914 results about "Gaussian" patented technology

The present invention lies in the field of 3D surveying, mapping, and imaging. In particular, the present invention relates to the rotational alignment of 3D datasets. Embodiments include an apparatus method and program for rotational and optionally also translational alignment of 3D datasets. The 3D datasets being stored as point clouds, transformed into vector sets, and the vector sets being represented as a unit sphere or Gaussian sphere and compared for best alignment. The found best alignment is used to rotate the two 3D datasets into rotational and translational alignment with one another.

A pitch estimating method and apparatus in which mixture Gaussian distributions based on candidate pitches having high period estimating values are generated, a mixture Gaussian distribution having a high likelihood is selected and dynamic programming is executed so that the pitch of the speech signal can be accurately estimated. The pitch estimating method comprises computing a normalized autocorrelation function of a windowed signal obtained by multiplying a frame of a speech signal by a window signal and determining candidate pitches from a peak value of the normalized autocorrelation function of the windowed signal, interpolating a period of the determined candidate pitches and a period estimating value representing a length of the period, generating Gaussian distributions for the candidate pitches for each frame for which the interpolated period estimating value is greater than a first threshold value, mixing the Gaussian distributions which are located at a distance less than a second threshold value to generate mixture Gaussian distributions and selecting at least one of the mixture Gaussian distributions that a likelihood exceeding a third threshold value, and executing dynamic programming for the frames to estimate the pitch of each frame, based on the candidate pitches of each of the frames and the selected mixture Gaussian distributions.

The invention discloses a multi-background modeling method for speaker recognition, relating to a background modeling method in speaker recognition. The method is characterized by comprising the steps of: firstly, dividing training data according to a vocal tract length bending coefficient of voice, respectively training a UBM (Universal Background Model) in each group of data, then obtaining a target speaker GMM (Gaussian Mixture Model) by means of the self-adaption of each background model, obtaining multiple groups of GMM and UBM models, when recognizing a speaker, carrying out calculation on test data by each group of GMM and UBM models to obtain logarithm likelihood ratio fractions, and finally selecting a minimum logarithm likelihood ratio fraction from the logarithm likelihood ratio fractions to output. By adopting the invention, delicate depiction can be carried out on the background model, thereby improving the accuracy rate on speaker recognition.

A method of using linear array CCD to measure central position of light beam includes presenting radiation power density of CCD surface light spot in Gaussian or quasi-Gaussian distribution , setting error limit and selecting end-off threshold of digital signal output by CCD and relative circuits , applying weighted- regression method to calculate out estimation value of central position for effective digital signal of end-off quasi-Gaussian distribution and using Monte Carlo figure to simulate error distribution for setting out width range of light beam .

The present invention discloses a depth map based hand feature point detection method. The hand feature point detection method comprises the steps of: (1) acquiring a human body motion video sequence by utilizing Kinect for hand extraction, obtaining hand position information of a human body by utilizing OPENNI through the depth map, and preliminarily obtaining a palm point with a method of setting a search region and a depth threshold value; obtaining a hand contour by utilizing a find_contours function of OPENCV; accurately determining the palm point of the hand by finding the center of a maximum inscribed circle in the hand contour, and finding a maximum value M in the shortest distances by calculating the shortest distances m between all hand inner points and a contour point, wherein the hand inner point represented by M is the palm point, and the radius R of the inscribed circle is equal to M; (2)by continuously performing Gaussian smoothing on the hand contour, obtaining a CSS curvature graph in combination with a curvature threshold value, analyzing a limit value according to the CSS contour in the drawing to obtain coordinates of a finger tip point and a finger valley point of the hand, and completing the finger valley point unavailable according to the CSS curvature graph; and (3) completing a missing finger.

Provided is an apparatus and method for determining stereo disparity based on two-path dynamic programming and GGCP. The apparatus includes a pre-processing unit for analyzing texture distribution of an input image by using a Laplacian of Gaussian (LOG) filter and dividing the input image into a homogeneous region and a non-homogeneous region; a local matching unit for determining candidate disparities to be included in an each pixel of all pixels; a local post-processing unit for removing candidate disparities in a pixel of low reliability by performing a visibility test betweens candidate disparities in each pixel to improve the reliability of the candidate disparity; and a global optimizing unit for determining a final disparity for candidate disparities in an each pixel by performing a dynamic programming.

Casing signals generated by an EMAT in a borehole are processed using at least two orthogonal band-limited filters. The band-limited filters may include Gaussian or Cauchy Wavelet filters. By using the Hilbert transform, an envelope of the filtered signals is determined and amplitudes and arrival times of individual arrivals are estimated. These can be used to estimate casing and cement properties.

The invention discloses a method for controlling Gaussian projection deformation based on a normal section meridian ellipsoid. The basic technical thought is that: a new ellipsoid is constructed, so that the direction of a central meridian is basically consistent with a line extension direction, and projection deformation along a line central line extension direction is close to zero. The method comprises the following steps of: (1) constructing a normal section meridian ellipsoid E4; (2) converting the space rectangular coordinate of an ellipsoid E0 into the space rectangular coordinate of the ellipsoid E4; (3) converting the space rectangular coordinate of the ellipsoid E4 into a geodetic coordinate; (4) converting the geodetic coordinate of the ellipsoid E4 into a Gaussian plane coordinate; (5) calculating the Gaussian projection deformation value of the ellipsoid E4; and (6) comparing the Gaussian projection deformation value of the ellipsoid E4 with a projection deformation limitvalue, and verifying. By the method, the Gaussian length projection deformation of a long and large line can be effectively controlled, the number of projection zones is greatly reduced, and the method is particularly suitable for east, west, non-south and non-north long and large line engineering.

A small vocabulary speech recognizer suitable for implementation on a 16-bit fixed-point DSP is described. The input speech xt is sampled at analog-to-digital (A / D) converter 11 and the digital samples are applied to MFCC (Mel-scaled cepstrum coefficients) front end processing 13. For robustness to background noises, PMC (parallel model combination) 15 is integrated. The MFCC and Gaussian mean vectors are applied to PMC 15. The MFCC and PMC provide speech features extracted in noise and this is used to modify the HMMs. The noise adapted HMMs excluding mean vectors are applied to the search procedure to recognize the grammar. A method of computing MFCC comprises the steps of: performing dynamic Q-point computation for the preemphasis, Hamming Window, FFT, complex FFT to power spectrum and Mel scale power spectrum into filter bank steps, a log filter bank step and after the log filter bank step performing fixed Q-point computation. A polynomial fit is used to compute log2 in the log filter bank step. The method of computing PMC comprises the steps of: computing noise MFCC profile, computing cosine transform MFCC into mel-scale filter bank, converting log filter bank into linear filter bank with an exponential wherein to compute exp2 a polynomial fit is used, performing a model combination in the linear filter bank domain; and converting the noise compensated linear filter bank into MFCC by log and inverse cosine transform.

The invention relates to a method for image mosaic based on a feature detection operator of second order difference of Gaussian, which carries out mosaic on sequence images which vary in viewpoint, rotation, proportion, illumination and the like to a certain degree. The method provided by the invention adopts zero crossing point detection of second order difference of Gaussian (D<2>oG) pyramid to replace extreme point detection of the original difference of Gaussian (DoG) pyramid so as to extract scale invariant feature points, thereby effectively simplifying the structure of the Gaussian pyramid. The method comprises the steps of: first, extracting image feature points by using an improved SIFT (Scale Invariant Feature Transform) algorithm; then, searching a rough matching point pairs for the extracted feature points through a BBF (Best-Bin-First) algorithm, and purifying the feature point matching pairs by adopting an RANSAC algorithm so as to calculate an invariant transformation matrix H; and finally, completing seamless mosaic for the images by adopting a fading-in-and-out smoothing algorithm. Experimental results show that the method improves the accuracy and the real-time performance of image mosaic, can well solve problems such as illumination, rotation, scale variation, affine and the like, and realizes automatic mosaic without manual intervention.

The invention relates to the field of automatic voice recognition, in particular to a method and a device for recognizing a speaker based on tensor subspace analysis. The method comprises steps of training a general background Gaussian mixture model and a general projection matrix through voice data; then establishing a speaker model by utilizing the trained general model and the voice of an objective speaker; and finally, calculating related coefficients of the objective speaker model and a low-dimensional embedding matrix of the tested voice, and utilizing the related coefficients as a reference for recognizing the speaker. The device comprises a voice preprocessing module, a feature extracting module, the trained general background Gaussian mixture model, an adaptive module, the trained general projection matrix module, a low-dimensional embedding calculating module, a related coefficient calculating module, a score judging module and a storage module. The method and the device can obviously reduce the amount of calculation when the speaker modal is established, have the advantages of low modal freedom, high robustness and the like, and is applicable to recognizing speakers unrelated to a text under the condition that the length of voice is limited.

The subject invention leverages standard probabilistic inference techniques to determine a log-likelihood for a conditional Gaussian graphical model of a data set with at least one continuous variable and with data not observed for at least one of the variables. This provides an efficient means to compute gradients for CG models with continuous variables and incomplete data observations. The subject invention allows gradient-based optimization processes to employ gradients to iteratively adapt parameters of models in order to improve incomplete data log-likelihoods and identify maximum likelihood estimates (MLE) and / or local maxima of the incomplete data log-likelihoods. Conditional Gaussian local gradients along with conditional multinomial local gradients determined by the subject invention can be utilized to facilitate in providing parameter gradients for full conditional Gaussian models.

The invention relates to a Mask R-CNN-based reinforced concrete crack identification and measurement method, which comprises the following steps of obtaining a to-be-detected picture which is a picture with a concrete crack; searching and positioning a crack region on the to-be-detected picture based on Mask R-CNN, and segmenting a crack target in the to-be-detected picture; carrying out edge detection on the crack target through combination of a Sobel edge filtering operator, a Laplacian filtering operator and a gaussian smoothing filter, and generating a crack mask; and calculating the length and width of the crack and the crack according to the pole coordinates of the generated crack mask and the minimum bounding rectangle of the mask. Compared with other technologies, the method has relatively high efficiency in crack measurement, has more accurate extraction of crack pixels, and realizes crack detection and identification and calculation of crack length, width and area.

The invention discloses an atmospheric pollution traceability diffusion analysis system and method based on a Gaussian diffusion model, and the system employs an analogue simulation module which comprises GIS information, pollution diffusion simulation and visual rendering, carries out the simulation of a regional environment through the information, forms a GIS map, and according to the monitoring data in combination with the Gaussian diffusion model after terrain correction, through an image processing technology, pollution diffusion concentration changes are overlaid on a GIS map, wind field distribution after topographic correction is obtained according to topographic data and meteorological data, then a Gaussian diffusion model formula is selected, the concentration influence of each pollution source on a grid center point of an evaluation area is obtained, and an accumulated concentration value is calculated. And visual rendering is performed on a GIS map according to the calculated pollutant concentration value of the central point of the grid, and a visual result is superposed with a basic geographic information base map of the evaluation area to realize visual display of a prediction result.

The subject invention leverages the conditional Gaussian (CG) nature of a continuous variable stochastic ARMAxp time series model to efficiently determine its parametric gradients. The determined gradients permit an easy means to construct a parametric structure for the time series model. This provides a gradient-based alternative to the expectation maximization (EM) process for learning parameters of the stochastic ARMAxp time series model. Thus, gradients for parameters can be computed and utilized with a gradient-based learning method for estimating the parameters. This allows values of continuous observations in a time series to be predicted utilizing the stochastic ARMAxp time series model, providing efficient and accurate predictions.

The invention relates to a lane changing model parameter optimization method based on a mixed Gaussian-hidden Markov model. Operation states of a vehicle and surrounding vehicles are analyzed, and a driving lane changing intention recognition method based on the hidden Markov model is established, that is, prediction of an implicit driver lane changing intention is achieved by using lane changingcharacterization parameter observation states capable of being observed; a highway vehicle database with concentrated NGSIM data serves as the basis, a lane changing characterization parameter sampleis extracted, training and verification are carried out on the hidden Markov model by using an HMM tool box programming algorithm in MATLAB, and the number of mixed Gaussian models and the number of states are optimized by using a particle swarm algorithm. The number of the mixed Gaussian models and the number of the states are optimized by using the particle swarm algorithm, so that the accuracyof driver lane changing intention recognition is improved; better accuracy is finally obtained for the prediction result of the driver lane change intention.

A fast accurate multi-channel frequency dependent scheme for analyzing noise in a signal processing system is described herein. Noise is decomposed within each channel into frequency bands and sub-band noise is propagated. To avoid the computational complexity of a convolution, traditional methods either assume the noise to be white, at any point in the signal processing pipeline, or they just ignore spatial operations. By assuming the noise to be white within each frequency band, it is possible to propagate any type of noise (white, colored, Gaussian, non-Gaussian and others) across a spatial transformation in a very fast and accurate manner. To demonstrate the efficacy of this technique, noise propagation is considered across various spatial operations in an image processing pipeline. Furthermore, the computational complexity is a very small fraction of the computational cost of propagating an image through a signal processing system.

The invention relates to a calculation method for predicting the multi-polar expansion attribute of a dipeptide model through a BP neural network. The calculation method for predicting the multi-polar expansion attribute of the dipeptide model through the BP neural network comprises the following steps that the structures of different dipeptide conformations are optimized through quantum mechanics calculation software Gaussian, and physicochemical parameters of the dipeptide conformations and the distance between atoms are calculated; the physicochemical parameters of part of the atoms of the dipeptide conformations and the distance between the atoms are selected for training the BP neural network, so that the physicochemical parameter of the BP neural network is obtained; the other dipeptide conformations serve as a testing set to verify a prediction result based on the BP neural network. According to the calculation method for predicting the multi-polar expansion attribute of the dipeptide model through the BP neural network, quantum mechanics calculation is conducted through the BP neural network instead of quantum mechanics calculation Gaussian software, the physicochemical parameter information such as the energy and the multi-polar distance of dipeptide can be quickly provided according to different conformations on the basis of molecular mechanics simulation based on force field information. Calculation time is greatly shortened and calculation quantity is greatly reduced within the acceptable error range, and the precision in the dynamic simulation process is greatly improved.

The invention discloses a joint inversion method based on magnetotelluric and DC resistivity data, and the method comprises the steps: determining an inversion region according to observation data, carrying out the subdivision to obtain an initial grid, and setting an initial value of an inversion parameter vector; performing Gaussian-Newton inversion iteration based on the current value of the inversion parameter vector to calculate a model change amount and a linear search step length, and then calculating an iterative update value of the inversion parameter vector; judging whether an inversion termination condition is met or not, and if not, carrying out next iteration; if so, judging whether a progressive grid inversion termination condition is reached or not, and if so, ending the inversion; and if not, carrying out refined subdivision on the grid cell, updating the inversion parameter vector value, and returning to carry out Gauss-Newton inversion iteration until inversion is finished. According to the invention, a joint inversion technology of inversion grid adaptive adjustment is researched and developed, the problem of multiplicity of solutions of joint inversion interpretation of a magnetotelluric method and direct current resistivity data is effectively reduced, and the joint interpretation accuracy is improved.

Systems and methods for automatic generation and efficient exploration of a large number of webpage layouts to discover a layout with superior empirical performance. A set of variants are displayed to visitors in accordance with a display probability distribution. Data related to visitors' interactions to the variants are collected and processed to evaluate their respective profit-related performances. The display probability distribution may be dynamically adjusted based on the profit-based evaluation. A profit brought by a webpage layout may be ascribed to a number of revenue sources. These difference revenues may be tracked and summed together to yield a profit assessment for a layout variant. Profit performance of a layout variant may be calculated using a Gaussian with NΓ−1 prior model, or a Gaussian-Dirac delta mixture mode.

The invention discloses a grid feature point extraction method based on fast line extraction, and the method is used for extracting coordinates of a feature point in a grid image quickly and accurately. The execution steps are as follows: firstly carrying out the binary processing of an image, enabling an initial color image to be converted into a grey-scale map, and reducing a calculation quantity; secondly sequentially carrying out the erosion and dilation conversion and Gaussian filtering of the image, and eliminating interference and noises; carrying out the edge detection of a preprocessed image through employing a Canny edge detector, obtaining a grid contour, then carrying out fast line extraction through employing improved Hough conversion, and obtaining a line set; enabling the line set to be divided into two classes according to the slope of a line, and reducing the number of times of intersection point solving operation; enabling lines which are perpendicular to each other to be combined together so as to obtain the coordinates of intersection points, and obtaining a feature point set; setting a threshold value for the aggregation of the extracted feature points, eliminating interference, and finally obtaining a precise feature point set. The method is higher in speed and precision of feature point extraction, can eliminate the interference, and is better in fault tolerance performance.

System and method for performing Time Domain Reflectometry (TDR) on a Device Under Test (DUT) using Gaussian pulses. A signal is received comprising an initial Gaussian pulse and one or more reflected pulses from the DUT. Each pulse is characterized by determining a set of estimated parameters, permuting the estimated parameter set to generate one or more permuted parameter sets, generating linear equations from the parameter sets, including parameter variables for the corresponding Gaussian pulse, and determining values for the parameter variables by solving the linear equations. The determined parameters characterize the Gaussian pulse. If there are N parameters to determine and M permutations generated, where M is greater than or equal to N, M+1 linear equations are solved to overdetermine the N parameters. The determined parameters of the initial pulse and the one or more reflected pulses are useable to perform TDR analysis on the DUT.

The invention discloses an abnormal traffic monitoring method based on statistics, which comprises the following steps: collecting user access log records in a preset time period, and carrying out cleaning and transformation processing to generate standard user access data; counting the distribution of statistical characteristics corresponding to the standard user access data in different time dimensions; mapping the distribution of the statistical characteristics on different time dimensions into corresponding multivariate Gaussian distribution and respectively performing parameter estimation; calculating Gaussian distribution probability values respectively corresponding to the statistical characteristics corresponding to the current network flow in each time dimension; judging whether the Gaussian distribution probability value is smaller than a preset alarm threshold value of the current network flow in the time dimension or not; If so, judging that the current network traffic is abnormal traffic. The invention further discloses an abnormal flow monitoring device and equipment based on statistics and a storage medium. The method is easy to deploy and low in implementation cost,and can flexibly respond to abnormal flow real-time alarms of different service scenes in different time periods.

The invention discloses a semantic sparse Web service discovery method based on Gaussian LDA and word embedding, and specifically relates to the technical field of service computing. The method specifically comprises the following steps in order: collecting a Web service description document, preprocessing the collected Web service description document, obtaining characteristic vocabularies in the Web service description document, preprocessing the vocabularies in the Web service description document, and obtaining a set of a group of prototype words; using a word embedded training model Word2Vec to train the set of the prototype words obtained in step 1, and obtaining a continuous vector representation of each word in the set of the prototype words; using a Gaussian LDA model to train the set of the prototype words obtained in step 1, and acquiring each Web service hierarchical structure; enriching user query by using a set of continuous vectors obtained in step 2 by training, and obtaining extended user query; and using the service hierarchical structure obtained in step 2 and using a probability ranking method to obtain query output corresponding to the extended user query.

The invention relates to a motor bearing fault diagnosis method based on generalized S transformation and WOA-SVM, and the method comprises the steps: inputting a motor bearing vibration signal, and obtaining two time-frequency matrixes after two times of different generalized S transformation; respectively obtaining a time domain cumulative characteristic curve with high time resolution and a frequency domain cumulative characteristic curve with high frequency resolution; obtaining a time domain feature and a frequency domain feature of the original signal; combining the time domain featuresand the frequency domain features to form a feature vector sample set, and dividing the feature vector sample set into training samples and test samples; inputting the training sample into a support vector machine optimized by a whale optimization algorithm WOA, and training a classifier; and inputting the test sample into the trained classifier WOA-SVM for testing, and outputting a fault diagnosis type. The method overcomes the defect that the Gaussian window function of S transformation cannot be adjusted along with the frequency and lacks flexibility, has better time-frequency analysis capability, and is more suitable for processing complex non-stationary and non-linear bearing vibration signals.

The invention discloses a tumor prognosis prediction system based on a deep belief network. The system comprises a data collecting module, a data preprocessing module, a data learning prediction module and a prediction result display module, wherein the data collection module is used for collecting tumor information; the data preprocessing module is used for carrying out missing value processing and normalization processing on tumor original data; the data learning prediction module is used for carrying out deep learning and prediction modeling on the tumor data; and the prediction result display module is used for displaying a relative risk output by the data learning prediction module. A Gaussian restricted Boltzmann machine is used to keep the nonlinear characteristics of the data; according to the dimension of input data, the amount of output categories and the accuracy of a model, the deep belief network can be flexibly expanded; and in a model training process, no restriction and hypothesis is adopted, an influence way of a variable for the result and a mutual function among variables can be fully mined, the influence way of different factors for tumor prognosis can be comprehensively revealed, and tumor prognosis prediction accuracy is improved.

The invention discloses a table recognition method which comprises the following steps: acquiring an initial table image, and performing horizontal correction and vertical correction on the initial table image to generate a corrected table image; converting the correction table image into a grey-scale map, generating a grey-scale table image, and performing Gaussian adaptive binarization processing on the grey-scale table image to generate a foreground mask; performing corrosion operation and expansion operation on the foreground mask by adopting a method based on combined morphology to extract a table line, and generating a table foreground; searching the outline of the table foreground to generate a table area; and identifying the text image line of each table area by using a convolutional neural network, and outputting an identified character. By adopting the method, inclination correction processing is carried out when the table image is identified, and shape distortion and unevenbrightness are processed, so that the table identification accuracy is improved.

The invention discloses an area-based remote sensing image registration method, and mainly solves the problems that registration of multispectral images with high image gray scale difference cannot be performed by existing algorithms. The technical scheme is that 1. a Gaussian scale space is established for a reference image and an image under registration respectively; 2. a sketch graph of the reference image and the image under registration is generated respectively; 3. the Gaussian scale space graphs and the corresponding sketch graphs of the reference image and the image under registration generate respective structural areas and non-structural areas; 4. the features points of the structural areas and the non-structural areas are detected; 5. the features points of the two structural areas of the reference image and the image under registration are added; 6. descriptors of the added feature points are obtained by using an SIFT algorithm, and registration parameters are obtained by utilizing the feature points descriptors; and 7. affine transformation is performed on the image under registration by utilizing the registration parameters so that a registration result is obtained. Registration of multi-source and multispectral remote sensing images with high gray scale difference can be realized.

A first party has a data vector x and a second party has a classifier defined as a set of multivariate Gaussian distributions. A secure inner dot product procedure is applied to each multivariate Gaussian distribution and the data vector x to produce a vector ai for the first party and a vector bi for the second party for each application. The secure inner dot product is then applied to each vector bi and the data vector x to produce a scalar ri for the first party and a scalar qi for the second party for each application. A summed vector of elements [(a1x+q1+r1), . . . , (aNx+qN+rN)] is formed, and an index I of the summed vector for a particular element having a maximum value is the class of the data vector x.

The invention discloses a full-flow-field simulation method based on a central five-order WENO format of a staggered grid. The invention discloses a full-flow-field simulation method based on a five-order basic weighted oscillation-free format under a central framework. The full-flow-field simulation method comprises the following steps: reconstructing a variable unit mean value at a half point byadopting a finite volume five-order WENO format in a new form; discretizing the time integral term by using a Gauss quadrature formula to obtain a Gaussian node combination of flux with respect to time at the whole point of space; reconstructing a variable point value at the integral point; reconstructing a value of a derivative of the flux at an integer point; obtaining a Gaussian node of the flux with respect to time at the whole space point; combining the unit mean value variable at the half point with the Gaussian node of the flux with respect to time at the whole point of space to obtaina unit mean value at the half point of the next time layer; and sequentially iterating to obtain a numerical result of the flow field at the termination moment in the calculation area. According to the full-flow-field simulation method, better precision can be achieved; better capture discontinuity is achieved, and meanwhile, part of problems which cannot be solved by a traditional algorithm canbe simulated; and the double-Mach reflection problem and the step problem which cannot be well solved by a traditional algorithm under the two-dimensional condition are solved.