors. Furthermore, this paper addresses the issue of inadequate Tikhonov regularization caused by the inherent scaling differences between the rotational and translational components within the FRF matrix. To rectify this challenge, a unit balancing method is proposed to mitigate the influence of scaling differences. This method enhances the accuracy and reliability of FRF matrix inversion processes, contributing to enhanced noise reduction and robustness in practical applications.

Direct Identification of the Continuous Relaxation Time and Frequency Spectra of Viscoelastic Materials

Relaxation time and frequency spectra are not directly available by measurement. To determine them, an ill-posed inverse problem must be solved based on relaxation stress or oscillatory shear relaxation data. Therefore, the quality of spectra models has only been assessed indirectly by examining the fit of the experiment data to the relaxation modulus or dynamic moduli models. As the measures of data fitting, the mean sum of the moduli square errors were usually used, the minimization of which was an essential step of the identification algorithms. The aim of this paper was to determine a relaxation spectrum model that best approximates the real unknown spectrum in a direct manner. It was assumed that discrete-time noise-corrupted measurements of a relaxation modulus obtained in the stress relaxation experiment are available for identification. A modified relaxation frequency spectrum was defined as a quotient of the real relaxation spectrum and relaxation frequency and expanded into a series of linearly independent exponential functions that are known to constitute a basis of the space of square-integrable functions. The spectrum model, given by a finite series of these basis functions, was assumed. An integral-square error between the real unknown modified spectrum and the spectrum model was taken as a measure of the model quality. This index was proved to be expressed in terms of the measurable relaxation modulus at uniquely defined sampling instants. Next, an empirical identification index was introduced in which the values of the real relaxation modulus are replaced by their noisy measurements. The identification consists of determining the spectrum model that minimizes this empirical index. Tikhonov regularization was applied to guarantee model smoothness and noise robustness. A simple analytical formula was derived to calculate the optimal model parameters and expressed in terms of the singular value decomposition. A complete identification algorithm was developed. The analysis of the model smoothness and model accuracy for noisy measurements was carried out. The equivalence of the direct identification of the relaxation frequency and time spectra has been demonstrated when the time spectrum is modeled by a series of functions given by the product of the relaxation frequency and its exponential function. The direct identification concept can be applied to both viscoelastic fluids and solids; however, some limitations to its applicability have been pointed out. Numerical studies have shown that the proposed identification algorithm can be successfully used to identify Gaussian-like and Kohlrausch–Williams–Watt relaxation spectra. The applicability of this approach to determining other commonly used classes of relaxation spectra was also examined.

Noise-robust modal parameter identification and damage assessment for aero-structures

PurposeGround vibration testing is critical for aircraft design and certification. Fast relaxed vector fitting (FRVF) and Loewner framework (LF), recently extended to modal parameter extraction in mechanical systems to address the computational challenges of time and frequency domain techniques, are applied for damage detection on aeronautically relevant structures.Design/methodology/approachFRVF and LF are applied to numerical datasets to assess noise robustness and performance for damage detection. Computational efficiency is also evaluated. In addition, they are applied to a novel damage detection benchmark of a high aspect ratio wing, comparing their performance with the state-of-the-art method N4SID.FindingsFRVF and LF detect structural changes effectively; LF exhibits better noise robustness, while FRVF is more computationally efficient.Practical implicationsLF is recommended for noisy measurements.Originality/valueTo the best of the authors’ knowledge, this is the first study in which the LF and FRVF are applied for the extraction of the modal parameters in aeronautically relevant structures. In addition, a novel damage detection benchmark of a high-aspect-ratio wing is introduced.

WPConvNet: An Interpretable Wavelet Packet Kernel-Constrained Convolutional Network for Noise-Robust Fault Diagnosis.

Deep learning (DL) has present great diagnostic results in fault diagnosis field. However, the poor interpretability and noise robustness of DL-based methods are still the main factors limiting their wide application in industry. To address these issues, an interpretable wavelet packet kernel-constrained convolutional network (WPConvNet) is proposed for noise-robust fault diagnosis, which combines the feature extraction ability of wavelet bases and the learning ability of convolutional kernels together. First, the wavelet packet convolutional (WPConv) layer is proposed, and constraints are imposed to convolutional kernels, so that each convolution layer is a learnable discrete wavelet transform. Second, a soft threshold activation is proposed to reduce the noise component in feature maps, whose threshold is adaptively learned by estimating the standard deviation of noise. Third, we link the cascaded convolutional structure of convolutional neutral network (CNN) with wavelet packet decomposition and reconstruction using Mallat algorithm, which is interpretable in model architecture. Extensive experiments are carried out on two bearing fault datasets, and the results show that the proposed architecture outperforms other diagnosis models in terms of interpretability and noise robustness.

Noise-robust automated sudden damage detection using blind source separation enhanced by variational mode decomposition and support vector machine based on shapelet transform

Compared to parametric counterparts, non-parametric (aka, model-free) damage detection methods have no requirements of accurate models, with the potential of autonomous monitoring of various complex structures. However, noises, or low signal-to-noise ratio (SNR), are one of the main challenges. This study is aimed at improving blind source separation (BSS)-based damage detection method, one of the most advanced non-parametric methods, in both aspects of noise robustness and autonomous operation. In particular, the measured acceleration responses are processed by variational mode decomposition (VMD) and wavelet transform (WT) in sequential, acting as the input for a BSS model. The BSS is then solved by independent component analysis (ICA), which approves to be more noise-robust compared to the state-of-the-art counterparts. Furthermore, shapelet transform is applied to extract the universal shape-based spike-like feature from the BSS model for training a support vector machine (SVM) model, applicable to different structures; it finally automates the sudden damage detection process and enables online monitoring. The effectiveness of the proposed method is illustrated by a numerical example and an experimental test, and demonstrated by a real-world seismic-excited structure. The results show that both single and multiple sudden damages can be automatically detected with high accuracy. Compared with the existing BSS methods, the proposed BSS method is more capable to detect small damages at relatively low SNR. In addition, the classification accuracy of SVM is also improved when shapelet-based feature is used for training, which reduces the malfunction of automated damage detection as shown by the numerical example. Therefore, the proposed strategy has the potential for rapid condition assessment of structures during rare/extreme events, before engineers are sent for further post-event inspection.

Representing Noisy Image Without Denoising.

A long-standing topic in artificial intelligence is the effective recognition of patterns from noisy images. In this regard, the recent data-driven paradigm considers 1) improving the representation robustness by adding noisy samples in training phase (i.e., data augmentation) or 2) pre-processing the noisy image by learning to solve the inverse problem (i.e., image denoising). However, such methods generally exhibit inefficient process and unstable result, limiting their practical applications. In this paper, we explore a non-learning paradigm that aims to derive robust representation directly from noisy images, without the denoising as pre-processing. Here, the noise-robust representation is designed as Fractional-order Moments in Radon space (FMR), with also beneficial properties of orthogonality and rotation invariance. Unlike earlier integer-order methods, our work is a more generic design taking such classical methods as special cases, and the introduced fractional-order parameter offers time-frequency analysis capability that is not available in classical methods. Formally, both implicit and explicit paths for constructing the FMR are discussed in detail. Extensive simulation experiments and robust visual applications are provided to demonstrate the uniqueness and usefulness of our FMR, especially for noise robustness, rotation invariance, and time-frequency discriminability.

Analog Sequential Hippocampal Memory Model for Trajectory Learning and Recalling: A Robustness Analysis Overview

The rapid expansion of information systems in all areas of society demands more powerful, efficient, and low‐energy consumption computing systems. Neuromorphic engineering has emerged as a solution that attempts to mimic the brain to incorporate its capabilities to solve complex problems in a computationally and energy‐efficient way in real time. Within neuromorphic computing, building systems to efficiently store the information is still a challenge. Among all the brain regions, the hippocampus stands out as a short‐term memory capable of learning and recalling large amounts of information quickly and efficiently. Herein, a spike‐based bio‐inspired hippocampus sequential memory model is proposed that makes use of the benefits of analog computing and spiking neural networks (SNNs): noise robustness, improved real‐time operation, and energy efficiency. This model is applied to robotic navigation to learn and recall trajectories that lead to a goal position within a known grid environment. The model is implemented on the special‐purpose SNNs mixed‐signal DYNAP‐SE hardware platform. Through extensive experimentation together with an extensive analysis of the model's behavior in the presence of external noise sources, its correct functioning is demonstrated, proving the robustness and consistency of the proposed neuromorphic sequential memory system.

A Fault Diagnosis Method for Pumped Storage Unit Stator Based on Improved STFT-SVDD Hybrid Algorithm

Stator faults are one of the common issues in pumped storage generators, significantly impacting their performance and safety. To ensure the safe and stable operation of pumped storage generators, a stator fault diagnosis method based on an improved short-time Fourier transform (STFT)-support vector data description (SVDD) hybrid algorithm is proposed. This method establishes a fault model for inter-turn short circuits in the stator windings of pumped storage generators and analyzes the electrical and magnetic states associated with such faults. Based on the three-phase current signals observed during an inter-turn short circuit fault in the stator windings, the three-phase currents are first converted into two-phase currents using the principle of equal magnetic potential. Then, the STFT is applied to transform the time-domain signals of the stator’s two-phase currents into frequency-domain signals, and the resulting fault current spectrum is input into the improved SVDD network for processing. This ultimately outputs the diagnosis result for inter-turn short circuit faults in the stator windings of the pumped storage generator. Experimental results demonstrate that this method can effectively distinguish between normal and faulty states in pumped storage generators, enabling the diagnosis of inter-turn short circuit faults in stator windings with low cross-entropy loss. Through analysis, under small data sample conditions, the accuracy of the proposed method in this paper can be improved by up to 7.2%. In the presence of strong noise interference, the fault diagnosis accuracy of the proposed method remains above 90%, and compared to conventional methods, the fault diagnosis accuracy can be improved by up to 6.9%. This demonstrates that the proposed method possesses excellent noise robustness and small sample learning ability, making it effective in complex, dynamic, and noisy environments.

Operating modal parameter identification of railway vehicle bogie based on k-value optimization variational mode decomposition and second-order blind identification method

Due to the variable operating conditions and strong background noise, the accuracy of modal parameter identification of the railway vehicle bogie is low. This paper proposes a method for identifying operational modal parameters by combining the K-value optimization variational mode decomposition (KVMD) and second-order blind identification (SOBI) method (KVMD–SOBI method). Firstly, the variational mode decomposition (VMD) algorithm is used to adaptively decompose a single-channel vibration signal into intrinsic mode function (IMF) components with different scale characteristics. The feature IMF components are selected through the K-value optimization method based on energy ratio and stability graph. Then, the SOBI algorithm is used to separate the matrix composed of the feature IMF components. And the mode shape and single-mode signal are obtained. Finally, the natural frequency and damping ratio of each order mode are extracted by the single-mode parameter recognition method. Through simulation and experimental verification, it is shown that the proposed method has good noise robustness and can effectively identify the first three modal parameters of railway vehicle bogies in variable-speed operation. At the same time, the feature IMF components extracted by KVMD algorithm eliminate the sensitivity of sensor installation positions and operating conditions to detection results. Therefore, this method is an excellent operational mode identification method based on a single-channel signal with good noise and working condition robustness.

Energy bubble entropy guided symplectic geometry mode decomposition for rotating machinery incipient fault feature extraction

Abstract Extracting incipient fault features is a critical aspect of monitoring the rotating machinery operation condition. However, existing methods based on symplectic geometry mode decomposition (SGMD) suffer from limited parameter adaptability and noise robustness. Therefore, this paper proposes an Energy bubble entropy guided SGMD method to extract incipient fault feature. Firstly, the SGMD method is employed to initially separate fault characteristic components from noisy signal. Furthermore, the energy bubble entropy(EbEn) is constructed to evaluate the attributes of incipient feature within the signal, which requires almost no parameter setting with good robustness and computational efficiency. Thirdly, the Empirical Bayes shrinkage method effectively mitigates irrelevant noise and enhances the significance of incipient fault feature. Simulated and experimental signals are employed to substantiate the efficacy of the EbEn guided SGMD method. The comparison analysis with relevant methods exhibits that this method has greater robustness and adaptivity.

Robustness enhancement in neural networks with alpha-stable training noise

With the increasing use of deep learning on data collected by non-perfect sensors and in non-perfect environments, the robustness of deep learning systems has become an important issue. A common approach for obtaining robustness to noise has been to train deep learning systems with data augmented with Gaussian noise. In this work, the common choice of Gaussian noise is challenged and the possibility of stronger robustness for non-Gaussian impulsive noise is explored, specifically alpha-stable noise. Justified by the Generalized Central Limit Theorem and evidenced by observations in various application areas, alpha-stable noise is widely present in nature. By comparing the testing accuracy of models trained with Gaussian noise and alpha-stable noise on data corrupted by different noise, it is found that training with alpha-stable noise is more effective than Gaussian noise, especially when the dataset is corrupted by impulsive noise, thus improving the robustness of the model. Moreover, in the testing on the common corruption benchmark dataset, training with alpha-stable noise also achieves promising results, improving the robustness of the model to other corruption types and demonstrating comparable performance with other state-of-the-art data augmentation methods. Consequently, a novel data augmentation method is proposed that replaces Gaussian noise, which is typically added to the training data, with alpha-stable noise.

Regulating Modality Utilization within Multimodal Fusion Networks.

Multimodal fusion networks play a pivotal role in leveraging diverse sources of information for enhanced machine learning applications in aerial imagery. However, current approaches often suffer from a bias towards certain modalities, diminishing the potential benefits of multimodal data. This paper addresses this issue by proposing a novel modality utilization-based training method for multimodal fusion networks. The method aims to guide the network's utilization on its input modalities, ensuring a balanced integration of complementary information streams, effectively mitigating the overutilization of dominant modalities. The method is validated on multimodal aerial imagery classification and image segmentation tasks, effectively maintaining modality utilization within ±10% of the user-defined target utilization and demonstrating the versatility and efficacy of the proposed method across various applications. Furthermore, the study explores the robustness of the fusion networks against noise in input modalities, a crucial aspect in real-world scenarios. The method showcases better noise robustness by maintaining performance amidst environmental changes affecting different aerial imagery sensing modalities. The network trained with 75.0% EO utilization achieves significantly better accuracy (81.4%) in noisy conditions (noise variance = 0.12) compared to traditional training methods with 99.59% EO utilization (73.7%). Additionally, it maintains an average accuracy of 85.0% across different noise levels, outperforming the traditional method's average accuracy of 81.9%. Overall, the proposed approach presents a significant step towards harnessing the full potential of multimodal data fusion in diverse machine learning applications such as robotics, healthcare, satellite imagery, and defense applications.