Search

The picking efficiency of seismic first breaks (FBs) has been greatly accelerated by deep learning (DL) technology. However, the picking accuracy and efficiency of DL methods still face huge challenges in low signal-to-noise ratio (SNR) situations. To address this issue, we propose a regression approach to pick FBs based on bidirectional long short-term memory (BiLSTM) neural network by learning the implicit Eikonal equation of 3D inhomogeneous media with rugged topography in the target region. We employ a regressive model that represents the relationships among the elevation of shots, offset and the elevation of receivers with their seismic traveltime to predict the unknown FBs, from common-shot gathers with sparsely distributed traces. Different from image segmentation methods which automatically extract image features and classify FBs from seismic data, the proposed method can learn the inner relationship between field geometry and FBs. In addition, the predicted results by the regressive model are continuous values of FBs rather than the discrete ones of the binary distribution. The picking results of synthetic data shows that the proposed method has low dependence on label data, and can obtain reliable and similar predicted results using two types of label data with large differences. The picking results of 9380 shots for 3D seismic data generated by vibroseis indicate that the proposed method can still accurately predict FBs in low SNR data. The subsequent stacked profiles further illustrate the reliability and effectiveness of the proposed method. The results of model data and field seismic data demonstrate that the proposed regression method is a robust first-break picker with high potential for field application. [ABSTRACT FROM AUTHOR]

Magnetic resonance sounding is a geophysical method for quantitatively determining the state for groundwater storage that has gained international attention in recent years. However, the practical acquisition of magnetic resonance sounding signals, which are on the nanovolt scale, is susceptible to various types of interference, such as power-line harmonics, random noise, and spike noise. Such interference can degrade the quality of magnetic resonance sounding signals and, in severe cases, be completely drowned out by noise. This paper introduces an adaptive Gaussian filtering algorithm that is well-suited for handling intricate noise signals due to its adaptive solving characteristics and iterative sifting approach. Notably, the algorithm can process signals without relying on prior knowledge. The adaptive Gaussian filtering algorithm is applied for the envelope extraction of noisy magnetic resonance sounding signals, and the reliability and effectiveness of the method are rigorously validated. The simulation results reveal that, even under strong noise interference (with original signal-to-noise ratios ranging from −7 dB to −25 dB), the magnetic resonance sounding signal obtained after algorithmic processing is compared to the ideal signal, with 16 sets of data statistics, and the algorithm ensures an initial amplitude uncertainty within 4nV and restricts the uncertainty of the relaxation time within a 6 ms range. The signal-to-noise ratio can be boosted by up to 53 dB. The comparative assessments with classical algorithms such as empirical mode decomposition and the harmonic modeling method confirm the superior performance of the adaptive Gaussian filtering algorithm. The processing of the field data also fully proved the practical application effects of the algorithm. [ABSTRACT FROM AUTHOR]

Insurance pricing systems should fulfill the auto-calibration property to ensure that there is no systematic cross-financing between different price cohorts. Often, regression models are not auto-calibrated. We propose to apply isotonic recalibration to a given regression model to restore auto-calibration. Our main result proves that under a low signal-to-noise ratio, this isotonic recalibration step leads to an explainable pricing system because the resulting isotonically recalibrated regression function has a low complexity. [ABSTRACT FROM AUTHOR]

In ocean acoustic fields, extracting the normal-mode interference spectrum (NMIS) from the received sound intensity spectrum (SIS) plays an important role in waveguide-invariant estimation and underwater source ranging. However, the received SIS often has a low signal-to-noise ratio (SNR) owing to ocean ambient noise and the limitations of the received equipment. This can lead to significant performance degradation for the traditional methods of extracting NMIS at low SNR conditions. To address this issue, a new deep neural network model called SSANet is proposed to obtain NMIS based on unrolling the traditional singular spectrum analysis (SSA) algorithm. First, the steps of embedding and singular value decomposition (SVD) in SSA is achieved by the convolutional network. Second, the grouping step of the SSA is simulated using the matrix multiply weight layer, ReLU layer, point multiply weight layer and matrix multiply weight layer. Third, the diagonal averaging step was implemented using a fully connected network. Simulation results in canonical ocean waveguide environments demonstrate that SSANet outperforms other traditional methods such as Fourier transform (FT), multiple signal classification (MUSIC), and SSA in terms of root mean square error, mean absolute error, and extraction performance. [ABSTRACT FROM AUTHOR]

Conducted electromagnetic emissions from interconnecting cables in computer systems can lead to internal information leakage and cause information security problems. However, unintentionally leaked EM signals are characterized by low signal-to-noise ratio and random noise, making it difficult to recover the original signal. In this paper, we propose a denoising model (S-DnCNN) based on an improved DnCNN to better recover the original signal. The network structure consists of three parts: feature mapping generation, low-dimensional feature extraction, and original reconstruction. To improve the noise extraction capability, we use Leaky ReLU as the activation function of the CNN, and introduce a residual structure and a convolutional attention module. The residual structure uses residual hopping to implicitly remove potentially clean images by hidden layer operations, thus training noisy data to recover clean data. We construct a one-dimensional selective convolution kernel (SKConv1d) and fuse it with local paths to form a feature extraction network, which improves the performance of the network. The experimental results show that our proposed method can preserve the details in the effective signal during denoising and shows good generalization to complex SNR data. [ABSTRACT FROM AUTHOR]

In low signal-to-noise ratio (SNR) conditions, to address the problem of poor speech enhancement effect of traditional neural networks, this paper combines Convolution Recurrent Neural Network (CRN) with Gated Linear Units (GLU) to extract speech characteristics. Meanwhile, considering the advantages of the Empirical Mode Decomposition (EMD) algorithm, this paper put forward a model for speech enhancement that is based on Adaptive Mean Median Empirical Modal Decomposition and Multilayer Gated Convolutional Recurrent Neural Networks (ME-MGCRN). The model first uses the improved EMD algorithm to split the original noisy speech into low-frequency and high-frequency components. After that, the noisy speech is denoised using noise correlation and completes the feature extraction process. Finally, the extracted feature information is fed into the MGCRN network to realize speech enhancement further. This paper analyzes the performance of the Adaptive Mean Median Empirical Modal Decomposition algorithm using the LibriSpeech ASR dataset. This paper compares and investigates the enhanced performance of the ME-MGCRN model with the baseline model and the traditional model in terms of evaluation metrics such as the perceptual evaluation of speech quality (PESQ) and short-time objective intelligence (STOI). The research indicates that the ME-MGCRN model proposed in this paper has improved in the evaluation indexes, such as PESQ and STOI, respectively, by 0.22% and 5.6% than the baseline model and the traditional model, and the speech enhancement effect is better when Huber is selected as the loss function.

In the window function spectrum of the low signal-to-noise ratio photon Doppler signal after the short-time Fourier transform, the weak beat frequency cannot be obtained by extracting the maximum amplitude spectrum frequency of each window because the amplitude of the noise spectrum exceeds the weak beat frequency. In this article, the value of the beat frequency is first estimated by the Kalman filter. Then, a multiple analytical bandpass filter is constructed to refine the spectrum of the high noise signal with the value of the beat frequency estimate as the spectrum band center. This technique extracts only the beat frequencies from the narrowband refinement spectrum by removing the interference spectrum band range. We use this technique to process the photon Doppler signals from the tiny high-speed flying fragments explosion experiment. After data processing, the beat frequency value heavily affected by high-frequency noise can reduce the error by up to 64.9%. The beat frequency value of the low noise signal can be accurate to 106 Hz, equivalent to 0.775 m/s after velocity demodulation. This method fully considers the positioning and protection of the beat frequency characteristics. It makes weak beat frequencies more obvious in the refined narrowband spectrum without changing the signal amplitude. This article describes a method for extracting the weak beat frequency of the photon Doppler signal or a signal optimization algorithm for needing high-precision beat frequencies in a test environment.

Seismic data collected from desert areas contain a large amount of low-frequency random noise with similar waveforms to the effective signals. The complex noise characteristics make it difficult to effectively identify and recover seismic signals, which will adversely affect subsequent seismic data processing and imaging. In order to recover the complex seismic events from low-frequency random noise, we propose an attention mechanism guided deep convolutional autoencoder network (ADCAE) to assign different importance to different features at different spatial position. In ADCAE, an attention module (AM) is connected to the deep convolutional autoencoder network (DCAE) with soft-thresholded symmetric skip connection that helps to enhance the ability of feature extraction. By combining the global features of the input data and the output local features of DCAE, AM generates an attention weight matrix, which assigns different weights to the features associated with the seismic events and random noise during the training process. In this way, AM can guide the update of the target gradient, thus retains the complex structure of the seismic events in the denoised results and improves the training efficiency of the model. The ADCAE is applied to the synthetic data and field seismic data, and denoised results show that ADCAE has achieved satisfactory denoising performance in signals recovery and low-frequency random noise suppression at the low signal-to-noise ratio. [ABSTRACT FROM AUTHOR]

Least squares (LS) is a commonly used pilot-based channel estimation algorithm in orthogonal frequency division multiplexing (OFDM) systems. This algorithm is simple and easy to implement because of its low computation complexity. However, it has poor performance, especially at low signal-to-noise ratio (SNR). To solve this problem, an improved LS channel estimation method based on convolutional neural network (CNN) is proposed on the basis of analyzing the traditional LS channel estimation methods. A channel estimation compensated network is designed based on CNN, which can solve the problem of performance degradation of the mean square error (MSE) through the online and offline modules. By designing the input-output relations, training data set, and testing data set, a CNN network is iteratively trained to learn the relevant features of the channels, so that the traditional LS estimation value can be corrected to improve the accuracy. Simulation results show that the proposed method can achieve better performance in bit error rate (BER) and MSE, compared with the traditional channel estimation methods.

Magnetic resonance sounding is a geophysical method for quantitatively determining the state for groundwater storage that has gained international attention in recent years. However, the practical acquisition of magnetic resonance sounding signals, which are on the nanovolt scale, is susceptible to various types of interference, such as power-line harmonics, random noise, and spike noise. Such interference can degrade the quality of magnetic resonance sounding signals and, in severe cases, be completely drowned out by noise. This paper introduces an adaptive Gaussian filtering algorithm that is well-suited for handling intricate noise signals due to its adaptive solving characteristics and iterative sifting approach. Notably, the algorithm can process signals without relying on prior knowledge. The adaptive Gaussian filtering algorithm is applied for the envelope extraction of noisy magnetic resonance sounding signals, and the reliability and effectiveness of the method are rigorously validated. The simulation results reveal that, even under strong noise interference (with original signal-to-noise ratios ranging from −7 dB to −25 dB), the magnetic resonance sounding signal obtained after algorithmic processing is compared to the ideal signal, with 16 sets of data statistics, and the algorithm ensures an initial amplitude uncertainty within 4nV and restricts the uncertainty of the relaxation time within a 6 ms range. The signal-to-noise ratio can be boosted by up to 53 dB. The comparative assessments with classical algorithms such as empirical mode decomposition and the harmonic modeling method confirm the superior performance of the adaptive Gaussian filtering algorithm. The processing of the field data also fully proved the practical application effects of the algorithm.

The traditional BP-PID control algorithm uses the gradient descent method to solve, which has problems such as slow convergence speed, easy trapping in local extremum, and performance degradation under low signal-to-noise ratio (LSNR) conditions. In order to solve the above problems, a BP PID belt conveyor speed control method (ImGSAA-BP-PID) based on improved genetic simulated annealing algorithm (ImGSAA) optimization is proposed. Firstly, the values of crossover and mutation probabilities are correlated with the iteration time. The inverse cosine function is introduced to enhance the dynamic adjustment and nonlinear change adaptability of GSAA. Secondly, by weighting the traditional Metropolis criterion, a weighted Metropolis criterion is proposed to modify the new population individuals and improve the noise robustness of genetic simulated annealing algorithm (GSAA). Finally, ImGSAA is used to optimize the initial parameters of BP-PID, automatically determining the optimal parameter combination for BP-PID. It improves its real-time parameter tuning, control precision, and adaptability to the LSNR environment. The experimental results show the following points. ① ImGSAA only needs 11 iterations to converge, indicating that optimizing the GSAA using the proposed improved crossover and mutation strategies and weighted Metropolis criteria can effectively improve the convergence speed and real-time performance of the algorithm. ② The control error of ImGSAA-BP-PID is −0.468 5-0.572 3 m/s, which is 224.88%, 104.07%, and 38.33% higher than the control methods based on genetic algorithm (GA)-BP PID, particle swarm optimization (PSO)-BP PID, and GSAA-BP-PID, respectively. ③ The performance of ImGSAA is least affected by LSNR. It converges to the global optimal solution after 15 iterations, which has strong noise robustness. ④ Under LSNR conditions, the average control error of ImGSAA-BP-PID decreases by 3.54%. The control performance is significantly better than GA-BP-PID, PSO-BP-PID, and GSAA-BP-PID, which better meets the practical engineering application requirements.

Least squares (LS) is a commonly used pilot-based channel estimation algorithm in orthogonal frequency division multiplexing (OFDM) systems. This algorithm is simple and easy to implement because of its low computation complexity. However, it has poor performance, especially at low signal-to-noise ratio (SNR). To solve this problem, an improved LS channel estimation method based on convolutional neural network (CNN) is proposed on the basis of analyzing the traditional LS channel estimation methods. A channel estimation compensated network is designed based on CNN, which can solve the problem of performance degradation of the mean square error (MSE) through the online and offline modules. By designing the input-output relations, training data set, and testing data set, a CNN network is iteratively trained to learn the relevant features of the channels, so that the traditional LS estimation value can be corrected to improve the accuracy. Simulation results show that the proposed method can achieve better performance in bit error rate (BER) and MSE, compared with the traditional channel estimation methods. [ABSTRACT FROM AUTHOR]

Aiming at the nonlinear and low signal-to-noise ratio characteristics of mine hoist vibration signal in complex environments, a mine hoist vibration signal fusion denoising algorithm based on complete EEMD with adaptive noise (CEEMDAN) and adaptive wavelet threshold is proposed. Firstly, the CEEMDAN algorithm is used to decompose the noisy mine hoist vibration signal to obtain the intrinsic mode component (IMF) and the residual. The IMF component is judged for high and low frequency. The t-test method is used to test whether the mean value is significantly different from 0. The IMF component which tends to 0 is the high-frequency component, and the IMF component which is significantly different from 0 is the low-frequency component. Secondly, the appropriate wavelet basis function and decomposition level are selected. The high-frequency IMF component is denoised by using the adaptive wavelet threshold method. Finally, the processed high-frequency IMF components and the unprocessed low-frequency IMF components are reconstructed with the residuals to obtain the de-noised vibration signal from the fusion algorithm. The CEEMDAN denoising method, CEEMD-wavelet threshold combined denoising method, CEEMDAN-wavelet threshold combined denoising method and CEEMDAN-adaptive wavelet threshold fusion denoising method are used to denoise the simulated signal respectively. The results show the following points. ① The signal denoised by the CEEMDAN-adaptive wavelet threshold fusion denoising method is similar to the original signal in local waveform features and signal peak values. Some features of the signal waveform have been restored well. The feature information of the original signal has been well preserved in the process of denoising. ② The composite evaluation index H is used as the objective evaluation standard. The H value of the CEEMDAN-adaptive wavelet threshold fusion denoising method is the smallest. This shows that the denoising effect of the fusion denoising algorithm for the simulation signal is better than that of other denoising methods. The experiment is carried out on the running mine hoist in a mine in Heilongjiang Province. The results show the following points. ① The db4 wavelet basis function is used to decompose the noisy IMF component in four layers. The signal de-noised by CEEMDAN-adaptive wavelet threshold fusion de-noising method is smooth. Some waveform features of the signal have also been restored well. While removing the noise, the feature information of the original signal has been retained to the greatest extent. ② In the actual mine hoist vibration signal denoising process, the CEEMDAN-adaptive wavelet threshold fusion denoising method has the smallest H value and the best denoising effect.

Most of the existing acoustic imaging studies in reverberant field ignore the influence of signal-to-noise ratio. As a result, commonly used beamforming algorithms in reverberant backgrounds have poor imaging accuracy for low signal-to-noise ratio sound sources. In response to that problem, an improved adaptive beamforming algorithm called SC-DAMAS is put forward in this paper. The algorithm replaces the free-field Green's function with the impulse response function, making the algorithm more suitable for acoustic imaging of low signal-to-noise ratio in a reverberant environment. Besides, the comparative simulation results with the conventional beamforming method and orthogonal matching pursuit algorithm-based DAMAS, as well as sound source acoustic imaging experiments are carried out to analyze its effectiveness. It is indicated that, in the reverberation field, the SC-DAMAS has no obvious sidelobes and achieves higher positioning accuracy for acoustic imaging of low signal-to-noise ratio sound source than the abovementioned counterparts, and its imaging test result is consistent with the actual situation, which verifies the effectiveness of the algorithm. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Mine Automation is the property of Industry & Mine Automation Editorial Department and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Because the traditional pulse laser high-precision ranging method has the problems of low ranging accuracy and long ranging time, a pulse laser high-precision ranging algorithm under low signal-to-noise ratio is proposed. Based on the analysis of the basic principle of pulse laser ranging, the pulse laser signal under low signal-to-noise ratio is obtained by modeling the pulse laser echo analog signal waveform. The wavelet denoising method of improved threshold function is used to denoise the pulse laser signal under low signal-to-noise ratio. According to the processing results, the pulse laser high-precision ranging is carried out through sinusoidal amplitude time conversion. The simulation results show that the proposed algorithm has high precision and short ranging time.

With the increasing demand for the wide-area refined detection of aircraft targets, remote sensing cameras have adopted an ultra-large area-array detector as a new imaging mode to obtain broad width remote sensing images (RSIs) with higher resolution. However, this imaging technology introduces new special image degradation characteristics, especially the weak target energy and the low signal-to-noise ratio (SNR) of the image, which seriously affect the target detection capability. To address the aforementioned issues, we propose an aircraft detection method for RSIs with low SNR, termed L-SNR-YOLO. In particular, the backbone is built blending a swin-transformer and convolutional neural network (CNN), which obtains multiscale global and local RSI information to enhance the algorithm's robustness. Moreover, we design an effective feature enhancement (EFE) block integrating the concept of nonlocal means filtering to make the aircraft features significant. In addition, we utilize a novel loss function to optimize the detection accuracy. The experimental results demonstrate that our L-SNR-YOLO achieves better detection performance in RSIs than several existing advanced methods. [ABSTRACT FROM AUTHOR]

The traditional speech enhancement algorithm based on generative adversarial networks (SEGAN) enhances speech in the time domain,and completely ignores the distribution of speech samples in frequency domain.Under the condition of low signal-to-noise ratio,the speech signal will be submerged in noise,and the time-domain distribution information of noisy speech is difficult to capture.Therefore,the enhancement performance of SEGAN will drop sharply,and the speech quality and speech intelligibility of its enhanced speech are very low.To solve this problem,this paper proposes a speech enhancement algorithm (time-frequency domain SEGAN,TFSEGAN) based on time-frequency domain generation confrontation network.TFSEGAN adopts the model structure of the time-frequency domain dual discriminator,and a time-frequency L1 loss function.The input of time domain discriminator is time domain feature of the speech sample,and the input of frequency domain discriminator is frequency domain feature of the speech sample.In the training process,time-domain discriminator uses the time-domain distribution information of speech sample as the criterion,and frequency-domain discriminator uses the frequency-domain distribution information of the speech sample as the criterion.Under the action of two discriminators,the generator of TFSEGAN could simulta-neously learn the distribution rules and information of speech samples in time domain and frequency domain.Experiments prove that,compared with SEGAN,the speech quality and intelligibility of TFSEGAN improve by about 17.45% and 11.75% respectively at low signal-to-noise ratio.

Copyright of Journal of Mine Automation is the property of Industry & Mine Automation Editorial Department and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The Shack-Hartmann wavefront sensor produces incorrect wavefront measurements when some sub-spots are weak and missing. In this paper, a method is proposed to predict the centroids of these sub-spots for the Shack-Hartmann wavefront sensor based on the deep neural network. Using the centroid information of present sub-spots, the method is able to predict the absent sub-spots’ positions. The feasibility and effectiveness of this method are verified by a large number of numerical simulations. The method is applied to wavefront measurement of light with non-uniform near-field intensity. The simulation results show that the proposed method is of great help to improve the measurement accuracy and the Strehl ratio of the focal spot. For wavefronts outside of the training sample, the proposed method shows good generalization and adaptability. In addition, the experiment results demonstrate that the proposed method can predict the missing sub-spots’ centroid displacements accurately even though a large proportion of sub-spot is lost randomly.

Under the condition of low signal-to-noise ratio, the target detection performance of radar decreases, which seriously affects the tracking and recognition for the long-range small targets. To solve it, this paper proposes a target detection algorithm using convolutional neural network to process graphically expressed range time series signals. First, the two-dimensional echo signal was processed graphically. Second, the graphical echo signal was detected by the improved convolutional neural network. The simulation results under the condition of low signal-to-noise ratio show that, compared with the multi-pulse accumulation detection method, the detection method based on convolutional neural network proposed in this paper has a higher target detection probability, which reflects the effectiveness of the method proposed in this paper. [ABSTRACT FROM AUTHOR]

Aiming at the problem that the traditional deep learning algorithms have unsatisfactory speech enhancement effect under low Signal-to-Noise Ratio (SNR) conditions, this paper improves Deep Neural Network (DNN) and Multiresolution Residual U Network (MultiResU_Net) to discuss the speech enhancement effect. In this paper, the Minimum Mean-Square Error Short-Time Spectral Amplitude Estimator (MMSE-STSA) is used to reduce the noise of the Multi-Resolution Cochleagram (MRCG) features that including the local and global characteristics of speech and using the improved features Improved Multi-Resolution Cochleagram (I-MRCG) as the input of Skip-DNN network to build Skip-DNN speech enhancement model. The short-time Fourier transformed features of the speech signal at low SNR are used as the input of MultiResU_Net network and the output features of residual path and upsampling are rearranged by the hybrid channel to build an improved MultiResU_Net speech enhancement model. Using LibriSpeech ASR corpus, under different SNR conditions, analyze the speech enhancement effects of different network models and different features as the input of the Skip-DNN model. The experimental results show that when the SNR is -5dB, the perceptual evaluation effect of the average speech quality is better that I-MRCG is used as the feature input of the Skip-DNN model. The average short-time objective intelligibility (STOI), the average perceptual evaluation of speech quality (PESQ) and Source-to-Distortion Ratio (SDR) of the improved MultiResU_Net speech enhancement model are higher than other models. It can be seen that both the improved Skip-DNN and MultiResU_Net proposed in this paper have better speech enhancement effect, especially the improved MultiResU_Net network is more suitable for speech enhancement at low SNR. [ABSTRACT FROM AUTHOR]

With the increasing demand for the wide-area refined detection of aircraft targets, remote sensing cameras have adopted an ultra-large area-array detector as a new imaging mode to obtain broad width remote sensing images (RSIs) with higher resolution. However, this imaging technology introduces new special image degradation characteristics, especially the weak target energy and the low signal-to-noise ratio (SNR) of the image, which seriously affect the target detection capability. To address the aforementioned issues, we propose an aircraft detection method for RSIs with low SNR, termed L-SNR-YOLO. In particular, the backbone is built blending a swin-transformer and convolutional neural network (CNN), which obtains multiscale global and local RSI information to enhance the algorithm’s robustness. Moreover, we design an effective feature enhancement (EFE) block integrating the concept of nonlocal means filtering to make the aircraft features significant. In addition, we utilize a novel loss function to optimize the detection accuracy. The experimental results demonstrate that our L-SNR-YOLO achieves better detection performance in RSIs than several existing advanced methods.

Copyright of Systems Engineering & Electronics is the property of Journal of Systems Engineering & Electronics Editorial Department and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Speech enhancement in the frequency-domain can achieve a good noise reduction effect and improve speech communication quality under the condition of a high signal-to-noise ratio. However, speech enhancement performance in frequency-domain decreases significantly under low SNR conditions. To address this problem, this paper applied a mask based on speaker embedding to a frequency-domain speech enhancement network, and used the a priori information of speaker embedding to improve the network's ability to distinguish between speakers and noise. To further improve the performance of speech enhancement under low SNR conditions, this paper proposes a speech enhancement network with the introduction of mapping-based speaker embedding to avoid the problem of speech distortion, which may be caused by using the mask scheme. The experimental results show that the speech enhancement network with the introduction of mapping-based speaker embedding performs better in the low signal-to-noise ratio condition when the same speaker information is applied, especially in improving the speech distortion. Compared with the mask- based network, the mapping-based network achieves a relative improvement of 6.40%, 1.46% and 24.84% in the three metrics of PESQ, STOI and SSNR, respectively. respectively. respectively. [ABSTRACT FROM AUTHOR]

Copyright of Journal of National University of Defense Technology / Guofang Keji Daxue Xuebao is the property of NUDT Press and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Journal of Signal Processing is the property of Journal of Signal Processing and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Under the condition of low signal-to-noise ratio, the target detection performance of radar decreases, which seriously affects the tracking and recognition for the long-range small targets. To solve it, this paper proposes a target detection algorithm using convolutional neural network to process graphically expressed range time series signals. First, the two-dimensional echo signal was processed graphically. Second, the graphical echo signal was detected by the improved convolutional neural network. The simulation results under the condition of low signal-to-noise ratio show that, compared with the multi-pulse accumulation detection method, the detection method based on convolutional neural network proposed in this paper has a higher target detection probability, which reflects the effectiveness of the method proposed in this paper.

A deep learning-based cyclic shift keying spread spectrum (CSK-SS) underwater acoustic (UWA) communication system is proposed for improving the performance of the conventional system in low signal-to-noise ratio and multipath effects. The proposed deep learning-based system involves the long- and short-term memory (LSTM) architecture-based neural network model as the receiving module of the system. The neural network is fed with the communication signals passing through known channel impulse responses in the offline stage, and then directly used to demodulate the received signal in the online stage to reduce the influence of the above factors. Numerical simulation and actual data results suggest that the deep learning-based CSK-SS UWA communication system is more reliable communication than a conventional system. In particular, the collected experimental data show that after preprocessing, when the communication rate is less than 180 bps, a bit error rate of less than 10−3 can be obtained at a signal-to-noise ratio of −8 dB.

For a surface ship carrying a towed array, its in-band radiated self-noise is one of the near-field strong interference, which will seriously limit the performance of the underwater acoustic (UWA) signal detection system. Typically, the conventional ship noise cancellation methods requires time-synchronized hydrophone arrays, such as the towed linear array (TLA), to suppress its self-noise by using spatial filters. However, the spatial filter based methods fail when the direction of arrival of the desired signal is in the ship noise masking area. In cooperative scenarios, the prior knowledge of the template signal provides additional temporal information, which can be utilized to design a time-domain representations based detection system. In this paper, a multiscale time self-attention network (MTSA-Net) is proposed to mitigate the ship radiated self-noise and enhance the desired signal to improve the performance of signal detection system. Experimental results based on sea trial data indicate the effectiveness of our proposed method. • This study reformulates the issue of reducing ship radiated self-noise as a signal separation task involving two types of signals: the desired signal from a cooperative source and the self-noise from the towed ship. • This paper introduces a novel approach, the multiscale time self-attention network (MTSA-Net), which is designed based on the cooperative source structure. • The proposed model MTSA-Net can obtain a high coefficient gain when maintaining sufficient discrimination. [ABSTRACT FROM AUTHOR]

Live-cell Ca2+ fluorescence microscopy is a cornerstone of cellular signaling analysis and imaging. The demand for high spatial and temporal imaging resolution is, however, intrinsically linked to a low signal-to-noise ratio (SNR) of the acquired spatio-temporal image data, which impedes on the subsequent image analysis. Advanced deconvolution and image restoration algorithms can partly mitigate the corresponding problems but are usually defined only for static images. Frame-by-frame application to spatio-temporal image data neglects inter-frame contextual relationships and temporal consistency of the imaged biological processes. Here, we propose a variational approach to time-dependent image restoration built on entropy-based regularization specifically suited to process low- and lowest-SNR fluorescence microscopy data. The advantage of the presented approach is demonstrated by means of four datasets: synthetic data for in-depth evaluation of the algorithm behavior; two datasets acquired for analysis of initial Ca2+ microdomains in T-cells; finally, to illustrate the transferability of the methodical concept to different applications, one dataset depicting spontaneous Ca2+ signaling in jGCaMP7b-expressing astrocytes. To foster re-use and reproducibility, the source code is made publicly available.

Vibration greatly impairs the imaging quality of airborne synthetic aperture ladar (SAL), so the vibration phase error must be estimated and compensated before imaging. Limited by the laser's average power, the high pulse repetition frequency of the SAL will result in lower peak-power of the laser, thereby reducing the echo signal's single-pulse signal-to-noise ratio (SNR). In the case of a low SNR, the coherence between channels worsens, and the vibration phase-error compensation method based on interferometry is difficult to apply. We analyze the imaging SNR and resolution of the echo signal when the airborne SAL observes noncooperative targets by intentionally broadening the transmitting and receiving beams, discussing the imaging processing flow of the low SNR signal under conditions of beam-broaden and platform vibration. The data processing flow includes the phase correction of the echo signal using the transmitting reference channel signal. Based on the flight parameters of the airborne SAL, the echo was dechirped and narrow-band filtered to reduce noise. After subaperture division, the spatial correlation algorithm was used to estimate and compensate the vibration phase error, and the range-Doppler algorithm was used to achieve strip imaging. At the same time, the phase gradient autofocus algorithm was introduced to improve the focusing effect. The processing results of flight experiment data of the airborne SAL corresponding to noncooperative targets at 1 km demonstrated the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

• We propose the deep-learning based separation network for underwater acoustic signal detection, which is addressed for the first time in underwater scenario by a data-driven separation method up to our knowledge. • With the help of Encoder-Decoder design, we project the input signal into a high-dimensional latent space where informative features are extracted and preserved, leading to better robustness to the random range migration compared to conventional handcraft methods. • We theoretically revise the conventional mean square error loss, which is agnostic to correlation coefficients. To address this issue, we propose a novel scale invariant signal-to-noise ratio as our loss function to better leverage the temporal coherence. To achieve efficient underwater acoustic (UWA) signal detection in low signal-to-noise ratio (SNR) scenarios, the transmitted signals are designed with a large time-bandwidth product to get a high detection gain. The linear correlator (LC) is considered as the maximum SNR detector, whose detection gain is proportional to the time-bandwidth product. However, the detection performance of LC degrades significantly in time-variant multipath UWA channel and non-Gaussian UWA ambient noise. In this study, we present a deep-learning based two-stage UWA signal detection method in intensity fluctuation environments. This method takes the advantages of the Conv-TasNet and Encoder-Decoder network, which utilizes an encoder module to extract signal features, a separation module to enhance the signal components and then another decoder module to reconstruct the transmitted signal. To demonstrate the performance of the proposed method, the datasets used for training and testing originated from the ASIAEX 2001 South China Sea (SCS) experiment. The experimental results show that our model outperforms the classical LC and channel estimation based LC (CE-LC) in constant false alarm rate (CFAR) detection and also surpasses the TCDAE and Conv-Tasnet models as evaluated by DSI-SNR1 and DSI-SNR2. [ABSTRACT FROM AUTHOR]

In tight reservoirs, the signal-to-noise ratio (SNR) of the echo train of nuclear magnetic resonance (NMR) logging is very low, which leads to inaccurate inversion results using the current echo train processing method. Therefore, a new processing method for improving inversion accuracy was developed, which includes four steps: echo-preferring, filtering, signal superimposition, and inversion. First, the echo-preferring method, which chooses high-quality echo trains, was studied in detail by analyzing low SNR NMR measurements. The SVD filtering method was then improved, and the echo train decomposed and filtered. Next, the echo trains of the adjacent depth points were stacked to further improve the SNR. Finally, the Butler–Reeds–Dawson (BRD) method was used to perform the inversion. Through numerical simulation experiments, the effects of echo optimization and high-efficiency filtering on inversion accuracy were studied, and the reliability of the method was verified. A case study of the gas reservoir in the Daqing oilfield shows that the porosity obtained by the new method agrees favorably with the core porosity, and the T 2 distribution is also more reasonable. The new method improves inversion accuracy and is equivalent to improving the original echo train of NMR logging, which lays the foundation for the calculation of movable and bound fluid volumes in tight reservoirs. • For ultra-low SNR NMR logging, a new processing method and procedure is proposed. • The echo preferring method is developed for low SNR echo train and proved effective. • The SVD filtering method is improved, and the preferred echo train is decomposed. • The echo train stacking is effectively added in the NMR logging data processing. [ABSTRACT FROM AUTHOR]

Raman data with low signal-to-noise ratio (SNR) are often obtained in the process of rapid or low-power acquisition for biological samples. The noise removal out of a biological Raman spectrum is laborious owing to numerous and superimposed Raman peaks, and it often takes a lot of time to find the optimal preproceed parameters. In this study, an adaptive denoising method for biological Raman data has been proposed, using three comparisons and a quantitative prediction model to show its simplicity and effectiveness. In this work, we decomposed Raman spectra of human skin and rice leaf on Hilbert vibration decomposition (HVD) method, to obtain a series of initial values, for example as peak position, intensity and full width at half maximum (FWHM), then fitted Raman spectrum according the Vogit function until the indicator function F reached the minimum value. The SNR and the root mean square error（ R M S E) of HVD denoised spectrum were 57.5864 dB and 0.0108, and were both better than those of preproceed data (35.1826 dB, 0.0337; 32.4120 dB, 0.0352; 38.0685 dB, 0.0226) by Empirical Mode Decomposition (EMD), Ensemble Empirical Mode Decomposition (EEMD), and Local Complementary Ensemble Empirical Mode Decomposition (LCEMD), compared with adaptive denoising methods, EMD, EEMD and LCEEMD, the improved algorithm had the advantages of no modal aliasing, no noise residue and independence of normalized permutation entropy. Then HVD denoised data of rice leaf samples were compressed by principal component analysis (PCA) and used for chlorophyll content prediction model. The correlation coefficient and standard error for prediction set of HVD +PCA model were 0.856 and 5.38, and were also better than those of non-adaptive preproceed data models (Mobing Average Smoothing, Gaussian Filter Smoothing, Median Filter Smoothing, Savitzky-Golay Smoothing and Wavelet Transform Denoising). The final experimental comparison showed that the HVD denoising method can also be applied to the Raman spectrum noise removal of inorganic samples, but better performances had been shown in the noise removal of complex biological Raman spectra. Accurate denoising of biological Raman spectrum is of great significance for accurate quantitative prediction and qualitative analysis, and no prior setting of the preproceed parameters makes the HVD denoising method easy to operate. [ABSTRACT FROM AUTHOR]

Live-cell Ca 2 + fluorescence microscopy is a cornerstone of cellular signaling analysis and imaging. The demand for high spatial and temporal imaging resolution is, however, intrinsically linked to a low signal-to-noise ratio (SNR) of the acquired spatio-temporal image data, which impedes on the subsequent image analysis. Advanced deconvolution and image restoration algorithms can partly mitigate the corresponding problems but are usually defined only for static images. Frame-by-frame application to spatio-temporal image data neglects inter-frame contextual relationships and temporal consistency of the imaged biological processes. Here, we propose a variational approach to time-dependent image restoration built on entropy-based regularization specifically suited to process low- and lowest-SNR fluorescence microscopy data. The advantage of the presented approach is demonstrated by means of four datasets: synthetic data for in-depth evaluation of the algorithm behavior; two datasets acquired for analysis of initial Ca 2 + microdomains in T-cells; finally, to illustrate the transferability of the methodical concept to different applications, one dataset depicting spontaneous Ca 2 + signaling in jGCaMP7b-expressing astrocytes. To foster re-use and reproducibility, the source code is made publicly available. [ABSTRACT FROM AUTHOR]

A deep learning-based cyclic shift keying spread spectrum (CSK-SS) underwater acoustic (UWA) communication system is proposed for improving the performance of the conventional system in low signal-to-noise ratio and multipath effects. The proposed deep learning-based system involves the long- and short-term memory (LSTM) architecture-based neural network model as the receiving module of the system. The neural network is fed with the communication signals passing through known channel impulse responses in the offline stage, and then directly used to demodulate the received signal in the online stage to reduce the influence of the above factors. Numerical simulation and actual data results suggest that the deep learning-based CSK-SS UWA communication system is more reliable communication than a conventional system. In particular, the collected experimental data show that after preprocessing, when the communication rate is less than 180 bps, a bit error rate of less than 10−3 can be obtained at a signal-to-noise ratio of −8 dB. [ABSTRACT FROM AUTHOR]