50 results about "Rényi entropy" patented technology

The invention relates to a radar radiation source identification method based on feature fusion. The method comprises the steps of generating a radar radiation source unintentional modulation signal set; extracting AR model coefficients, Renyi entropy features and spectral kurtosis features; computing smoothed pseudo Wigner-Ville distribution, generating time-frequency images, and carrying out graying and adaptive binarization to obtain adaptive binarized images; extracting pseudo Zernike matrix and Hu matrix features of the images; extracting signal time-frequency image unintentional modulation features through application of an AlexNet convolutional neural network, carrying out normalization, and carrying out feature fusion to obtain fused feature vectors; and inputting the fused featurevectors into a support vector machine, training the support vector machine optimized through particle swarm optimization, and inputting the radar radiation source signal set into a system to finish identifying radar radiation sources. According to the method, the signals are analyzed from a time domain, a frequency domain and a time-frequency domain, various unintentional modulation features areextracted comprehensively, and the problems that the extracted unintentional modulation features are low applicability and reliability and the radiation sources are difficult to identify is solved.

The invention belongs to the technical field of radar emitter signal modulation recognition, and particularly relates to a modulation recognition method for extracting time-frequency image features byjoint entropy and pre-training CNN. The method includes the following steps: firstly, performing time-frequency transformation on 9 types of radar signal sets to be identified to obtain a time-frequency image; and then based on a pre-training convolutional neural network model imagenet-vgg-verydeep-19 provided by the MatConvNet official website, constituting an FT-VGGNet-fc6 feature migration extraction module from an input layer to an fc6 full connection layer; and then, sending an adjusted image to the feature migration extraction module, and outputting time-frequency image features of radar signals; performing graying on the adjusted image, and manually extracting the Renyi entropy of the processed image; and then, dividing a training set and a test set according to a certain proportion, and selecting the training set to train an SVM classifier; and finally, adopting the trained SVM classifier to recognize the training set of time-frequency images, and adopting a data set composedof 9 types of radar signals with multiple signal-to-noise ratios to verify a recognition rate of an FT-VGGNET-fc6-SVM classifier.

The invention provides a heart rate variability feature classification method based on multi-scale Renyi entropy and belongs to the field of electrocardiosignal processing. R wave positioning is performed after pretreatment such as interference and baseline drift removal is performed on to-be-processed original electrocardiosignals, and the interval of adjacent R waves is calculated to obtain an HRV sequence; discrete wavelet coefficients are obtained through discrete wavelet conversion of the HRV sequence; a proper q value is selected according to requirements to calculate the Renyi entropy of the wavelet coefficient of each layer; feature vectors are constructed by using the calculated Renyi entropy values of all scales for classification and identification of the electrocardiosignals.

The invention discloses a semi-supervised hyperspectral remote sensing image classification method based on information entropies and relates to the field of remote sensing. The method specifically comprises the steps of (1) inputting a hyperspectral remote sensing image, (2) inputting a training sample set, (3) inputting a category set corresponding to training samples, (4) calculating the probability of the category which each image element in the hyperspectral remote sensing image represents through a multi-classification linear regression method, (5) outputting the categories corresponding to all the image elements according to the calculated probabilities of all the image elements, (6) outputting the classification result and judging the accuracy of the output result, (7) converting the probabilities of all the image elements in the remote sensing image into indeterminacy through the Renyi entropy algorithm, (8) converting unmarked label image elements in the hyperspectral remote sensing image into marked label image elements according to the indeterminacy, (9) adding new marked labels into the training set, and (10) conducting iteration operation. The hyperspectral remote sensing image classification method has the advantages of being easy to realize, low in calculation complexity and the like.

The invention discloses a non-local means filtering method for a speckle noise pollution image. The method comprises that iterative computation of a neurons firing state image series of the speckle noise pollution image is conducted through a pulse transmission cortex model, a Renyi entropy vector is extracted by the neurons firing state image series, and non-local means filtering is conducted on the speckle noise pollution image based on the Renyi entropy vector, so that a denoised gray value is obtained. By the aid of the method, rotation invariance, translation invariance and scaling invariance can be extracted from the image containing speckle noise, the method can use more image information for denoising than traditional methods, besides, the similarity between two image pixel blocks can be calculated reasonably, image noise can be suppressed significantly, and the peak signal-to-noise ratio of the image can be improved, so that detailed information of the image can be effectively protected.

The invention provides a combined classifier multi-signal modulation identification method based on fractional Fourier transform. The method includes steps: firstly, solving the problem of multi-signal separation by employing an FRFT technology, then realizing accurate classification and identification of a strong signal with the combination of a SVM by employing a pre-trained CNN in time and frequency domains, extracting characteristics including the spectrum kurtosis, a root-mean-square value, an AR characteristic coefficient and a Renyi entropy etc. of a weak signal to realize characteristic fusion, performing dimension reduction on fusion characteristics by employing a principal component analysis (PCA) to improve the system validity, and inputting the characteristics after dimension reduction to the SVM to realize second signal classification and identification and finally realize accurate and rapid identification of the multi-signal modulation mode with low signal to noise ratio.

The invention relates to a video significance processing method based on spectral analysis. The method comprises the following steps of: extracting a key frame from an input video by utilizing a method for extracting a video key frame according to motion information to acquire a video key frame sequence; extracting a video key frame gray level image sequence and a video key frame motion information characteristic map sequence; calculating Renyi entropy of the video key frame gray level image sequence and the video key frame motion information characteristic map sequence; measuring the distribution of significance points in the air domain significant maps by the Renyi entropy and using a result as a significant value of an air domain significant map; and finally linearly fusing the obvious vale of the video key frame air domain significant value and an obvious value of the time domain so as to acquire an overall obvious value of the video key frame. By the method, the significance information carried by magnitude spectra and phase spectrum in the video key frame field is fully utilized, the calculated amount is small and the effect is prominent; and the distribution of the significance points in the significant maps is measured by the Renyi entropy, and the result is used as the significant value of the significant pattern, so that the problem of overall significance measurement of the significant maps are well solved; and therefore, the method can be applied to video analysis processing systems of various of military use or civil use.

The invention provides an image registration method based on mixed mutual information and improved particle swarm optimization. The image registration method comprises the following steps that mutual information of a floating image and a reference template image is calculated by Renyi entropy; the population size, the maximum number of steps of algorithm iteration, a nonlinear adjustment parameter n and the minimum threshold are set; the new speed and position of each particle are calculated according to the speed updating and position updating formula of a PSO algorithm so that the local optimal solution Pid and the global optimal solution Pgd are updated; local optimization searching is performed on the global optimal solution Pgd with shannon entropy acting as similarity measure so that global optimal solution Rid is calculated; the global optimal solution Rid and the preset minimum threshold Fmin are compared until iteration ends or algorithm registration fails and then the algorithm ends; and mapping of the gray value of each pixel of the transformed images is performed. Compared with the methods in the prior art, the image registration method is high in calculation speed, short in registration time and more accurate in registration parameter.

The invention discloses a rolling bearing fault detection method based on an actual measurement signal. The rolling bearing fault detection method comprises the steps of: firstly, converting a rollingbearing fault time domain vibration signal to an angle domain through employing an order tracking technology; secondly, carrying out parameter optimization on variational mode decomposition through adpopting a longicorn beard search algorithm, and decomposing all state vibration signals of a rolling bearing to obtain a series of intrinsic mode functions, wherein frequency band energy in differentintrinsic mode functions can change when different faults happen to the bearing; thirdly, extracting Renyi entropy features from modal components containing main fault information, and constructing afeature subset; and finally, using normal state vibration signals easy to obtain for training, extracting fault characteristic quantities, establishing fault data samples and incremental learning data samples, acquiring a fault recognition model through training by adopting a single-class support vector machine incremental learning algorithm, judging whether the rolling bearing breaks down or notaccurately, and achieving fault early warning.

The invention relates to an underwater information anti-interference method based on compressed sensing and Renyi entropy, and belongs to the technical field of underwater analog information conversion and compressed sensing. The method comprises the following steps: 1) reading underwater acquired information, calculating the Renyi entropy of each row, solving the row corresponding to the maximumvalue, and calculating the pixel maximum value of the row as the Renyi entropy threshold T1; 2) filtering by taking W * W as a unit based on T1; 3) carrying out discrete two-dimensional wavelet transform; 4) calculating Renyi entropy denoising thresholds of the high-frequency coefficients according to rows, and then denoising the high-frequency coefficients; 5) constructing a Gaussian random observation matrix to perform compressed observation on the denoised high-frequency coefficient; 6) performing AMP reconstruction on all the filtered high-frequency coefficients; and 7), performing waveletinverse transform on the low-frequency sub-coefficient and the reconstructed high-frequency sub-coefficient for obtaining recovered underwawter acquired information; according to the method, part ofnoise can be filtered out, the peak signal-to-noise ratio of the denoised underwater acquired image is effectively improved, and the average structural similarity of the denoised underwater image is optimized.

The invention belongs to the technical field of radar signal processing in a non-Gaussian clutter environment, and discloses a radar signal identification method based on generalized variational modedecomposition combined with a support vector machine, which comprises the following steps: carrying out generalized variational mode decomposition on a received radar radiation source signal to obtainan intrinsic mode component IMF; computing smooth pseudo Wigner-Ville time-frequency distribution matrix for each intrinsic mode component, and extracting Renyi entropy features of time-frequency distribution to construct a feature vector T; and finally, carrying out classification and identification by utilizing a support vector machine. When the generalized signal-to-noise ratio is larger than5dB, the identification rate of various signals reaches 65% or above, especially when the generalized signal-to-noise ratio is larger than 10dB, the identification rate of various signals reaches 90%or above, and therefore the identification effect of the method is good.

The invention provides a control method and a device for a flywheel energy storage system. The method comprises the steps of a, with the angular velocity, the d-axis current and the q-axis current of A, a flywheel as state variables, and the energy storage capacity of the flywheel as a control variable, constructing a model for the energy storage system of the flywheel, and discretizing the model according to the sampling time; B, constructing a performance index based on a second-order Renyi entropy, and determining an optimal input control variable increment with the partial derivative of the input control variable increment to be zero based on the performance index; C, determining a current optimal input control variable based on the optimal input control variable increment. According to the control method and device, based on the simplified energy storage model of the flywheel, a controller for the energy storage system of the flywheel can be designed. Meanwhile, the excess energy of the flywheel is stored, or the deficient energy of the flywheel is supplemented. In this way, a wind turbine generator set is ensured to transmit a smoothing power to the power grid.

The invention belongs to the signal processing field, and specifically relates to clutter filtering on opto-acoustic signals through Renyi entropy so as to improve signal to noise ratio and opto-acoustic imaging quality of the opto-acoustic signals. The opto-acoustic image reconstruction prefilter based on Renyi entropy includes the following steps: 1) acquiring opto-acoustic signals; 2) performing time frequency distribution solution on the opto-acoustic signals; 3) performing Renyi entropy solution on each opto-acoustic signal point; 4) determining the threshold; 5) performing filtering processing; and 6) performing image reconstruction on the opto-acoustic signals which are processed through filtering. Compared with the prior art, the opto-acoustic image reconstruction prefilter based on Renyi entropy has the following advantages: 1) the signal to noise ratio of the reconstructed image is obviously improved when opto-acoustic image reconstruction is performed through the opto-acoustic signals which are processed by the prefilter; and 2) the mean square error of the reconstructed image is obviously reduced when opto-acoustic image reconstruction is performed through the opto-acoustic signals which are processed by the prefilter.

The invention relates to an information enhancement and transmission method based on wavelet, threshold filtering and compressed sensing, and belongs to the technical field of image enhancement. The method comprises the following steps: carrying out discrete two-dimensional wavelet sparse basis on information to obtain high-frequency and low-frequency coefficients; observing the high-frequency coefficient, and outputting an observation result matrix; performing quantization, channel coding, modulation, transmission, demodulation, channel decoding and inverse quantization on the observation result matrix and the low-frequency coefficient, and outputting recovered high-frequency and low-frequency coefficients; denoising the high-frequency coefficient based on Renyi entropy, and then performing compressed sensing reconstruction; and performing Butterworth high-pass filtering on the low-frequency coefficient, and performing wavelet inverse transformation on the low-frequency coefficient and the recovered high-frequency coefficient to obtain recovered information. According to the method, the contour of the image with the blurred contour can be enhanced through high-pass filtering so that the processed image is clearer; part of noise can be filtered out in the compressed sensing process, and wavelet coefficient reconstruction with unknown sparsity is achieved; and channel coding iscombined so that the cost is reduced and the bit error rate is also reduced.

The present invention discloses a medical image registration method based on the alpha-Renyi mutual information (RMIR). The method comprises the following steps: the step of acquiring image pixel information: acquiring pixel information of a reference image and a floating image; the step of acquiring image grayscale information: acquiring grayscale information of the reference image and the floating image; the step of acquiring image mutual information: acquiring the Renyi mutual information of the reference image and the floating image; and the step of image registration: performing rigid registration on the Renyi mutual information obtained in the step of acquiring image mutual information, and obtaining the optimal transformation through the registration function, so that the Renyi mutual information reaches a maximum value. The Renyi entropy is used to introduce the mutual information metric, and the registration function is used to register, so that the registration accuracy is effectively improved; and the appropriate alpha initial parameters are found out through a large number of registration experiments, so that the operation load in the registration process is effectivelyreduced.

The invention relates to the field of bearing fault diagnosis, and discloses a weighted sliding window second-order synchronous compression S transformation bearing fault diagnosis method, which is used for improving the precision of bearing fault feature extraction and fault type identification. According to the invention, an optimal weighted sliding window module is constructed according to bearing fault vibration signal characteristics; second-order synchronous compression S transformation is fused into a plurality of step modules, a Gaussian window best matched with a fault feature signalis obtained in a self-adaptive mode in the signal time-frequency analysis process through an optimization algorithm, the weight is calculated through the obtained time-frequency analysis Renyi entropy, and then overall signal reconstruction output is achieved. The bearing fault characteristics are accurately extracted and the fault type is accurately identified through the time-frequency spectrogram, and the method is suitable for rolling bearing fault diagnosis.

The invention provides an image data dimension reduction method based on two-dimensional kernel entropy component analysis. The method comprises the following steps that (1) image data are read in; (2) a kernel function is estimated through a Parzen window; (3) a kernel matrix for calculating all the image data in row is set; (4) the eigenvalue and eigenvector of a correlation matrix of the image data are calculated; (5) the Renyi entropy of the image data is calculated; (6) the eigenvector of the correlation matrix of the image data is mapped through a two-dimensional kernel entropy component analysis method, and data dimension reduction is achieved. According to the method, through the two-dimensional analysis method, kernel conversion is directly carried out on the rows or lines of an image, the entropy estimated by the kernel matrix of the image data is sorted, the intrinsic dimension of the image data obtained after dimension reduction is obtained, and the space structure information of the image data can be kept. According to the method, due to the fact that the kernel matrix directly calculates the image data in row or in line, two-dimensional image data do not need to be converted into one-dimensional vectors, and calculation complexity is reduced when the correlation matrix is obtained through kernel conversion.

The invention discloses an integrated circuit fault diagnosis method based on information entropy. The fault diagnosis method comprises steps as follows: the characteristic that entropy information is sensitive to parameters of a to-be-tested circuit is used, a probability density function of output responses of the to-be-tested circuit is derived with the Lagrange multiplier method, then a free parameter alpha in a Renyi entropy definition formula of an output of the to-be-tested circuit is acquired through estimation with a maximum likelihood method, finally, Renyi entropy of the output of the to-be-tested circuit is acquired through computation by the aid of the probability density function and the free parameter alpha, and fault diagnosis is finished by the aid of the difference between Renyi entropy corresponding to an output of the unknown fault circuit and Renyi entropy corresponding to an output of a fault-free circuit. Compared with the prior art, the method has the advantages that the diagnosis effect on parametric faults in noises is good, the robustness is high, only a single test point is required, the method is applicable to current signals and voltage signals, and the computation complexity is low.