65 results about "Generalized inverse matrix" patented technology

An estimation apparatus for estimating a channel response of a radio propagation path using a received signal including a first known signal, comprises a generator which generates a reference signal matrix, a calculator which calculates a generalized inverse matrix of the reference signal matrix including singular values which exceed a preset threshold value, an estimation unit configured to estimate an impulse response of the radio channel modeled using a transversal filter based on the first known signal and the generalized inverse matrix, and a converter which converts the estimated impulse response into a frequency-domain signal to acquire a frequency transfer function of the channel.

In a wireless communication system including a radio-signal receiver and a radio-signal transmitter, the radio-signal receiver includes a unit for making a generalized inverse matrix for carrying out maximum likelihood estimation, based on a matrix indicating a training series used for carrying out channel estimation, a unit for carrying out singular value decomposition to the generalized inverse matrix, a unit for truncating singular values in the generalized inverse matrix in accordance with a threshold by which a singular value is turned to zero in a greater order, a unit for estimating an impulse-response vector, based on the generalized inverse matrix, a unit for interpolating zero to the impulse-response vector, and a unit for acquiring channel estimation by carrying out Fourier transform, and the radio-signal transmitter includes a unit for quantifying spread of the singular values, and optimizing the training series in accordance with the quantified spread of the singular values, to thereby make a new training series.

The invention relates to a high-precision method for quickly positioning centers of two-dimensional Gaussian distribution spot images, which includes obtaining an approximate two-dimensional Gaussian distribution image through defocused imaging of a CCD (charge coupled device) camera, processing the image to preliminarily split multiple light spots in the image and mark serial numbers, determining the integer pixel coordinate of the light spot center before setting up local image coordinate system with the integer pixel coordinate serving as an origin, calculating variance of the Gaussian light spot and sub-pixel coordinate components of the light spot center in the x and y directions, and compositing the integer pixel coordinate with the sub-pixel coordinate to obtain the actual precise coordinate of the light spot center. All pixel grayscale information in a window is comprehensively utilized by the method, position of the sub-pixel center of the Gaussian distribution light spot canbe directly calculated through an analytic expression, solving of a generalized inverse matrix is not needed, efficiency is improved by more than 200 times and precision loss is avoided as compared with those of the traditional Gaussian quadrics fitting method, and real-time extracting of multiple light spot centers can be realized.

The invention discloses a single focal plane high-precision testing method for optical wavefront of an optical imaging system, relates to the technical field of optical testing, solves the problems that exit pupil amplitudes are not distributed uniformly and calculation errors are introduced by fast Fourier transform in the conventional phase retrieval algorithm, and provides the scheme for eliminating the influence of vibration in the process of image acquisition on detection accuracy. The method comprises the following steps of: establishing a detection platform of the optical imaging system; detecting the position of the focal plane of a lens to be detected by using a detection device in the detection platform and acquiring an out-of-focus stellar image of the lens to be detected by the detection device; selecting effective data according to the acquired out-of-focus stellar image and calculating a pupil function of an optical system; and extracting the phase of the acquired pupil function to obtain the optical wavefront of the optical imaging system. The pupil function of the optical system is calculated by a Zernike multinomial, an extended Nijboer-Zernike multinomial, and a generalized inverse matrix. The single focal plane high-precision testing method is low in cost, and high in accuracy and is suitable for manufacturing enterprises, scientific research and detection units of the optical imaging system.

The invention provides a one-dimensional mirror image synthetic aperture radiation imaging method. The method comprises the following steps of: constructing a linear equation set by utilizing the relation among relevant output, the cosine visibility and the scene brightness temperature distribution, solving the linear equation set to obtain a cosine visuality; and then reconstructing scene brightness temperature distribution through a mode of carrying out anti-cosine transformation or solving a generalized inverse matrix of an impulse response matrix. In the invention, the imaging performance of a traditional synthetic aperture imaging system with a large array element number is acquired by using a small array element number, thereby the complexity of the structure of the traditional synthetic aperture imaging system and the complexity of signal processing are effectively reduced.

A MIMO decoder, which is capable of changing a search area of a transmission signal vector in accordance with a change in a channel matrix, includes: a generalized inverse vector matrix calculation unit for calculating a generalized inverse matrix of Moore-Penrose derived from a channel matrix indicative of a radiowave propagation environment; a search area limiting processing unit for performing weighting for each eigenvector calculated from the channel matrix in inverse proportion to a square root of an eigenvalue corresponding to the eigenvector, and determines the search area of the transmission signal vector centered at the generalized inverse matrix solution of Moore-Penrose based on the weighted result; and a most likelihood estimation unit for searching for a transmission signal vector by use of a most likelihood estimation based on the search area determined by the search area limiting processing unit.

The invention discloses a redundant freedom robot inverse kinematics method. The redundant freedom robot inverse kinematics method comprises the steps of establishing a forward kinematics equation; establishing an MP generalized inverse matrix equation; according to a generalized MP inverse matrix relationship, establishing a Newton iteration equation set; performing calculation according to the Newton iteration equation set to obtain partial joint position values; according to the partial joint position values, calculating residual joint position values; and performing selection on multiple joint solutions obtained by calculation. According to the redundant freedom robot inverse kinematics method provided by the invention, the time problem and the solution screening problem caused by seven freedom kinematics redundant axis calculation are solved by adopting an MP inverse matrix solution method, so that a mapping relationship between a pose of a gripper at the end of an operation arm of a robot and a joint vector is established.

The invention discloses a method for eliminating a dead zone by using a generalized inverse matrix. The method comprises the following steps: modeling a magnetic target as a magnetic dipole model, solving the generalized inverse matrix of the gradient tensor matrix of the magnetic field of a magnetic dipole when the magnetic moment vector of the magnetic dipole is perpendicular to the position vector of the magnetic dipole, and substituting the generalized inverse matrix into a positioning formula to obtain an accurate positioning result of the magnetic target. The method has the advantages of further improvement of the practicability of the magnetic target detection and positioning by eliminating the detection and positioning dead zone through using the generalized inverse matrix, and realization of the detection of the magnetic target in any direction.

The invention discloses a raw cement material quality control method and system based on a generalized inverse matrix. The method comprises steps as follows: S1, determining a target value expected by raw cement material quality; S2, determining the required proportion of all raw materials in a raw cement material and giving the relation between the raw materials and effective ingredients; S3, calculating the corresponding relation between the expected target value and the raw materials by using a generalized inverse matrix theory according to the raw materials, the effective ingredient content and the expected target value; S4, adjusting the proportion of the raw materials by using the corresponding relation obtained in Step 3 according to a coefficient matrix of the corresponding relation among the expected target value, a given target value and the effective ingredients as well as a coefficient matrix of the effective ingredients contained in the raw materials. According to the method, the proportion of the raw materials is calculated according to the given target value, the stability of ingredients of the ground raw material is guaranteed, batching requirements under different process conditions are met, the proportion of the raw materials is adjusted in real time, the raw material rate fluctuation is reduced, and the quality of the raw materials and the robustness of the whole raw material quality control system are guaranteed.

A power grid multiple major failure recognition method based on a linear weighting method comprises the steps as follows: establishing an electrical power system graph theory model and a Laplacian matrix corresponding to the electrical power system graph theory model according to a topological connection relation; establishing a node active power matrix according to a node property and size of an active power; establishing multiple major failure objective function by using the linear weighting method; calculating product of a generalized inverse matrix of the Laplacian matrix and the node active power matrix; establishing a failure preliminary screening matrix according to a vector corresponding to an exact solution; calculating fault component number corresponding to each column vector in the failure preliminary screening matrix; determining weight coefficient in the objective function according to concern extent of an operating crew and screening the column vector which enables the objective function to obtain a minimum value. The power grid multiple major failure recognition method based on the linear weighting method of the invention has important guiding significance for formulating effective prevention and control measures and solving a column control strategy to recognize the multiple major failure.

The invention relates to an electric-thermal coupled integrated energy system optimal current calculation method based on a generalized inverse method and belongs to the field of power / integrated energy system operation control technologies. According to the method, firstly, a steady-state model of a regional electric-thermal coupled integrated energy system is established; then, a multi-target optimal current model of the regional integrated energy system is established based on a generalized inverse matrix theory, wherein multiple targets include minimum energy utilization cost and minimum power distribution system network loss; and solving is performed iteratively based on a Newton Raphson method, and information transmission between a power system and a thermodynamic system is realizedthrough coupling equipment operating power conversion. Through the method, operating conditions of the regional integrated energy system can be accurately reflected, the method has good numerical stability and convergence, the optimization process is an optimization process of a feasible solution, a more optimal feasible solution can be obtained in each step, calculation can be stopped anytime according to the actual situation, and the method is suitable for real-time system optimization, scheduling and application.

The invention provides a prediction method and device of movement track. The method comprises the steps that the n+1 frame before the frame to be predicted and the corresponding target coordinate areselected as the reference frame and the reference coordinate; a generalized inverse matrix can be established based on the n frame before the frame to be predicted and the reference frame; a coordinate matrix can be established based on the n frame before the frame to be predicted in the video and the said reference frame; a secondary matrix can be established based on the frame to be predicted inthe video and the reference frame; a target track prediction model can be established based on the generalized inverse matrix, the coordinate matrix, and the frame matrix; a coordinate of a mobile object in the frame image to be predicted can be predicted based on the target track prediction module. The invention is advantageous in that a more accurate track prediction can be provided, and a lotof calculation can be prevented.

The invention discloses an equation solver based on memristor arrays, and an operation method thereof. The equation solver comprises a first memristor array, a second memristor array, a comparison module and a conductivity modulation module, a three-layer neural network is established based on the memristor arrays, an inverse matrix or a generalized inverse matrix of a coefficient matrix of the system of linear equations is solved based on a training algorithm of a neural network, so the purpose of solving the equations is achieved. The equation solver can be used for solving various tasks such as a non-singular linear equation set, a compatible linear equation set, an incompatible linear equation set, a partial differential equation and a matrix feature vector, and is high in reconfigurability. Besides, the memristors have the advantages of high speed, low power consumption, easiness in integration, compatibility with a CMOS process and the like, so that vector matrix multiplication can be realized by conductively storing a data matrix for operation. The equation solver is simple in circuit structure, small in circuit area, low in circuit power consumption and simple in operationstep.

The invention discloses a PWM (Pulse-Width Modulation) rectifier fuzzy sliding mode variable structure control method when the three-phase network voltage is unbalanced, and belongs to the field of power electronics application. The method comprises the following steps of firstly sampling the direct current side voltage of a rectifier; obtaining a rated active power through a PI (Proportional-Integral) controller; sampling the network voltage and the rectifier current, and using a trap filter for performing positive and negative sequence decomposition on the sampling data; listing and writing the power matrix of the rated values by using the rated power conservation principle; obtaining the rated current through a generalized inverse matrix conversion formula; and finally, obtaining a space voltage vector by the fuzzy sliding mode variable structure control method on an electric current loop. The control method has the advantages that the current loop uses the variable structure control, so that the robustness and the dynamic performance of a system can be enhanced; meanwhile, the variable structure coefficient is regulated through fuzzy operation on state points; the adverse influence due to buffeting is reduced; and the stability of the system is improved.

The invention discloses a structural stress sensitivity solving method. The structural stress sensitivity solving method comprises the following steps of 1: performing structural finite element analysis calculation, and selecting out a displacement value corresponding to a current unit node from a structural displacement array; 2: calculating a generalized inverse matrix of a global stiffness matrix; 3: establishing design variables, taking an area or thickness of a metal bar plate beam, and combining single layers, with same ply angles, of materials of a laminated plate with composite material layers into single design variables; 4: selecting out stress components currently serving as constraints, and determining to establish a selection matrix according to the type of a current unit andstress components of calculated sensitivity; 5: calculating a differential stiffness matrix corresponding to the current unit; 6: through a series of matrix multiplication operations of the generalized inverse matrix, the stress selection matrix, a stress matrix, a conversion matrix from global coordinates to unit coordinates, a displacement selection matrix, the differential stiffness matrix of units to design and the displacement array, performing sensitivity calculation of the stress of the current unit to the design variables; and 7: circulating the steps 4-6 according to the units to finish sensitivity solving of stress constraints of all the units.

The invention discloses a combined sparse signal dimension reduction reconstruction improvement method based on a compressed sensing theory, and belongs to the field of analog information conversion. According to the method, an original multi-measurement vector is converted into a low-dimension multi-measurement vector, then a sparse solution is recovered from the low-dimension multi-measurement vector by using a gradient tracking algorithm, and dimension reduction reconstruction of signals is realized. According to a correlation theorem and parameter setting requirements based on a compressed sensing theory, a unique solution can be recovered when the signal width of a joint sparse signal with known sparseness reaches a certain value (called as a critical value L1) through theoretical derivation, so that when a high-dimensional signal is reconstructed, the signal width is reduced to L1 and then reconstruction is carried out. During reconstruction, a gradient tracking algorithm is used, calculation of an inverse matrix or a generalized inverse matrix is replaced by a gradient thought in an unconstrained optimization method, and QR decomposition does not need to be used. According to the method, the calculation complexity is reduced, the reconstruction success rate is improved, and the larger the signal width is, the more obvious advantages are achieved.

The embodiment of the invention discloses a frequency offset estimation method and device, relating to the technical field of wireless communication, which are used for improving the accuracy of the result of frequency offset estimation. The frequency offset estimation method comprises the following steps of: a receiving end receiving a signal and a training serial number sent by a sending end, obtaining a training sequence according to the training serial number, and constructing a generalized inverse matrix of a correlation matrix of the training sequence; generating the correlation matrix constructed by a received signal; obtaining a channel impulse response correlation vector with frequency offset according to the generalized inverse matrix and the correlation matrix constructed by the received signal; and carrying out frequency offset estimation according to the correlation vector. Based on a multipath transmission model, channel estimation results are unnecessary for the frequency offset estimation; and the accuracy of result of frequency offset estimation can be efficiently improved by adopting the frequency offset estimation method and device.

The invention brings forward a spectrum reconstruction method based on a reverse dual light path. A target spectrum image after forward dispersion and the target spectrum image after backward dispersion are obtained by utilizing a reverse dual light path structure; a generalized inverse matrix of an equivalent data matrix of slit movement and superposition effect in the reverse dual light path is applied to respectively act on a forward dispersion spectrum data matrix and a backward dispersion spectrum data matrix so that a forward spectrum data matrix approximate value and a backward spectrum data matrix approximate value of a target scene are obtained; and then solution is obtained via a cross iteration mode so that the spectrum data matrix of the target scene after reconstruction is obtained. A high amount of light transmission can be ensured and spectrum resolution can also be effectively enhanced by the spectrum reconstruction method.

The invention discloses a method for rapidly assessing multi-antenna system channel capacity stability. Aiming at the channel capacity stability problems in various wireless communication systems, the method includes that the influence of the spatial correlation of a MIMO system and the coupling effect of an antenna array element is comprehensively considered at a receiving end to obtain a channel transmission matrix of the MIMO system, and a generalized inverse matrix of the channel transmission matrix is constructed; and according to the related knowledge of the matrix analysis, the F-norm theory analytic expression of the inverse matrix, namely, the F inverse norm of the channel matrix, is calculated theoretically, and according to the changing trend of the F inverse norm of the channel matrix, the trend that the channel capacity changes along with the system key parameters can be seen. The method fully considers all of the feature values of the channel matrix, therefore, the change trend of the future wireless communication system channel capacity can be accurately estimated according to the method.

The invention provides a near-field acoustical holography reconstruction method based on the Fourier transformation and source strength simulation technology. The near-field acoustical holography reconstruction method based on the Fourier transformation and source strength simulation technology comprises the steps that continuous simple sources are arranged on a rectangular virtual source surface, and a source strength function is a three-dimensional free space Green function; double Fourier series expansion is carried out on an unknown source strength density function, and a sound pressure expression is deduced; expressions of the sound pressure on measuring points on a holographic surface are deduced through quick Fourier transformation and inverse transformation calculation, and a non-homogeneous linear equation set which uses a Fourier series expansion coefficient as an unknown term and the sound pressure as a constant term is formed through arrangement; the Fourier series expansion coefficient is obtained by solving the non-homogeneous linear equation set through a generalized inverse matrix, the Fourier series expansion coefficient is substituted into a reconstruction formula in a back substitution mode, the expression of the sound pressure on any field point in a radiation field is obtained, and reconstruction and prediction of the whole space sound field are achieved. According to the near-field acoustical holography reconstruction method based on the Fourier transformation and source strength simulation technology, implementation is easy, the calculating speed is high, limitation cannot be caused by the limited hole diameter, and the near-field acoustical holography reconstruction method is applicable to sound sources with any shapes.

The invention discloses a modeling application of an auto-encoder-based extreme learning machine in industrial production prediction, which comprises the steps of encoding input data by using the auto-encoder to obtain main characteristics of original data, and removing noise and redundant information in the data; comparing the auto-encoder reconstruction losses corresponding to different auto-encoder structures according to the average relative error, selecting the auto-encoder structure with the minimum loss as a final auto-encoder, and using the selected auto-encoder to encode the original data as the input of an extreme learning machine. According to output weight of the extreme learning machine calculated by Moore-Penrose generalized inverse matrix, the final output of the extreme learning machine is obtained. According to the method, noise and redundant information in the data are removed, the main characteristics extracted by the auto-encoder are used as the input of the extreme learning machine, the sensitivity of the extreme learning machine to hidden layer node selection is reduced, and the precision and robustness of model prediction are improved.

The invention discloses a millimeter wave radar angle super-resolution method. The method comprises the following steps: acquiring sampling data; performing FFT processing on the sampling data to obtain a distance phase sequence and a first signal covariance matrix; adding the first signal covariance matrix, and reconstructing to obtain a second signal covariance matrix; decomposing the second signal covariance matrix into a signal subspace matrix and a noise subspace matrix, and solving a weight vector; reconstructing a noise subspace matrix by using the weight vector, and processing a signal subspace matrix by using a generalized inverse matrix; and forming a spatial spectral density function equation by using the subspace matrix, wherein the angle is traversed to find a peak to estimate the angle. The millimeter wave radar is low in transmitting power, high in resolution and sensitivity and small in radar size, the MUSIC algorithm is optimized through information of the signal subspace and the noise subspace, the angular resolution is improved, and the detection capacity of the radar in a long-distance mode is remarkably improved.

The invention relates to the technical field of signal processing, and particularly relates to a NPCA-based post nonlinear blind source separation method. The NPCA-based post nonlinear blind source separation method of the invention comprises the following steps: step 1, observation signals of T sampling points are loaded; step 2, W^(t) is initialized, and elements thereof are randomly generated within [-11]; step 3, whether or not t<=T is judged, step 4 is executed if t<=T, or step 8 is executed; step 4, y(t) optimal estimation on the source signals is carried out, namely, the generalized inverse matrix W^(t) of an optimal time-varying system A^(t) is solved; step 5, column vectors of W^(t) are stacked to form a column vector with N<2> dimensions, W<rightward arrow>(t) with n<2>*n<2> dimensions is constructed, and a nonlinear function suitable for source signal separation is selected; step 7, the optimal W<rightward arrow>(t) is solved by a H-infinite adaptive algorithm; and step 8, y(t) of the T sampling points is output when t>T. Signals mixed and stacked in a post nonlinear system and subject to any distribution can be separated by the NPCA-based post nonlinear blind source separation method, and the method is applicable to online blind source separation.

The invention relates to a hyperspectral unmixing optical implementation method based on spectral imaging technology. The hyperspectral unmixing optical implementation method based on the spectral imaging technology solves the problems that a data cube needs to be reconstructed for existing hyperspectral unmixing, and the data cube comprises the massive amount of data information, so that a data transmission system, a storage system and a processing system have large burden and low spectral unmixing efficiency. The method comprises the following steps that (1) light of a spatial pixel is projected onto a wide-spectrum coding filter by an optical splitting system; (2) an end-member matrix E is formed, a generalized inverse matrix E<#> of the end-member matrix E is obtained, i is set to be equal to 0, and the number of end members is m; (3) i plus 1, if i is less than or equal to m, the step (4) is performed, otherwise, the step (8) is performed; (4) encoding control is performed on a DMD; (5) spectral modulation is performed on a spectrum after optical splitting through the DMD; (6) the modulated spectrum is projected onto a detector; (7) the detector obtains an abundance coefficient of the preceding spatial pixel for the end members, and the step (3) is performed; and (8) after spectral unmixing of a current line is completed, the spectral unmixing process is performed on a next line.

The invention relates to a face recognition method that suits for the small sample, which uses two-dimensional matrix, to establish sample data matrix, similar member matrix, getting typical correlation matrix and finally generalized inverse matrix. It solves the problem that the queer covariance matrix is added artificial disturbance to process correction, which brings disturbance data introducing more random and uncertainty, making the result of small sample is not satisfied, leading to a significant increase in calculation, discarding the image null space ahead due to the method of pre-treating samples by dimensionality reduction, losing some important distinguishing information. The invention replaces number 1 with a matrix, that is the same size as sample image, to make the similar member matrix not only reflect the affiliation on each sample and any kind, but retain the space information of rows and columns in the sample images.

The invention provides a precoding method and device, relates to the technical field of communication, and can improve the power utilization rate of an antenna. The method comprises the following steps: determining a generalized inverse matrix m = 1, 2,..., J of a channel value matrix Hm of K users in the mth iteration of J times of iteration operation, J being greater than or equal to 1 and lessthan or equal to M-K + 1, J being an integer, and M being the number of antennas; selecting an antenna index nm and an updating coefficient alpha m corresponding to the antenna index nm according to the set S1m-1, wherein S1m-1 is a set of n which is not selected in the antenna indexes n of the M antennas until the m-1th iteration is finished; determining a weight matrix Wm according to the sum alpha m; and assigning a row corresponding to nm in the Wm to a row corresponding to nm in a final weight matrix Wopt, the Wopt being used for adjusting the transmitting power of at least J antennas inthe M antennas to a preset maximum transmitting power P after the J iterative operations are finished.

The invention discloses a quantitative calculation method and a quantitative calculation system for the vibration response of a complicated combination shell structure in a fluid medium. The method comprises the steps of obtaining structure parameters and excitation parameters of a combined shell system in a fluid medium; carrying out point collocation on the surface of the combined shell in a bus-bar direction, calculating filed transfer matrixes and point transfer matrixes between all collocation points of the combined shell system by a fine integration method, and calculating an external force vector; building a dynamical model of a fluid medium-combined shell system according to the boundary conditions and structure continuing conditions of the two ends of the combined shell system; and solving to obtain the vibration response of the fluid medium-combined shell system through an Moore-Penrose generalized inverse matrix. According to the method provided by the invention, the vibration response generated by external excitation in the complicated combination shell system in the fluid medium can be quantitatively calculated, therefore the reference and basis can be provided to the characteristic analysis of the vibration of an underwater vehicle.

The invention discloses a method for fast identification of a dominant oscillation mode based on a random subspace. The method comprises the following steps: obtaining measurement information of a power system from a wide area measurement system, and constructing an extended observable matrix of the random subspace of the power system according to the measurement information; according to the extended observable matrix of the stochastic subspace, calculating the state matrix of the reduced order state equation of the system; computing the inverse matrix of the generalized inverse matrix of thestate matrix; calculating the eigenvalues of the state matrix and the inverse matrix in the continuous space, and only the eigenvalues in the frequency range of the low frequency oscillation of the power system are reserved. Based on modal similarity quasi-side, the dominant oscillation mode of the system is identified. On the premise of ensuring the identification accuracy of the dominant oscillation mode, the invention improves the identification efficiency of the dominant oscillation mode of the power system based on the random subspace identification.