Complex Noise Research Articles

Research on Transformer Voiceprint Feature Extraction Oriented to Complex Noise Environment

Transformer fault diagnosis based on acoustic characteristics is a new non-contact and non-destructive monitoring method. It has the advantages that the acoustic signal detection is not disturbed by electric and magnetic fields, and the monitoring process does not affect the normal operation of the transformer. Aiming at the difficulty of extracting transformer voiceprint features in complex noise environment, a transformer voiceprint feature extraction method based on Variable Mode Extraction (VME) is proposed. In this method, the center frequency of the Intrinsic Mode Function(IMF) is set according to the generation mechanism of transformer radiated noise, thus the uncertainty of decomposition results caused by random distribution and other frequency search methods is eliminated; Then, taking the frequency-domain energy aggregation of IMF and the minimum center frequency energy of residual signal as the optimization objectives, the cyclic iterative decomposition is used to identify and extract the transformer voiceprint features, so as to reduce the impact of environmental noise and other equipment noise. The analysis results of simulation signals and field signals show that this method can effectively reduce the impact of environmental noise and extract more clear and accurate transformer voiceprint features.

Direction of Arrival Estimation for Single Microelectromechanical Systems Vector Hydrophone Using Modified Wavelet Packet De-Noising

The complex and variable ocean environment introduces complex noise to the signal received by microelectromechanical systems (MEMS) vector hydrophones, which seriously affects target location. To improve the detection efficiency and broaden the application range of vector hydrophone, this paper proposed a novel modified Wavelet packet de-noising method, known as the Wavelet packet energy (WPE) de-noising method. This method utilizes Shannon entropy to determine the optimal decomposition levels of Wavelet packets, and calculates the energy of reconstructed signals in each frequency band to construct feature vectors. The Wavelet packet with the larger energy is selected to reconstruct the signal until the root mean square error of the original signal and the reconstructed signal is minimized. thereby obtaining an optimal solution. The performance of this proposed method is evaluated through both indoor and outdoor experiments. The experimental results demonstrate that the developed method is effective in Direction of Arrival (DOA) estimation of single frequency signals and is feasible in low signal-to-noise ratio situations. The method improves the azimuth discrimination ability of the Minimum Variance Distortionless Response algorithm, and DOA estimation errors are all within 1°. Furthermore, the indoor and outdoor experiments also demonstrate that the WPE de-noising method solves the group delay and distortion of signals caused by conventional filtering.

An Unsupervised Image Denoising Method Using a Nonconvex Low-Rank Model with TV Regularization

In real-world scenarios, images may be affected by additional noise during compression and transmission, which interferes with postprocessing such as image segmentation and feature extraction. Image noise can also be induced by environmental variables and imperfections in the imaging equipment. Robust principal component analysis (RPCA), one of the traditional approaches for denoising images, suffers from a failure to efficiently use the background’s low-rank prior information, which lowers its effectiveness under complex noise backgrounds. In this paper, we propose a robust PCA method based on a nonconvex low-rank approximation and total variational regularization (TV) to model the image denoising problem in order to improve the denoising performance. Firstly, we use a nonconvex γ-norm to address the issue that the traditional nuclear norm penalizes large singular values excessively. The rank approximation is more accurate than the nuclear norm thanks to the elimination of matrix elements with substantial approximation errors to reduce the sparsity error. The method’s robustness is improved by utilizing the low sensitivity of the γ-norm to outliers. Secondly, we use the l1-norm to increase the sparsity of the foreground noise. The TV norm is used to improve the smoothness of the graph structure in accordance with the sparsity of the image in the gradient domain. The denoising effectiveness of the model is increased by employing the alternating direction multiplier strategy to locate the global optimal solution. It is important to note that our method does not require any labeled images, and its unsupervised denoising principle enables the generalization of the method to different scenarios for application. Our method can perform denoising experiments on images with different types of noise. Extensive experiments show that our method can fully preserve the edge structure information of the image, preserve important features of the image, and maintain excellent visual effects in terms of brightness smoothing.

Denoising distributed acoustic sensing data using unsupervised deep learning

Distributed acoustic sensing (DAS) technology has been widely used in seismic exploration to acquire high-quality data due to its noteworthy advantages, such as high coverage, high resolution, low cost, and strong environmental friendliness. However, the seismic signals acquired in DAS are often masked by various types of noise (e.g., high-frequency random, high-amplitude erratic, horizontal, and coupled noise), which seriously decreases the signal-to-noise ratio. We develop a fully connected neural network with dense and residual connections to attenuate various complex noises in real DAS data. The network is designed to learn the features of useful reflection signals and remove various noises in an unsupervised way, therefore enjoying the convenience of label-free processing. Our network uses several encoders and decoders to compress and reconstruct the abstract waveform features, respectively. Each encoder/decoder consists of one dense block with stacked fully connected blocks (FCBs). To transfer the shallow-level features to the deep level for reuse, we add the skip connections with one FCB between the corresponding encoders and decoders. Our method provides encouraging results when applied to synthetic and real DAS data sets. Compared with several traditional and advanced deep-learning methods, our method can more effectively attenuate strong noise and better extract hidden signals.

A three-dimensional microseismic downhole noise suppression based on polarization filtering method in Shearlet transform

Microseismic noise suppression is widely used in the exploration of unconventional oil and gas resources. The effective microseismic downhole signals have extremely weak energy and are contaminated by strong interference, making data processing and interpretation difficult. The need for high-frequency effective signal reservation presents a basic problem in the design of noise suppression methods. The effective signals represent as the continuous reflection event and have more concentrated features in the transform domain, which can be used to tell the signal from the irregular microseismic noise. However, the high-frequency signal and extremely complex noise bring difficulty in accurately separating them by a single threshold. In this study, we propose a novel denoising method called Shearlet-polarization filtering to effectively suppress the microseismic noise. In general, Shearlet-polarization filtering is the combination of polarization filtering and conventional Shearlet transform. Specifically, the Shearlet transform can decompose the microseismic data into multi-directional and multi-scale information, providing a solid foundation for the separation of effective signals and background noise. From this basis, polarization filtering achieves signal reservation and noise attenuation by making full use of the three-dimensional information. To evaluate the performance, we also compare the proposed method with conventional Shearlet threshold filtering and polarization filtering. Experimental results both in synthetic and field data processing indicate that the Shearlet-polarization filtering is superior to the competing methods because it can significantly improve the continuity and smoothness of the microseismic events, even in low SNR conditions.

TelecomTM

We introduce the TelecomTM system that uses pre-existing telecommunication fiber-optic cables as virtual strain sensors to sense vehicle-induced ground vibrations for fine-grained and ubiquitous traffic monitoring and characterization. Here we call it a virtual sensor because it is a software-based representation of a physical sensor. Due to the extensively installed telecommunication fiber-optic cables at the roadside, our system using redundant dark fibers enables to monitor traffic at low cost with low maintenance. Many existing traffic monitoring approaches use cameras, piezoelectric sensors, and smartphones, but they are limited due to privacy concerns and/or deployment requirements. Previous studies attempted to use telecommunication cables for traffic monitoring, but they were only exploratory and limited to simple tasks at a coarse granularity, e.g., vehicle detection, due to their hardware constraints and real-world challenges. In particular, those challenges are 1) unknown and heterogeneous properties of virtual sensors and 2) large and complex noise conditions. To this end, our TelecomTM system first characterizes the geographic location and analyzes the signal pattern of each virtual sensor through driving tests. We then develop a spatial-domain Bayesian filtering and smoothing algorithm to detect, track, and characterize each vehicle. Our approach uses the spatial dependency of multiple virtual sensors and Newton's laws of motion to combine the distributed sensor data to reduce uncertainties in vehicle detection and tracking. In our real-world evaluation on a two-way traffic road with 1120 virtual sensors, TelecomTM achieved 90.18% vehicle detection accuracy, 27x and 5x error reduction for vehicle position and speed tracking compared to a baseline method, and ±3.92% and ±11.98% percent error for vehicle wheelbase and weight estimation, respectively.

A new weak fault diagnosis approach for train bearings based on improved grey wolf optimizer and adaptive variational mode decomposition

Accurately identifying the health status of train running gear bearings is crucial to ensure the quality of operation. As the early fault information of bearings is weak and submerged in the complex noise environment, which is difficult to diagnose. Therefore, a new weak fault diagnosis approach for train running gear bearings based on variational mode decomposition (VMD) with improved performance and refined weighted kurtosis (RWK) index is proposed to solve this problem. First, an improved grey wolf optimizer (IGWO) based on a variety of strategies is proposed. Secondly, the VMD performance is improved using the IGWO algorithm, and the improved VMD is used to process the early weak signals of bearings. A new fault-sensitive index called the RWK is proposed to detect the mode with the most fault information. Finally, the envelope analysis of the characteristic signals is performed to achieve the early weak fault diagnosis of bearings. Compared with the other nine optimization algorithms, the IGWO algorithm has strong optimization ability, stable performance and a fast convergence speed. Four cases verify that the RWK index has the highest sensitivity to fault information and can more effectively filter out modal components containing rich fault information than the comparison methods.

Adaptive noise suppression for low-S/N microseismic data based on ambient-noise-assisted multivariate empirical mode decomposition

Microseismic monitoring data may be seriously contaminated by complex and nonstationary interference noises produced by mechanical vibration, which significantly impact the data quality and subsequent data-processing procedure. One challenge in microseismic data processing is separating weak seismic signals from varying noisy data. To address this issue, we proposed an ambient-noise-assisted multivariate empirical mode decomposition (ANA-MEMD) method for adaptively suppressing noise in low signal-to-noise (S/N) microseismic data. In the proposed method, a new multi-channel record is produced by combining the noisy microseismic signal with preceding ambient noises. The multi-channel record is then decomposed using multivariate empirical mode decomposition (MEMD) into multivariate intrinsic mode functions (MIMFs). Then, the MIMFs corresponding to the main ambient noises can be identified by calculating and sorting energy percentage in descending order. Finally, the IMFs associated with strong interference noise, high-frequency and low-frequency noise are filtered out and suppressed by the energy percentage and frequency range. We investigate the feasibility and reliability of the proposed method using both synthetic data and field data. The results demonstrate that the proposed method can mitigate the mode mixing problem and clarify the main noise contributors by adding additional ambient-noise-assisted channels, hence separating the microseismic signal and ambient noise effectively and enhancing the S/Ns of microseismic signals.

Periodic Detection Mode Decomposition and Its Application in Bearing Fault Diagnosis

In the complex background noise, it is difficult to extract the periodic pulse of the rolling bearing fault signal. In this paper, a new modal decomposition method supported by singular value ratio (SVR) spectrum is proposed to find the optimal period of bearing fault data, which can be named as periodic detection mode decomposition (PDMD). In order to detect the intervals of periods and eliminate the interference of useless periods, the initial measurement intervals in this method are divided according to the theoretical fault periods of different fault types of bearings. In each interval, the SVR spectrum is used to detect the appropriate period and suppress the influence of noise on the recognition process. This period is used to construct the optimal Ramanujan subspace (RS). Finally, the harmonic spectral kurtosis (HSK) is used to identify the extracted period as false information, interference or fault. Simulation and experimental data verify the effectiveness of the proposed method. This method can effectively extract and screen periodic pulses, and can successfully identify the outer and inner ring faults of bearings.

Cascaded Denoising Transformer for UAV Nighttime Tracking

The automation of unmanned aerial vehicles (UAVs) has been greatly promoted by visual object tracking methods with onboard cameras. However, the random and complicated real noise produced by the cameras seriously hinders the performance of state-of-the-art (SOTA) UAV trackers, especially in low-illumination environments. To address this issue, this work proposes an efficient plug-and-play cascaded denoising Transformer (CDT) to suppress cluttered and complex real noise, thereby boosting UAV tracking performance. Specifically, the novel U -shaped cascaded denoising network is designed with a streamlined structure for efficient computation. Additionally, shallow feature deepening (SFD) encoder and multi-feature collaboration (MFC) decoder are constructed based on multi-head transposed self-attention (MTSA) and multi-head transposed cross-attention (MTCA), respectively. A nested residual feed-forward network (NRFN) is developed to focus more on high-frequency information represented by noise. Extensive evaluation and test experiments demonstrate that the proposed CDT has a remarkable denoising effect and improves UAV nighttime tracking performance.

Beyond Gaussian Noise: A Generalized Approach to Likelihood Analysis with Non-Gaussian Noise

Likelihood analysis is typically limited to normally distributed noise due to the difficulty of determining the probability density function of complex, high-dimensional, non-Gaussian, and anisotropic noise. This is a major limitation for precision measurements in many domains of science, including astrophysics, for example, for the analysis of the cosmic microwave background, gravitational waves, gravitational lensing, and exoplanets. This work presents Score-based LIkelihood Characterization, a framework that resolves this issue by building a data-driven noise model using a set of noise realizations from observations. We show that the approach produces unbiased and precise likelihoods even in the presence of highly non-Gaussian correlated and spatially varying noise. We use diffusion generative models to estimate the gradient of the probability density of noise with respect to data elements. In combination with the Jacobian of the physical model of the signal, we use Langevin sampling to produce independent samples from the unbiased likelihood. We demonstrate the effectiveness of the method using real data from the Hubble Space Telescope and James Webb Space Telescope.

W-Net: A boundary-enhanced segmentation network for stroke lesions

Accurate lesion segmentation is a critical technology basis for the treatment and prognosis of stroke. Stroke lesion segmentation suffers from complex background and noise interferes due to the similar grayscale characteristics of tissue structures and imaging mode in medical image. To resolve the difficulty of fuzzy boundary for lesion segmentation, we propose a novel two-stage brain MRI lesion segmentation method, called W-Net. W-Net leverages the CNN and transformer-based method as the backbone network, and introduces boundary deformation module (BDM), boundary constraint module (BCM) to handle the tough condition of fuzzy boundary. The BDM utilizes circular convolution to correct the initial boundary and BCM exploits dilated convolution to dynamically constrain the object boundary. Moreover, we design a multi-task learning loss function to optimize the W-Net from both region and boundary perspectives. Extensive experiments are conducted using the two ischemic stroke lesion segmentation datasets, and the results are compared with other advanced models. The experimental results demonstrate that the proposed method is effective in preserving the edge information of stroke lesions, thus achieving competitive performance on the segmentation task.

PPA-Net: Pyramid Pooling Attention Network for Multi-Scale Ship Detection in SAR Images

In light of recent advances in deep learning and Synthetic Aperture Radar (SAR) technology, there has been a growing adoption of ship detection models that are based on deep learning methodologies. However, the efficiency of SAR ship detection models is significantly impacted by complex backgrounds, noise, and multi-scale ships (the number of pixels occupied by ships in SAR images varies significantly). To address the aforementioned issues, this research proposes a Pyramid Pooling Attention Network (PPA-Net) for SAR multi-scale ship detection. Firstly, a Pyramid Pooled Attention Module (PPAM) is designed to alleviate the influence of background noise on ship detection while its parallel component favors the processing of multiple ship sizes. Different from the previous attention module, the PPAM module can better suppress the background noise in SAR images because it considers the saliency of ships in SAR images. Secondly, an Adaptive Feature Balancing Module (AFBM) is developed, which can automatically balance the conflict between ship semantic information and location information. Finally, the detection capabilities of the ship detection model for multi-scale ships are further improved by introducing the Atrous Spatial Pyramid Pooling (ASPP) module. This innovative module enhances the detection model’s ability to detect ships of varying scales by extracting features from multiple scales using atrous convolutions and spatial pyramid pooling. PPA-Net achieved detection accuracies of 95.19% and 89.27% on the High-Resolution SAR Images Dataset (HRSID) and the SAR Ship Detection Dataset (SSDD), respectively. The experimental results demonstrate that PPA-Net outperforms other ship detection models.

Optimized interferometric encoding of presaturated TurboFLASH B1 mapping for parallel transmission MRI at 7 T: Preliminary application for quantitative T1 mapping in the spinal cord.

The acquisition of accurate B1 maps is critical for parallel transmit techniques (pTx). The presaturated turboFLASH (satTFL) method has been widely used in combination with interferometric encoding to provide robust and fast B1 maps. However, typical encodings, mostly evaluated on brain, do not necessarily fit all coils and organs. In this work, we evaluated and improved the accuracy of the satTFL for cervical spine at 7 T, proposing a novel interferometric encoding optimization. The benefits of such improvements were investigated in an exploratory study of quantitative T1 mapping with pTx-MP2RAGE. Global optimization of interferometric encoding was implemented by simulating the ability of the satTFL to reconstruct B1 maps, with varying encoding and inclusion of complex noise, inside a region of interest covering the cervical spine. The performance of satTFL before and after optimization was compared to actual flip angle imaging. Optimized and non-optimized B1 maps were then used to calculate pTx pulses for MP2RAGE T1 mapping. Interferometric encoding optimization resulted in satTFL closer to actual flip angle imaging, with substantial gain of signal in regions where non-optimized satTFL could fail. T1 maps measured with non-adiabatic pTx pulses were closer to standard non-pTx results (which used adiabatic pulses) when using optimized-satTFL, with substantially lower specific absorption rate. Optimization of the satTFL interferometric encoding improves B1 maps in the spinal cord, in particular in low SNR regions. A linear correction of the satTFL was additionally shown to be required. The method was successfully used for quantitative phantom and in vivo T1 mapping, showing improved results compared to non-optimized satTFL thanks to improved pTx-pulse generation.

Fault Diagnosis Method of Roadheader Bearing Based on VMD and Domain Adaptive Transfer Learning.

The roadheader is a core piece of equipment for underground mining. The roadheader bearing, as its key component, often works under complex working conditions and bears large radial and axial forces. Its health is critical to efficient and safe underground operation. The early failure of a roadheader bearing has weak impact characteristics and is often submerged in complex and strong background noise. Therefore, a fault diagnosis strategy that combines variational mode decomposition and a domain adaptive convolutional neural network is proposed in this paper. To start with, VMD is utilized to decompose the collected vibration signals to obtain the sub-component IMF. Then, the kurtosis index of IMF is calculated, with the maximum index value chosen as the input of the neural network. A deep transfer learning strategy is introduced to solve the problem of the different distributions of vibration data for roadheader bearings under variable working conditions. This method was implemented in the actual bearing fault diagnosis of a roadheader. The experimental results indicate that the method is superior in terms of diagnostic accuracy and has practical engineering application value.

Adaptive parameter selection variational mode decomposition based on a novel hybrid entropy and its applications in locomotive bearing diagnosis

The emission signal of locomotive bearings is non-stationary, nonlinear and mixed with complex noise when operating online. The adaptive decomposition method is widely employed for denoising and feature extraction. Variational mode decomposition (VMD) is a powerful adaptive decomposition method, but the number of modes K and the penalty factor α are difficult to set in advance and the unreasonable parameter selection may probably bring faulty feature loss. To address this problem, a novel hybrid entropy, which includes the envelope fuzzy and dispersion characteristics of entropy, is proposed to describe the complexity and detect dynamic change of intrinsic mode functions component of VMD. Furthermore, an adaptive parameter selection variational mode decomposition (APSVMD) based on the novel hybrid entropy is presented to obtain the optimal parameter combination (K and α). Then the sensitive intrinsic mode function (IMF) can be chosen by novel hybrid entropy based on its new features. In order to verify the performance of the proposed diagnostic approach, the locomotive online running test was performed. The experimental results demonstrated that the APSVMD represents the signal modes explicitly and can process different types of signals. As a result, different kinds of bearings faults were diagnosed successfully.

Study on signal decomposition-based pump noise cancellation method for continuous-wave mud pulse telemetry

Continuous mud pulse telemetry is an advanced downhole data transmission technology in measurement while drilling (MWD) systems. However, various noise interference during transmission has resulted in continuous mud pulse telemetry often encountering difficulties in decoding under harsh working conditions, with the influence of pump noise being the most severe. On the bases of the narrowband signal characteristics of pump noise harmonics and a deep study of the variational mode decomposition (VMD) algorithm, this study proposes an improved version called constant center frequency VMD (CVMD), the mode function center frequencies of which are fixed in the initial state. The denoising performance of the proposed algorithm is thoroughly analyzed by simulating a series of continuous mud pulse signals with different noise intensities under stable/unstable mud pump conditions. Results demonstrate that the CVMD algorithm has an advantage over conventional filters because the former can maximize the preservation of useful information within the critical frequency band. Compared with the Kalman filter based on the same pump noise model, the CVMD algorithm has a higher signal-to-noise ratio, smaller mean square error, higher correlation with the ideal signal, and is also applicable when the pump state is unstable. In addition, the proposed method has been successfully applied to the processing of real-field experimental data transmitted over thousands of meters and disturbed by complex noise. Consequently, a distinctive waveform of the transmitted pulse is obtained, and powerful technical support is provided for the realization of continuous mud pulse downhole data transmission in harsh mud circulation environments.

An efficient calculation method of large-region dynamic traffic noise maps based on hybrid modeling

The construction of noise maps is of great significance for the management and control of urban noise and the protection of residents’ physical and mental health. The European Noise Directive recommends using computational methods to construct strategic noise maps when possible. The current noise maps based on model calculation rely on complex noise emission and propagation models, and their huge number of regional grids needs to consume a lot of calculation time. This seriously restricts the update efficiency of noise maps, making it difficult to realize large-scale application and real-time dynamic update of noise maps. In order to improve the computational efficiency of noise maps, based on big data-driven technology, this paper combines the traditional CNOSSOS-EU noise emission modeling method with the multivariate nonlinear regression modeling method, and proposes an efficient calculation method of large-region dynamic traffic noise maps based on hybrid modeling method. First, this paper constructs the (daily and nightly) noise contribution prediction models of road sources with different classes, considering the daily and nightly periods and different urban road classes. Parameters of the proposed model are evaluated by using the multivariate nonlinear regression method to replace the complex nonlinear acoustic mechanism modeling. On this basis, in order to further improve the computational efficiency, noise contribution attenuations of the constructed models are parameterized and evaluated quantitatively. And then, the database containing the index table of the road noise sources-receivers and the corresponding noise contribution attenuations is constructed. The experimental results show that compared with the traditional calculation methods based on acoustic mechanism model, the noise map calculation method based on hybrid model proposed in this paper greatly reduces the model computations of noise map, improves the efficiency of noise mapping. It will provide technical support for constructing dynamic noise maps of large urban regions.

Medical image categorization combining image segmentation and machine learning

The death rate has increased in recent years due to the rising prevalence of encephaloma tumors across all age brackets. Because of their complex structure and background noise, tumors are difficult to detect in medical imaging and require a great deal of time and effort on the part of professionals. This is crucial since locating the tumor early on is key to successful treatment. Scans can detect and even forecast the presence of cancer at a variety of stages. A combination of these scans with segmentation and relegation techniques can aid in a rapid diagnosis, saving valuable time for the treating physician. Due to the complex nature of tumors and the gradual evolution of noise in MR imaging data, physical tumor identification has become a complicated and time-consuming process for medical professionals. Hence, early detection and localization of the tumor site is essential. Using segmentation and relegation techniques, medical imaging can pinpoint cancer tumors at multiple stages for a precise diagnosis. This study presents a machine learning-based method for automatically segmenting and labelling MRI scans of the brain to help in the detection of malignant growths. In addition, this framework employs a number of machine learning algorithms for tasks including image pre-processing, segmentation, feature extraction, and classification, including Nave Bayes, Nearest Neighbours, and Decision Table.

Optimization of W-DCGAN Fault Diagnosis Method Based on Self-attention Mechanism

Bearing component is the most important component of wind turbine, the harsh operating environment makes the bearing prone to failure, and the maintenance cost is very expensive when a failure occurs. Complex environmental noise brings serious noise pollution to fault samples. In addition, the serious sample imbalance between fault samples and normal samples brings great challenges to bearing fault diagnosis. In view of the above mentioned problems, this paper uses a self-attention mechanism optimization and Wasserstein distance improvement deep convolutional adversarial network model based on self-attention mechanism optimization and Wasserstein distance improvement deep Bearing fault diagnosis method based on the convolution generative Adversarial Network Model (SAW-DCGAN). The experimental results show that W-DCGAN has excellent generating ability and can generate samples like real samples to achieve the balance of samples. The addition of self-attention makes the classification features more expressive, and can accelerate the training speed of the model while having a higher fault diagnosis rate, which verifies the practicability of the method.