Noise Characteristics Research Articles

DAF‐Retinex: Preserve the image detailed features and restore the reflected image

AbstractCurrently, deep learning methods for low‐light image enhancement tasks mainly focus on the illumination of images, while neglecting the problems of image noise and feature loss. To address this issue, this paper proposes a novel low‐light image enhancement network called DAF‐Retinex, based on the Retinex‐Net. To address the issue of image noise, different from traditional image denoise methods, this paper utilizes a fully convolutional neural network to denoise the reflection component, additionally, a denoising loss function is introduced to suppress noise. For preserving image details and extracting features, this paper creatively introduces self‐calibrated convolutions into low‐light image enhancement tasks, furthermore, a feature augmented attention block consisting of feature‐guided attention (FGA) is designed for feature learning to effectively enhance image illumination and extract image detail features. Experimental results demonstrate that the proposed algorithm in this paper effectively removes image noise and extracts detailed features, resulting in visually improved outcomes. On public datasets, the average improvement in objective evaluation metrics of image quality such as PSNR, SSIM, and NIQE are 1.13%, 4.12%, and 1.28%, respectively.

THz-TDS with gigahertz Yb-based dual-comb lasers: noise analysis and mitigation strategies

We investigate terahertz time-domain spectroscopy using a low-noise dual-frequency-comb laser based on a single spatially multiplexed laser cavity. The laser cavity includes a reflective biprism, which enables generation of a pair of modelocked output pulse trains with slightly different repetition rates and highly correlated noise characteristics. These two pulse trains are used to generate the THz waves and detect them by equivalent time sampling. The laser is based on Yb:CALGO, operates at a nominal repetition rate of 1.18 GHz, and produces 110 mW per comb with 77 fs pulses around 1057 nm. We perform THz measurements with Fe-doped photoconductive antennas, operating these devices with gigahertz 1 µm lasers for the first time, to our knowledge, and obtain THz signal currents approximately as strong as those from reference measurements at 1.55 µm and 80 MHz. We investigate the influence of the laser’s timing noise properties on THz measurements, showing that the laser’s timing jitter is quantitatively explained by power-dependent shifts in center wavelength. We demonstrate reduction in noise by simple stabilization of the pump power and show up to 20 dB suppression in noise by the combination of shared pumping and shared cavity architecture. The laser’s ultra-low-noise properties enable averaging of the THz waveform for repetition rate differences from 1 kHz to 22 kHz, resulting in a dynamic range of 55 dB when operating at 1 kHz and averaging for 2 s. We show that the obtained dynamic range is competitive and can be well explained by accounting for the measured optical delay range, integration time, as well as the measurement bandwidth dependence of the noise from transimpedance amplification. These results will help enable a new approach to high-resolution THz-TDS enabled by low-noise gigahertz dual-comb lasers.

Numerical Simulation Analysis of the Correlation between Hydrodynamic Noise of Hydraulic Turbines and Defects in Runner Blades

To investigate the hydrodynamic noise characterization of hydraulic turbines with runner blade defects, this article establishes the intact machine model and three kinds of models with runner blade defects. Using the Computational Fluid Dynamic (CFD) and Computational Acoustic (CA) hybrid simulation computational methods, the hydrodynamic noise field of the hydraulic turbine is numerically simulated, and the results of the acoustic near field and acoustic far field are shown. 1. The double-row leaf grille and the runner are the primary sound source areas of the hydraulic turbine, and the intensity of sound radiation from these areas is positively correlated with the degree of runner blade defects. 2. As the runner blade defects develop, the sound power level (SWL) increases more significantly in the guide vanes near the nose of the spiral case in the double-row leaf grille. The most pronounced increase in the SWL is observed at the defective craters on the runner blades. 3. The frequency of the defective noise signal is primarily concentrated in the low-frequency band. The dominant frequency amplitude associated with runner blade defects increases and rises after the occurrence of defects. Secondary frequency changes are also observed, and the location of these changes varies at different receiving points.

Unsupervised complex semi-binary matrix factorization for activation sequence recovery of quasi-stationary sources

Industry 5.0 advocates for a sustainable industry, particularly in terms of energy. One of the most fundamental elements of comprehension is knowing when the systems are operated, as this is key to locating energy intensive subsystems and operations. Such knowledge is often lacking in practice. Activation statuses can be recovered from sensor data though. Some non-intrusive sensors (accelerometers, current sensors, etc.) acquire mixed signals containing information about multiple actuators at once. This paper considers different industrial settings in which the identification of binary sub-system activation sequences is sought. In this context, source signals are assumed to be quasi-stationary and they may be correlated. Existing methods either restrict these assumptions, e.g., by imposing orthogonality or noise characteristics, or lift them using additional assumptions, typically using nonlinear transforms.This paper addresses these limitations, and introduces the unsupervised complex semi-binary matrix factorization (ℂSBMF) as its main contribution. In particular, we show that the exact recovery of source activation sequences from non-intrusive sensor data is intrinsically tied to the presence of problematic phase shifts, the causes of which are detailed. A greedy algorithm is proposed, iteratively resynchronizing sources to converge towards the maximum decomposition of each operation despite these phase shifts. The ℂSBMF is verified and compared to existing techniques on synthetic use cases, then validated on experimental data with signals of different nature. To that occasion, the CAFFEINE dataset for unsupervised time series multi-label classification is introduced.

Assessing Traffic Noise and Its Impact on High-Rise Apartment Buildings Adjacent to an Urban Expressway: A Case Study in Chengdu, China

With rapid urbanization, traffic growth has accelerated in some cities in China. Due to strict urban construction land utilization policies, many high-rise apartment buildings have been constructed adjacent to expressways. To better understand the impact of urban traffic noise on the residents of high residential buildings adjacent to expressways and the differences in noise impacts on different floors, on-site noise monitoring and questionnaires for building residents have been conducted. The characteristics of traffic noise were analyzed based on the measured data, and factors, including time periods and building floors, were considered. According to the results of the questionnaire survey, 56.06% of the male respondents and 54.55% of the female respondents think that the impact of traffic noise on high-rise buildings is “high” or “very high”; 50.53% of the respondents who were in “good” or “very good” condition thought that the traffic noise has a “significant impact” or “very significant impact” on their sleep and daily life. In addition, 25% of respondents living on floors 4–10 and 62.5% of respondents living on floors 11–20 considered the traffic noise to have a “significant impact” or “very significant impact” on their sleep and daily life. The on-site noise monitoring results show that the noise levels (LAeq) outside the windows of the studied buildings remain significantly elevated, with daytime noise on working days ranging from 56 to 70 dB(A), and on weekends ranging from 50 to 65 dB(A). During the four time periods on weekdays from 7 a.m. to 9 a.m., 11 a.m. to 1 p.m., 5 p.m. to 7 p.m., and 10 p.m. to 12 a.m., the average LAeq levels on floors 11–20 are higher than those on floors 4–10 by 4.04 dB(A), 4.92 dB(A), 4.06 dB(A), and 2.67 dB(A), respectively. Similarly, during these time periods on weekends, the levels on floors 11–20 are higher than those on floors 4–10 by 4.96 dB(A), 6.32 dB(A), 5.28 dB(A), and 5.24 dB(A), respectively. This indicates that floors 4–10 of the building experience relatively lower noise levels, while floors 11–20 are subjected to comparatively higher levels of noise disturbance.

Wi-Cyclops: Room-Scale WiFi Sensing System for Respiration Detection Based on Single-Antenna

Recent years have witnessed the emerging development of single-antenna wireless respiration detection that can be integrated into IoT devices with a single transceiver chain. However, existing single-antenna-based solutions are all limited by the short sensing range within 2-4 m due to noise interference, which makes them difficult to be adopted in most room-scale scenarios. To deal with this dilemma, we propose a room-scale, noise-resistance, and accurate respiration monitoring system, named Wi-Cyclops , 1 which captures CSI changes induced by respiratory movements only via one antenna on commercial WiFi devices. To push the limits of effective sensing distance, we innovatively supply a new perspective to review the CSI samples along the sub-carrier dimension. From this dimension, we find that the interrelationship between sub-carriers with different timestamps still shows a high correlation even though the SNR decreases. Based on that, we analyze the noise characteristics along the sub-carrier dimension and correspondingly design a series of denoising schemes. Specifically, we carefully design a PCA-based denoising method to filter out ambient noises. After that, considering the low distribution densities of the AGC-induced noise, we then remove it by optimizing the DBSCAN denoising method with the K-Means-based adaptive radius search. Extensive experiments demonstrate that our system can work effectively in three typical family scenarios. Wi-Cyclops can achieve 98% accuracy even when the person is 7 m away from the transceiver pair. Compared with the start-of-art single-antenna-based approaches in real scenarios, Wi-Cyclops can improve the sensing range from 3 m to 7 m, which can meet the requirements of room-scale respiration monitoring. Additionally, to show the high compatibility with smart home devices, Wi-Cyclops is deployed on seven commercial IoT devices and still achieves a low average absolute error with 0.41 bpm.

National-scale nighttime high-temperature anomalies from Landsat-8 OLI images

Identifying and characterizing the distribution, abundance, and structural composition of national high-temperature anomalies (HTAs) are essential for mitigating natural disasters, regulating industrial emissions, and assessing societal impacts. The Landsat-8 Operational Land Imager (OLI) provides a unique capability for monitoring nighttime HTAs (N-HTAs) at high spatial resolutions without contamination from reflected solar radiation. However, the extraction of N-HTAs from nighttime OLI imagery poses significant challenges due to the interference of bright up-light, especially for smaller and cooler hotspots, and the absence of spectral information on backgrounds complicates the classification of N-HTAs at night. Here, we explored the background noise characteristics of Bands 2–7 in nighttime OLI imagery and developed an object-oriented approach for the enhanced detection of small and cool N-HTAs. Furthermore, we introduced a deep learning model that leverages high-resolution aerial imagery and land cover data to categorize N-HTAs into industrial sources, living sources, vegetation fires, and others. Applying these methodologies to nighttime OLI imagery from 2013 to 2020 allowed us to map and classify N-HTAs across the contiguous United States (CONUS), facilitating the creation of a detailed spatiotemporal inventory of industrial N-HTAs. Our analysis revealed that N-HTAs from living sources dominate the total N-HTAs, with a pronounced holiday effect, and gas flaring activities in oil and gas development areas account for >85 % of industrial N-HTAs. The outcomes of this national-scale investigation underscore the significant potential and practical utility of nighttime OLI imagery for N-HTA monitoring and advocate for expanding the spatial and temporal coverage of nighttime OLI observations globally.

Characteristics of jet noise: A synthesis

It is a distinct privilege to produce this article to honor Professor Tam, the foremost authority on jet aeroacoustics. The fundamental characteristics of jet noise have been studied for 70 years, since the pioneering work of Lighthill in the 1950s. The acoustic analogy, with many variants, has served as the leading theory for nearly 50 years. Many leading researchers in the 1970s formulated theories to interpret the measured trends from subsonic and supersonic jets, using acoustic analogy and flow features as the framework. Quadrupoles, dipoles and monopoles were believed to constitute the sources of noise. The discovery of large-scale organized structures in free shear layers and jets sparked a different avenue of thinking about their importance for noise generation. Tam was the first to clearly demonstrate that these structures are efficient generators of noise and constitute the dominant noise sources, especially in the downstream direction. Now, two schools of thought emerged on the sources of jet noise. Experimental measurements showed that the mean flow as well as the turbulence statistics exhibit a self-similarity in the mixing layer and another similarity in the fully developed jet. Based on these observations, Tam proposed that since noise is generated by the turbulence of the jet, the noise spectra generated by fine-scale and large-scale turbulence should also exhibit self-similarity. By examining a large set of supersonic jet noise data acquired at NASA Langley, Tam offered evidence that the turbulent mixing noise of high-speed jets does consist of two independent self-similar components. In this paper, experimental evidence is compiled from an extensive database that quantifies the effect of several parameters that affect jet spectra. A new scaling method is developed and extended to noise predictions for realistic dual-stream nozzle geometries. The objectives of this paper are to serve as a synthesis of noise characteristics and to focus on application to real-world problems. Results from five different experimental measurements are examined: (1) farfield spectral characteristics; (2) azimuthal and polar correlations in the farfield; (3) correlations of jet turbulence fluctuations and farfield sound; (4) measurement of source distributions with an elliptic mirror; and (5) space-time correlation measurements in the nearfield with a cage array, and nearfield-farfield correlations. Two distinctly different trends are observed in the angular ranges of 50° – ∼120° and ∼120° – 165° for all the parameters investigated with the above five approaches. The salient observations are mutually supporting, and the cumulative weight lends credence to the proposition that there are two distinct sources of turbulent mixing noise.

The role of the damping component deformation in improving the tribological behavior at the high-speed train braking interface

Aiming to the issue of the unclear mechanism and lack of design basis for the influence of the deformation of the damping component installed in the brake pads of high-speed train, various perforated configurations were implemented on the damping component to obtain different typical deformation. The damping component was then installed on the back of the friction block and the tests on friction braking were performed using a test rig designed to assess high-speed train braking performance. A finite element model (FEM) was established for wear simulation to obtain block worn surfaces that closely matched the test results, succeeded by conducting both complex eigenvalue analysis (CEA) and implicit dynamic analysis (IDA). The findings indicate that the perforated configurations effectively provided damping components with different deformation capabilities, and increasing the deformation weakened its damping effect on vibration. The damping component deformation not only significantly affected the non-uniform wear on the block surface but also altered the wear patterns. Inappropriate deformation exacerbated the degree of non-uniform wear, leading to complex abrasive wear and adhesive wear, and added to the interface damage. Therefore, the damping component deformation significantly altered the friction-induced vibration and noise (FIVN) characteristics. Enhancing the deformation disrupted the continuity of FIVN but intensified its strength. The major reason is that the damping component deformation determined the time length of its compression and rebound, and inappropriate deformation significantly amplifying the oscillation intensity of contact area and friction force. This posed a challenge in establishing stable contact at the braking interface, ultimately leading to the generation of intense FIVN. Installation of damping components in brake pad to improve interface contact characteristics and suppress FIVN should carefully consider the structural of the braking system, and design damping components with appropriate deformation. Otherwise, it may lead to opposite effects.

Study on structure-borne noise characteristics of long-span steel truss bridge in urban rail transit

Study on structure-borne noise characteristics of long-span steel truss bridge in urban rail transit

Reduction of relative intensity noise in a diamond Raman laser

The relative intensity noise (RIN) characteristics of a continuous-wave diamond Raman laser are investigated for the first time. The results reveal the parasitic stimulated Brillouin scattering (SBS) that usually occurred with higher-order spatial modes in the diamond Raman resonator is a pivotal factor impacting the Raman longitudinal modes and deteriorating the RIN level. The diamond Raman laser automatically switches to single-longitudinal-mode operation and the RIN level is significantly decreased in the frequency range of 200 Hz to 1 MHz after the parasitic SBS is effectively suppressed through inserting a spatial aperture or a χ(2) nonlinear crystal into the cavity. Due to the introduction of additional nonlinear loss to the high intensity Raman fluctuations and the non-lasing spontaneous Raman modes, the χ(2) nonlinear crystal enables better performance in the RIN-level reduction compared to the spatial aperture which can only achieve SBS inhibition. The RIN reduction routes are well suited for various crystalline Raman media to achieve high power and low intensity noise laser at different wavelengths.

Measurement of sub-Poissonian shot noise in a quantum cascade detector

In a Quantum Cascade Detector, photocurrent is generated by the absorption of infrared and terahertz radiation in the quantum-well-based modules arranged in series. Consequently, the current responsivity is by construction inversely proportional to the number of cascading modules. Upon absorption of a photon, the electron travels through only a single period of the detector, with a mean free path corresponding to the period length. Therefore, the shot noise power density is expected to decrease by the same factor under sufficiently high illumination, reflecting the same inverse relationship with the number of cascading modules. This phenomenon leads to sub-Poissonian noise characteristics. We experimentally observe this effect in a 90-period Quantum Cascade Detector operating at 4.5 μm, confirming a reduction in the shot noise contribution by the anticipated Fano factor of 1/90. This measurement underscores the suitability of these detectors for coherent detection scenarios, particularly where shot noise dominates.

Combined Cubature Kalman and Smooth Variable Structure Filtering Based on Multi-Kernel Maximum Correntropy Criterion for the Fully Submerged Hydrofoil Craft

This paper introduces a novel filter algorithm termed as an MKMC-CSVSF which combined square-root cubature Kalman (SR-CKF) and smooth variable structure filtering (SVSF) under multi-kernel maximum correntropy criterion (MKMC) for accurately estimating the state of the fully submerged hydrofoil craft (FSHC) under the influence of uncertainties and multivariate heavy-tailed non-Gaussian noises. By leveraging the precision of the SR-CKF and the robustness of the SVSF against system uncertainties, the MKMC-CSVSF integrates these two methods by introducing a time-varying smooth boundary layer along with multiple fading factors. Furthermore, the MKMC is introduced for the adjustment of kernel bandwidths across different channels to align with the specific noise characteristics of each channel. A fuzzy rule is devised to identify the appropriate kernel bandwidths to ensure filter accuracy without undue complexity. The precision and robustness of state estimation in the face of heavy-tailed non-Gaussian noises are improved by modifying the SR-CKF and the SVSF using a fixed-point approach based on the MKMC. The experimental results validate the efficacy of this algorithm.

Impact of pump-phase noise in PPLN-based optical parametric amplifier and wavelength converter for digital coherent transmission

Wideband signal amplification and optical signal processing with a high gain using an optical parametric amplifier based on a periodically poled LiNbO3 (PPLN) waveguide is attractive for constructing wideband optical fiber networks. We experimentally investigate the transfer characteristics of the phase noise of a pump laser in χ(2)-based optical parametric amplification and wavelength conversion on the basis of second-harmonic-generation and differential-frequency-generation processes. We also evaluate the effect of the transferred phase noise on signal quality in dispersion-unmanaged digital coherent fiber transmission systems. We show that the phase noise is transferred only to the wavelength-converted idler and does not affect the amplified signal even by using a pump laser with a MHz-order linewidth. We also show that the phase noise transferred to the idler light can have a similar impact on signal quality as equalization-enhanced phase noise (EEPN) in digital coherent transmission. The signal penalty including EEPN was evaluated with several pump lasers and at symbol rates of 32, 64, and 96 Gbaud. We also propose a method of using correlated pump lights between a wavelength converter pair to cancel out the transfer of phase noise.

DeepNRMS: Unsupervised deep learning for noise-robust CO2 monitoring in time-lapse seismic images

Monitoring stored [Formula: see text] in carbon capture and storage projects is crucial for ensuring safety and effectiveness. We introduce DeepNRMS, a novel noise-robust method that effectively handles time-lapse noise in seismic images. The DeepNRMS leverages unsupervised deep learning to acquire knowledge of time-lapse noise characteristics from preinjection surveys. By using this learned knowledge, our approach accurately discerns [Formula: see text]-induced subtle signals from the high-amplitude time-lapse noise, ensuring fidelity in monitoring while reducing costs by enabling sparse acquisition. We evaluate our method using synthetic data and field data acquired in the Aquistore project. In the synthetic experiments, we simulate time-lapse noise by incorporating random near-surface effects in the elastic properties of the subsurface model. We train our neural networks exclusively on preinjection seismic images and subsequently predict [Formula: see text] locations from postinjection seismic images. In the field data analysis from Aquistore, the images from preinjection surveys are used to train the neural networks with the characteristics of time-lapse noise, followed by identifying [Formula: see text] plumes within two postinjection surveys. The outcomes demonstrate the improved accuracy achieved by DeepNRMS, effectively addressing the strong time-lapse noise.

Intelligent and Automated Product Quality Assurance System using Machine Learning

The paper discusses the use of advanced technologies like the Internet of Things (IoT) and cyber-physical systems (CPS) to enhance product quality assurance in production industries. It proposes an intelligent approach using machine learning and termed the model as ETLViT, which integrates transfer learning models and vision transformers (ViT). The proposed model is a three-layered architecture integrated with a cyber-physical system (CPS). The first layer collects images of the processed products using a CPS, while the second layer transmits the data to a cloud server. The third layer analyzes the input image and predicts any faults in the product, with results also stored in the cloud server. The learning process involves noise removal, feature extraction, and classification, which are performed simultaneously by the ensemble model called ETLViT, designed using transfer learning and vision transformers. The result was presented with transfer learning models, ensemble model with machine learning and ETLViT. Among all of them, ETLViT outperforms the best.

Influence of start-up modes on the noise characteristics of mixed-flow pump during start-up process

The evolutions of noise characteristics of the pump have significant implications for monitoring the stability of the mixed-flow pump (MFP) start-up process. To investigate the variations in noise characteristics under different start-up modes and start-up times, the noise, shaft vibration and pressure pulsation of MFP under linear and nonlinear start-up modes were collected by an acoustic and vibration experimental system. Subsequently, an analysis was conducted based on acoustic signal processing methods. The results show that the high-amplitude sound pressure under different start-up modes appears in the later stage of start-up, and the time for the high-amplitude noise region to appear is the shortest under the concave exponential (CCEF) start-up mode. The shock noise under the CCEF start-up mode is relatively small, indicating that this start-up mode is conducive to avoiding severe mechanical collisions and shocks during the start-up process. The highest A-weighted sound pressure level (SPL) of the noise under different start-up modes appears at the center frequency of 250 Hz, and the CCEF start-up mode brings about the smallest total effective SPL. Adopting the CCEF start-up mode and extending the start-up time is beneficial to avoid severe mechanical collisions, reduce the SPL in each frequency band, and weaken the continuous impact of noise on the environment during the MFP start-up process. In addition, CCEF start-up modes can effectively reduce the tone of the noise during the start-up of the MFP.

Acoustic wind tunnel experimental study of aerodynamic noise distribution in high-rise buildings

Wind-induced aerodynamic disturbance has become a major noise source affecting residents in super high-rise buildings. To investigate the distribution characteristics of aerodynamic noise on high-rise building surfaces in strong wind conditions and the impact of wind speed and direction, this paper pioneers the application of the acoustic wind tunnel method based on microphone array localization to the experimental study of the surface acoustic pressure field of a high-rise building model, the far-field noise spectrum of which is obtained using the microphone array sensor. The beamforming calculation method was used to obtain the spectrum and localization cloud. By comparing the spectra and cloud diagrams under different process conditions and based on existing building wind pressure reports, it is concluded that: the noise source is always located at the height of the connecting corridor where there is large wind pressure pulsation in the two building models; the aerodynamic noise is more pronounced around building gaps where the airflow is more disturbed and also in the oncoming wind trailing direction where the wind pressure pulsation is more pronounced. An increase in wind speed increased noise source amplitude from 30 dB to 60 dB, and the location of the noise source shifted backwards to the tail of the incoming wind. This paper shows that the acoustic wind tunnel method based on microphone array localization technology can obtain the aerodynamic noise intensity on the surfaces of super high-rise buildings and reasonably predict the location of the sound source. This article puts forward suggestions to reduce noise in high-rise buildings based on experimental results.

Analysis of submarine seismic ambient noise data in Bohai Sea and Yellow Sea and its application

Submarine seismic ambient noise imaging combines current marine and on-land seismic detection technologies. Based on data from several broadband shallow-sea type ocean bottom seismometers (SOBSs) deployed in the Bohai Sea and north Yellow Sea, this paper analyzes the characteristics of the submarine seismic ambient noise and explores the theory, technology, method and application of the submarine seismic ambient noise imaging by using the single point horizontal and vertical spectral ratio method (HVSR). The observations yield the following results: 1) Submarine seismic ambient noise has consistent and constant energy, making it an appropriate passive seismic source for submarine high frequency surface wave investigation. 2) Using the HVSR approach, a single three-component OBS could differentiate between the basement and sediments. Array seismic observation could be utilized to extract the frequency dispersion curve and invert it to obtain the velocity structure for more accurate stratification. 3) The SOBS we use is suitable for submarine surface wave exploration. 4) Tomography results with greater resolution and deeper penetration could be obtained by combining the use of active and passive sources in a simultaneous inversion of the HVSR and frequency dispersion curve. Seamless land-to-ocean seismic research can be accomplished with submarine seismic ambient noise imaging technologies.

Real-time Abnormal Detection of GWAC Light Curve based on Wavelet Transform Combined with GRU-Attention

Nowadays, astronomy has entered the era of Time-Domain Astronomy, and the study of the time-varying light curves of various types of objects is of great significance in revealing the physical properties and evolutionary history of celestial bodies. The Ground-based Wide Angle Cameras telescope, on which this paper is based, has observed more than 10 million light curves, and the detection of anomalies in the light curves can be used to rapidly detect transient rare phenomena such as microgravity lensing events from the massive data. However, the traditional statistically based anomaly detection methods cannot realize the fast processing of massive data. In this paper, we propose a Discrete Wavelet (DW)-Gate Recurrent Unit-Attention (GRU-Attention) light curve warning model. Wavelet transform has good effect on data noise reduction processing and feature extraction, which can provide richer and more stable input features for a neural network, and the neural network can provide more flexible and powerful output model for wavelet transform. Comparison experiments show an average improvement of 61% compared to the previous pure long-short-term memory unit (LSTM) model, and an average improvement of 53.5% compared to the previous GRU model. The efficiency and accuracy of anomaly detection in previous paper work are not good enough, the method proposed in this paper possesses higher efficiency and accuracy, which incorporates the Attention mechanism to find out the key parts of the light curve that determine the anomalies. These parts are assigned higher weights, and in the actual anomaly detection, the star is detected with 83.35% anomalies on average, and the DW-GRU-Attention model is compared with the DW-LSTM model, and the detection result f1 is improved by 5.75% on average, while having less training time, thus providing valuable information and guidance for astronomical observation and research.