110 results about "Soft thresholding" patented technology

The invention provides a compressed sensing theory-based reconstruction method of a magnetic resonance random sampled K space data image. The reconstruction method applies a contourlet conversion and iterative soft thresholding method to realize reconstruction of a magnetic resonance image. The method comprises the following steps: collecting K space data in a magnetic resonance image scanner according to a preset observation matrix phi to generate a measurement value, and keeping y; acquiring y from a coil of the magnetic resonance image scanner, and transmitting y to a computer; and finallyconstructing a same phi, constructing any orthogonal transformation psi, and recovering from y by adopting a compressed sensing theory-based magnetic resonance random sampled K space data image reconstruction method according to reconstruction. According to the method, scanning time is saved, quick imaging is realized, high-quality reliable image information is provided to medical nuclear magnetic resonance imaging detection, and solid theoretical and practical foundation is established for further development and large-scale popularization and application of the medical imaging detection technology.

The invention discloses a partial echo compressed sensing-based quick magnetic resonance imaging (MRI) method. The conventional imaging method has low speed and high hardware cost. The method comprises the following steps of: acquiring echo data of a random variable density part, namely intensively acquiring data in a central area of a k-space and acquiring the data around the k-space randomly and sparsely to generate a two-dimensional random mask, adding the two-dimensional random mask into every data point which needs to be acquired on a frequency coding shaft to form a three-dimensional random mask, and acquiring the data of the k-space according to the generated three-dimensional random mask; re-establishing by projection onto convex sets based on a wavelet domain which is de-noised by soft thresholding; and nonlinearly re-establishing a minimum L1 normal number based on finite difference transformation, namely sparsely transforming an image space signal x, determining an optimization objective and solving the optimization objective. By the method of the invention, partial echo technology and compressed sensing technology are combined and applied to data acquisition of MRI, sothat echo time is shortened, and data acquisition time is shortened at the same time.

The invention provides a signal identification and classification method. The method comprises the followings steps of: carrying out noise reduction on initial data containing higher noise by utilizing a wavelet transform method, decomposing signals into high-frequency information and low-frequency information in data analysis, carrying out noise cancelling on the signals by adopting a soft thresholding method and then carrying out signal reconstruction; carrying out further decomposition on the high-frequency part which is not detailedly classified by multiscale analysis while inheriting allthe favorable time-frequency localization advantages of the wavelet transform; analyzing the signals within different frequency bands after multi-layered decomposition by utilizing the wavelet packettransform to extract out characteristic information reflecting a system state; transforming the characteristic vectors of input signals into a high-dimensional characteristic space through non-lineartransform and then solving for an optimal linear classification plane in the high-dimensional characteristic space. The invention overcomes the defects of difficult determination of a network structure, low convergence rate, requirement on large quantities of data samples during training, and the like in neural network learning and enables the neural network learning to be with the characteristics of high precision and strong real time in the aspect of practical application of engineering.

The invention relates to a feature flow-based method for generating an abstract line drawing. The invention aims to design a novel filtering method-FGaD filtering. A feature flow-based anisotropic filtering framework is referenced, a core thought is that the sum of the absolute value of Gaussian first-order filtering results and the absolute value of Gaussian difference filtering results is defined as a differential response value for edge detection, and the differential response value is subjected to soft thresholding so as to acquire a highly abstract line drawings result. According to the FGaD filtering method provided by the invention, large-scale visual features can be effectively detected, and adjacent tiny edge lines can be fused, and highly abstract line drawings with strong stylized effects are generated.

The invention discloses a self-adaptive enhancement method based on a dark-light color image. The method specifically includes the following steps: carrying out HSV color space conversion on the dark-light color image to obtain a channel-V image containing brightness information, and carrying out second wavelet decomposition on the channel-V image to obtain a low-frequency sub-image of W and six high-frequency sub-images of W<v><H>, W<v><V>, W<v><D >, W<v><H'>, W<v><D'> and W<v><V'>; using a semi-soft threshold value to process the high-frequency sub-band images of W<v><H>, W<v><V>, W<v><D>, W<v><H'>, W <v><D'> and W<v><V'> obtained by decomposition; carrying out image enhancement on the low-frequency sub-band image obtained by decomposition; and carrying out image reconstruction on theprocessed high-frequency sub-band images and the processed low-frequency sub-band image, and carrying out inverse HSV conversion on a reconstructed image to obtain an enhanced color image. Accordingto the method, a method of wavelet transform is employed to enhance noise portion thresholds and non-noise high-frequency coefficients, and denoising problems of existing dark-light color images are solved.

The invention provides an empirical mode decomposition denoising method based on a arevised wavelet threshold value. The method is characterized by comprising the following steps of first carrying out the empirical mode decomposition on an original signal to acquire a plurality of intrinsic mode functions I with the frequency being gradually reduced and a remainder term; calculating the smoothness of each intrinsic mode function I; calculating a threshold value of each intrinsic function I by utilizing a wavelet threshold value method; revising the threshold value obtained through the wavelet method according to the smoothness and a serial number of each intrinsic mode function I; carrying out the soft threshold value treatment on each intrinsic mode function I by utilizing the revised threshold value to obtain an intrinsic mode function II; finally reconstructing the intrinsic mode function II to obtain a denoised signal. The method is good in self-adaptability, the threshold value calculated by adopting the wavelet threshold value method is revised through the smoothness index, a signal with high signal-to-noise ratio is obtained on the premise of guaranteeing the smoothness, and the method can be used for denoising the ultrasonic signal.

The invention discloses a method and a system for denoising partial discharge signals of power cables and extracting effective signals. A new threshold function is proposed to perform threshold processing on wavelet coefficients after wavelet decomposition. The problems existing in an existing hard threshold method and a soft threshold method for processing signals can be solved. Denoising processing and extraction of the effective signals can be performed on cable operation data laid in the cable trench or directly buried underground. Accurate positioning of cable faults is realized, and fault points are effectively found out in time, thereby guaranteeing safe and effective operation of electric power facilities.

Methods and associated apparatuses for imaging a target. An echo sequence image of the target is acquired and a log spectrum of at least a portion of the echo sequence image is computed. A point spread function is estimated by one of two methods. According to the first method, a low-resolution wavelet projection of the echo sequence log spectrum is used as an estimate of the log spectrum of the point spread function. According to the second method, an outlier-resistant wavelet transform of the echo sequence log spectrum is effected, followed by soft-thresholding and an inverse wavelet transform. Under both methods, a frequency domain phase of the point spread function also is estimated, the relevant portion of the echo sequence image is deconvolved using the estimated point spread function.

The invention provides an image enhancing method and an image enhancing system. A gray level of an original image is modified by using a preset gray level histogram algorithm, and therefore the local thinning of the original image is realized. High frequency information extracted from the original image is denoised, and an acquired target high frequency information after the above mentioned denoising processing is reintegrated with low frequency information extracted from the modified image to acquire a target enhanced image. The high frequency information is processed by adopting a semi-soft thresholding method, and an unpolluted image is protected, and at the same time, continuity of a processed wavelet system is guaranteed. High frequency signals and low frequency signals of wavelets are separately processed, and therefore the local thinning processing of the original image is realized, and the integral denoising processing of the original image is carried out, and therefore the integral enhancing effect of the enhanced image is guaranteed.

The invention discloses a fault diagnosis device and a method for an antifriction bearing based on the analysis on the morphological component of an acoustic signal. During the rotation process, the bearing generates an acoustic signal. Through analyzing the time domain features of the acoustic signal, a novel bearing fault diagnosis method based on the MCA and Hilbert spectrum analysis of the acoustic signal is provided. The result of the simulation and example analysis shows that, the fault acoustic signal of the bearing is formed through compounding an impulse component, a resonance component and a noise component. Through constructing a redundant dictionary composed of a coif 4 wavelet dictionary and a local cosine dictionary, the good approximation to the original signal is realized. Based on the MCA method, the generalized soft thresholding operation and the de-noising are realized. Meanwhile, the impulse component of the original signal is sparsely isolated. Finally, in combination with the impulse component of the original signal and the Hilbert frequency spectrum thereof, isolated based on the MCA method, the fault type of the antifriction bearing can be accurately identified.

The invention discloses a magnetic resonance super undersampled K data imaging method based on studying a generalized double-layer Bergman non-convex-type dictionary. The imaging method includes the steps that 1 prior information with non-convex function p norm is blended into a double-layer Bergman dictionary study frame, the dictionary study and coefficient sparsity are conducted, and an image sparse representation model is established; 2 by using an increasing auxiliary variable and alternating technology, the dictionary study and coefficient sparsity are updated in an inner-layer iteration of the double-layer Bergman iterative dictionary study, an objective function of the prior information with the non-convex p norm is obtained by using a generalized soft threshold iterative method particularly, and the sparse coefficients are updated; 3 the image is updated in an outer-layer iteration of the double-layer Bergman dictionary study, and a reconstructed image is obtained. By means of the generalized soft threshold iterative method, the objective function of the prior information with non-convex p norm is obtained, small coefficients can be punished in a larger range and large coefficients are smaller in deviation, a sparse representation image can be further obtained, the image can be accurately reconstructed with less scan measurement, artifacts of the reconstructed image are reduced, and more image details are recovered.

The present invention relates to a computer-implemented method and apparatus for data processing for the purpose of blind separation of nonnegative correlated pure components from smaller number of nonlinear mixtures of mass spectra. More specific, the invention relates to preprocessing of recorded matrix of mixtures spectra by robust principal component analysis, trimmed thresholding, hard thresholding and soft thresholding; empirical kernel map-based nonlinear mappings of preprocessed matrix of mixtures mass spectra into reproducible kernel Hilbert space and linear sparseness and nonnegativity constrained factorization of mapped matrices therein. Thereby, preprocessing of recorded matrix of mixtures mass spectra is performed to suppress higher order monomials of the pure components that are induced by nonlinear mixtures. Components separated by each factorization are correlated with the ones stored in the library. Thereby, component from the library is associated with the separated component by which it has the highest correlation coefficient. Value of the correlation coefficient indicates degree of pureness of the separated component. Separated components that are not assigned to the pure components from the library can be considered as candidates for new pure components. Identified pure components can be used for identification of compounds in chemical synthesis, food quality inspection or pollution inspection, identification and characterization of compounds obtained from natural sources (microorganisms, plants and animals), or in instrumental diagnostics—determination and identification of metabolites and biomarkers present in biological fluids (urine, blood plasma, cerebrospinal fluid, saliva, amniotic fluid, bile, tears, etc.) or tissue extracts.

The invention discloses a micro-seismic signal multi-scale denoising method and device and a readable storage medium, and the method comprises the steps: 1, obtaining a micro-seismic signal, carryingout the EMD or EEMD decomposition, and filtering the high-frequency noise in the decomposed signal; 2, respectively constructing a Hankel matrix of each IMF component; 3, carrying out singular value decomposition is based on the Hankel matrix of each IMF component to obtain a score vector of principal component analysis, carrying out primary denoising, and carrying out soft threshold secondary denoising on each component signal and residual component after primary denoising; and 4, superposing the component signal subjected to secondary denoising and the residual component to obtain a denoisedmicro-seismic signal. According to the method, singular value decomposition is associated with principal component analysis, information of singular value decomposition serves as a score vector of principal component analysis, the PCA calculation process is simplified, and the defect that denoising cannot be conducted on a single column vector through the score vector of singular value decomposition is overcome.

The invention provides a seismic data de-noising method adopting an adaptive wavelet threshold function. The method addresses the problem that the quality of seismic data in a low signal-to-noise ratio region of oil and gas seismic exploration cannot meet the requirements of treatment and interpretation. By establishing a new threshold function overcoming the defects of hard and soft thresholdsand using wavelet transform as a core idea, the de-noising goal of seismic data is gained and a new threshold function is determined. Compared with the current similar technologies, the method has theadvantages that the de-noising result is closer to original useful signals and the quality of the seismic data is greatly improved.

The invention discloses a short-term wind speed prediction method integrating deep learning and data denoising. The method comprises the following steps: 1) decomposing a wind speed sequence A by using discrete wavelet transform; 2) distinguishing a noise signal and an effective signal after discrete wavelet transform by using a wavelet soft threshold denoising method; 3) setting a wavelet transform coefficient of the noise signal as zero, and reconstructing a wind speed time sequence to obtain a denoised wind speed sequence B; 4) training a gating cycle unit neural network by using the wind speed sequence B to obtain a wind speed prediction model, and 6) inputting real-time wind speed into the wind speed prediction model to complete future multi-step wind speed prediction. The short-termwind speed high-quality prediction can be realized so as to ensure the economic dispatching and safe operation of a power system.

The invention discloses a grayscale image adaptive enhancement method for singular value decomposition and beetle antennae search algorithm. The method comprises subjecting an input original image todiscrete wavelet decomposition in order to convert the original image into the frequency domain; applying the beetle antennae search algorithm to decomposed wavelet sub band to solve the optimal adjustment parameter of a soft threshold function, and using the soft threshold function of the corresponding optimal adjustment parameter for threshold processing to enhance the contour; performing singular value decomposition on the low-frequency sub band, correcting the wavelet coefficients in the low-frequency sub band to realize the transformation of the luminance; and finally, performing inversewavelet transform on each processed wavelet sub band to achieve the image adaptive enhancement.

The disclosed apparatus, systems and methods relate to a framelet-based iterative algorithm for polychromatic CT which can reconstruct two components using a single scan. The algorithm can have various steps including a scaled-gradient descent step of constant or variant step sizes; a non-negativity step; a soft thresholding step; and a color reconstruction step.

A method and system for automatically removing noise from images. First, a current pixel is received. Next, an activity metric of the current pixel is computed by using the current pixel and a neighborhood of pixels related to the current pixel. Then, a distance metric that indicates the likelihood that the current pixel is a background pixel is computed by using the activity metric and at least one background parameter. The distance metric provides a soft thresholding framework for the background removal. The current pixel is modified based on the distance metric.

The invention provides a voice noise reduction method based on spectral subtraction and wavelet transform. The method comprises the following steps of removing a direct current component of an input pure voice signal; performing normalization processing on an amplitude value; superimposing white gaussian noise on the processed voice signal so as to obtain the voice signal with noise; setting a leading no-word section; calculating the frame number of the leading no-word section; processing the voice signal with noise by spectral subtraction; performing wavelet transform on the obtained voice signal subjected to spectral subtraction to obtain a wavelet signal with a discrete telescopic factor and a discrete translation factor; according to a low-frequency component in the wavelet signal anda high-frequency component of each decomposition layer, obtaining a low-frequency coefficient and a high-frequency coefficient of each decomposition layer; respectively processing the high-frequency coefficient of each decomposition layer by a zero setting processing method and a soft threshold processing method so as to obtain the processed high-frequency coefficient of each decomposition layer;and performing inverse wavelet transform to obtain a final noise reduction voice signal. The music noise can be obviously reduced; in addition, only little high-frequency distortion exists; and the noise reduction capability is further improved.

The invention discloses an improved wavelet threshold denoising-based fault diagnosis method and an improved wavelet threshold denoising-based fault diagnosis system. The method comprises the steps ofobtaining to-be-diagnosed TE process data; standardizing the acquired data; carrying out wavelet transform decomposition on the standardized data, decomposing the data into a plurality of layers, andobtaining a wavelet coefficient of each layer; for the wavelet coefficient of each layer, calculating a skewness coefficient and a kurtosis coefficient; according to the skewness coefficient and thekurtosis coefficient of each layer, judging whether the wavelet coefficient of the current layer conforms to normal distribution or not is judged, and if yes, a hard threshold method is adopted for denoising; otherwise, denoising by adopting a soft threshold method; after denoising, obtaining a processed high-frequency coefficient and a processed low-frequency coefficient of each layer of waveletcoefficient, and reconstructing a data signal by using the processed high-frequency coefficient and the processed low-frequency coefficient to obtain denoised data; and analyzing the denoised data byutilizing a PCA model, and judging whether the TE process data to be diagnosed has fault data or not.

The invention provides an adaptive empirical mode decomposition denoising method for satellite-borne full-waveform signals. The adaptive empirical mode decomposition denoising method comprises the following steps: acquiring noisy full-waveform signals; carrying out EMD decomposition to acquire all IMF components; determining a Hurst index value of each IMF component; judging whether the componentis a high-frequency IMF component or not according to the index value; if the component is a high-frequency IMF component, performing soft threshold processing to obtain a denoised high-frequency IMFcomponent and reserving the denoised high-frequency IMF component, otherwise, directly reserving the denoised high-frequency IMF component; and superposing the denoised high-frequency IMF component and the denoised low-frequency IMF component, and reconstructing a denoised full-waveform signal. According to the self-adaptive empirical mode decomposition denoising method, rapid and large-scale judgment of high-frequency noise components is achieved by constructing an H value; meanwhile, soft threshold processing is carried out on the selected high-frequency IMF components, so that possible signal components are searched again in the components needing to be removed, effective signal loss caused by direct removal of the high-frequency components is avoided, and the signal-to-noise ratio is higher.

The invention discloses a sparse subspace clustering algorithm based on semi-supervision. The sparse subspace clustering algorithm comprises the steps that data prior information is converted into a constraint matrix suitable for a sparse subspace model in the form of point pair constraint; interference of flag-free bits is eliminated in the form of Hadamard product, the state of the coefficient represented by different constraint conditions is also considered and corresponding constraint terms are established; and a semi-supervised sparse subspace model of two hard threshold and soft threshold forms is established by using the constraint terms, and a semi-supervised framework is accordingly established on the sparse subspace clustering algorithm. The clustering accuracy of the sparse subspace algorithm can still be maintained by the algorithm without prior information. Meanwhile, the performance advantages of the sparse subspace clustering algorithm are also absorbed so that the high-dimensional clustering problem containing interference information data can be directly and effectively processed, the clustering performance is ensured to be effectively enhanced under the condition of less known prior information and thus the algorithm applicability can be increased.

In the field of machinery, vibration signal feature extraction is of great significance to state recognition, fault diagnosis and other problems, but vibration signals are inevitably interfered by noise in the acquisition process, so that signal features are difficult to extract. The invention discloses a mechanical vibration signal feature extraction method based on variational mode decompositionand grey correlation analysis, and the method comprises the steps: decomposing a vibration signal into a plurality of modes through VMD, and determining the number K of important preset parameter modes of the VMD through grey correlation analysis; after VMD decomposition, selecting the noise dominant modes again through grey correlation analysis, and conducting soft threshold processing on the noise dominant modes to remove noise components in the noise dominant modes; and reconstructing the processed noise dominant mode and other modes to obtain a denoised vibration signal, and further extracting signal features. According to the invention, the technical problems of mechanical vibration signal noise reduction and feature extraction are effectively solved.

The invention aims to provide a method and device for reducing acceleration signal measurement noise, wherein a wavelet decomposition technology is mainly used for performing soft threshold noise reduction processing on two paths of acceleration signals respectively, two paths of wavelet reconstructed signals are compared layer by layer, and finally an effective acceleration signal is obtained, sothat the time domain and frequency domain characteristics of the acceleration signals and the relevant information of signal mutation can be reflected at the same time; more importantly, two acceleration sensors are adopted for measurement, the noise of the two acceleration sensors may be different, but the fluctuation characteristics of the effective signals of the two acceleration sensors are basically consistent, extraction can be carried out after wavelet processing so as to obtain a needed signal, if the difference between the two noise reduction signals is large, the problems existing in the acceleration signal noise reduction processing procedure need to be corrected, such as selection of a threshold value; thus, the accuracy of signal noise reduction processing can also be improved.

The invention relates to a noisy SAR image target recognition method based on wavelet denoising threshold self-learning, and belongs to the technical field of radar target recognition. Based on a threshold-learnable wavelet speckle suppression network and a compression-excitation convolutional neural network, the threshold-learnable wavelet speckle suppression network comprises a wavelet transformation module, a threshold-learnable module and an inverse wavelet transformation module. The wavelet speckle suppression network capable of learning the threshold extracts an image high-frequency component through DWT, puts the image high-frequency component into a convolution and full connection layer, adaptively carries out soft threshold processing on the high-frequency component, combines the high-frequency component with a low-frequency component, and obtains a denoised image through IWT; therefore, the threshold value of wavelet denoising is supervised and learned through a training set label, features are extracted from an image output by a wavelet speckle suppression network capable of learning the threshold value through network automatic layering, and a compression-excitation module is added to balance the contribution degree of a feature map from an original image and a denoising branch. The method has the advantages of high identification capability and high test precision.

An SR image reconstruction method based on deep convolutional sparse coding (DCSC) is provided. The method includes: embedding a multi-layer learned iterative soft thresholding algorithm (ML-LISTA) of a multi-layer convolutional sparse coding (ML-CSC) model into a deep convolutional neural network (DCNN), adaptively updating all parameters of the ML-LISTA with a learning ability of the DCNN, and constructing an SR multi-layer convolutional sparse coding (SRMCSC) network which is an interpretable end-to-end supervised neural network for SR image reconstruction; and introducing residual learning, extracting a residual feature with the ML-LISTA, and reconstructing a high-resolution (HR) image in combination with the residual feature and an input image, thereby accelerating a training speed and a convergence speed of the SRMCSC network. The SRMCSC network provided by the present disclosure has the compact structure and the desirable interpretability, and can generate visually attractive results to offer a practical solution for the SR reconstruction.

The invention discloses a variational level set image segmentation method based on Landmark simplex constraint, and belongs to the technical field of digital image processing. According to the method,prior landmark feature points of an image are converted into simplex constraints, level set expression of Landmark feature points is achieved, a variational level set image segmentation model based on Landmark simplex constraints is provided, and evolution of contours and prior points is achieved. For nonlinearity, non-convexity and non-smoothness of a segmentation model, auxiliary variables areintroduced to convert solving of a non-convex energy equation into a convex sub-problem, an alternating direction multiplier method is adopted, and a rapid projection method, a generalized soft threshold formula and a gradient descent method are comprehensively used for solving. Experimental results show that the variational level set image segmentation method based on the Landmark simplex constraint is high in segmentation performance, and the segmentation problem of noisy images, weak edge images and heterogeneous images can be solved robustly and efficiently. The obtained segmentation result is good in subjective visual effect and excellent in objective evaluation standard, and a foundation is laid for subsequent image feature extraction, interpretation and other applications.

The invention relates to a mechanical equipment residual life prediction method for optimizing BiLSTM based on DRSN and sparrow search, and belongs to the field of mechanical equipment monitoring. Firstly, self-adaptive feature learning is carried out on an original signal by utilizing DRSN without priori knowledge, an attention mechanism and a soft thresholding structure of the network can effectively eliminate influences of noise related features, and degradation features of mechanical equipment are excavated to construct health indexes; then, a residual life prediction model is constructed by utilizing a BiLSTM network, and the parameters are optimized by adopting a sparrow search algorithm aiming at the problem that the number of neurons and the learning rate of a hidden layer of the BiLSTM are difficult to set; and after health indexes extracted by the DRSN are smoothed, normalized life is used as a label and input into the optimized BiLSTM prediction model, and residual life prediction of the mechanical equipment is completed.

The invention relates to a denoising method in low-frequency oscillation analysis of power systems, in particular to a wavelet threshold denoising method for low-frequency oscillation Prony analysis.The invention improves the traditional wavelet threshold denoising algorithm by introducing new threshold selection and threshold function. The selection of new threshold can ensure that the wavelet coefficient varies with the distribution of scales, and fit the actual distribution of noise in different types of signals at each layer, so as to ensure sufficient denoising ability without distortionof reconstructed signals. Compared with the common soft threshold function, the new threshold function reduces the fixed deviation, and the continuity of the function is ensured compared with the common hard threshold function, thus eliminating the possible oscillation of the reconstructed signal.