38results about How to "Reduce matrix dimension" patented technology

The invention discloses an FPGA (Field Programmable Gata Array)-based unscented kalman filter system mainly solving the problem that the traditional unscented kalman filter hardware has great implementation difficulty and poor instantaneity and comprising a covariance matrix Cholesky decomposition model A, a Sigma point generation module B, a time updating module C, an observation and prediction module D, a part-mean value and covariance matrix computation module E, a population mean value computation module F, a population covariance matrix computation module G, an observation and predictioncovariance matrix inversion module H, a gain computation module I and a state quantity and state covariance matrix estimating module J, wherein the module A generates K group of column vector to the module B and the B, C, D and E modules are connected in series and respectively comprise K submodules adopting a parallel arithmetic modular construction; the F and G modules receive and process the Kgroup of results of the module E and the processed results pass through the modules H, I and J in sequence to obtain the present result. The invention has the advantages of quick filter speed and easy hardware implementation and can be used for target tracking and parameter estimation.

The invention provides an equivalent model method for performing electromagnetic transient simulation on an alternating current / direct current hybrid power grid. The method comprises the steps of (1) confirming an initial network, an internal network and an external network; (2) confirming an external equivalent network; (3) performing matrix calculation; (4) screening elements in node impedance matrixes; (5) modeling and checking the external equivalent network; (6) modeling and checking the internal network except the portion outside the direct current portion; (7) modeling and checking the direct current portion of the internal network. A direct current transmission line is completely retained, accurate research on performance of a relay protection device in the alternating current / direct current hybrid power grid is facilitated, and the influence of a direct current network to the relay protection device, the influence of an alternating current network to the relay protection device and mutual effects of the direct current network and the alternating current network under the breakdown condition in the alternating current / direct current hybrid power grid can be comprehensively analyzed.

The invention discloses a personalized recommendation method based on combination of content and collaborative filtering. The method is divided into a training stage and a personalized recommendationstage, wherein in the training stage, an existing user-new article pre-recommendation matrix is formed by acquiring data information, and in the personalized recommendation stage, article recommendation for a new user and further recommendation after scoring by the user are completed. Compared with the traditional personalized recommendation method, the method has the advantages that the problem of cold start can be solved, newly added articles are predicted and then recommended to the users, collaborative filtering recommendation is used after pre-recommendation, and diversity and accuracy ofrecommended articles are improved. Further, virtual users are generated with a clustering algorithm, matrix dimensions are reduced greatly, and computation complexity is reduced.

The invention discloses a distribution network fault location method with a high fault-tolerant rate belonging to the technical field of AC distribution network fault diagnosis. The method achieves the fault location by comparing an overcurrent information matrix uploaded during a FTU fault with an overcurrent information matrix formed according to a network topology, and specifically includes: mapping a distribution network into a directed graph; according to the directed graph of the distribution network, obtaining an ideal fault matrix D; obtaining an reported information matrix G accordingto the information actually uploaded by the FTU; making subtraction on the ideal fault matrix D and the reported information matrix G to obtain a fault discriminant matrix P, thereby obtaining a fault region; and in view of missed reporting and false reporting, establishing a fault location criterion, and simulating a fault condition that the FTU generates missed reporting and false reporting according to the fault location criterion and a corresponding conditional probability. The method is simple in reasoning and good in fault tolerance. Particularly for the missed reporting and false reporting of the FTU, the provided fault location discrimination method and the expanded fault location region can improve the positioning accuracy.

The invention discloses an IRS-based large-scale MIMO cascaded channel estimation scheme under a hybrid low-precision architecture, which comprises the following steps: firstly, a base station end performing uniform division and phase configuration of units on a large-scale reflective array plane by sending a downlink configuration information signaling, and selecting a central unit of each sub-array plane to perform channel estimation; secondly, designing an isotropic analog receiver which adopts a full-connection phase shifter network, and enabling phases of all phase shifters to be uniformly distributed; then, the base station obtaining sparse information of an angle domain millimeter wave channel by utilizing a compressed sensing technology, so that a digital estimator is optimally designed, and angle domain channel estimation is obtained; and finally, through sparse recovery and angle domain inverse transformation, obtaining the cascade channel estimation of the sub-array surface,and further deducing to obtain the cascade channel estimation of all the IRS array surfaces. According to the method, the performance of channel estimation can be remarkably improved, and the amplification effect of the IRS on quantization noise is effectively inhibited.

The invention discloses a foundation radar and aerial mobile platform radar real-time error registering method. The real-time error registering method includes the steps of reporting the measurement of radars on objects, local geographical coordinates of the radars and the measurement on an attitude angle of a platform to a fusion center, obtaining an equivalent measurement equation, then carrying out sequential least square initialization and using a sequential least square method to estimate system errors of each radar and the attitude angle deviation of the platform, and finally using the estimated system errors and the estimated attitude angle deviation of the platform to compensate the system errors of the radars and the attitude angle deviation of the platform so as to achieve error registering on radar networking. Through reasonable modeling, the foundation radar and aerial mobile platform radar real-time error registering method obtains the absolute system error estimation on each radar, and the foundation radar and aerial mobile platform radar real-time error registering method is applicable to the situation that the distance between the radars is long.

A blind frequency-deviation estimating method of OFDM communication system carrier wave includes distributing virtual sub-carrier wave by emitter end in equal interval way, structuring low dimensional self-covariance matrix at receiver end through under-sampling way and utilizing subspace means to decompose obtained covariance matrix for obtaining estimation on carrier wave frequency deviation.

The invention discloses a remotely-operated wearable remote control operating platform, which relates to the field of robot systems. The remotely-operated wearable remote control operating platform comprises an information collecting device, an information display device, a communication system, a mobile operating platform and a humanoid robot, wherein the information collecting device and the information display device are configured to complete upper body information collection, biped data collection of the operator, and environment information feedback display data display; the communication device is configured to complete data communication, sensor information acquisition, and data recording; the mobile operating platform is used for receiving mobile platform motion control commands and implementing mobile platform motion. In combination of the basic principles of movement and control of the mechanical arm, a system architecture supported by hardware, a sensor, an actuator and a communication network is built, a control loop of acquisition-transmission-control-feedback and remote control of a movable platform can be realized, and man-in-loop real-time control is achieved.

The invention provides a method and a device for multi-cell pilot frequency combined channel estimation. The method for multi-cell pilot frequency combined channel estimation includes that: frequency domain pilot frequency signals are extracted by a base station from uplink signals received by terminals; a multi-cell multi-terminal local pilot frequency sequence, which is with time offset and used in coordinated multi-point transmission, is reconstructed by the base station according to the frequency domain pilot frequency signals; and multi-cell multi-terminal combined channel estimation is carried out for the frequency domain pilot frequency signals by the base station according to the local pilot frequency sequence to obtain channel estimation results of the pilot frequency signals of all the terminals. With the method and the device for multi-cell pilot frequency combined channel estimation, interference rejection capability under time offset is enhanced, requirements for scheduling are lowered, and computation complexity is decreased.

The embodiment of the invention discloses a reconstruction method for an aerodynamic model in a ground flutter test, and the reconstruction method achieves the reconstruction of the aerodynamic modelthrough modal shape excitation, physical coordinate and modal coordinate conversion and interpolation point polycondensation. In a reconstruction process, vibration mode excitation is carried out on amode participating in flutter; obtaining an aerodynamic force influence coefficient matrix under frequency domain and modal coordinates; the method comprises the following steps: performing conversion of physical quantities under modal coordinates and physical coordinates, performing optimal configuration on interpolation points through a genetic algorithm to realize polycondensation of the interpolation points, and finally fitting frequency domain aerodynamic force to a time domain through a rational function fitting mode to obtain an aerodynamic force model meeting the requirements of a ground flutter test. The reconstruction method provided by the embodiment of the invention has the advantages that the precision is high, the calculation efficiency is high, and the model reconstructionprocess is simple and feasible; the test precision of the ground flutter test in a transonic region is improved, so that the effect of expanding the application range of the ground flutter test is achieved.

The invention provides a method of enhancing the down-link performance of TD-LTE, which relates to a wireless communication system. The method comprises the following steps that the number of antennaeThe invention provides a method of enhancing the down-link performance of TD-LTE, which relates to a wireless communication system. The method comprises the following steps that the number of antennae to be selected is confirmed on the basis of a maximized capacity criterion; the antenna subset selection and the minimum singular value calculation are carried out; the precoding matrix and the signato be selected is confirmed on the basis of a maximized capacity criterion; the antenna subset selection and the minimum singular value calculation are carried out; the precoding matrix and the signal-to-noise ratio are solved on the basis of a ZF criterion; transmission antennae are selected on the basis of a minimum error rate criterion; and finally, a base station finishes the sending of downwl-to-noise ratio are solved on the basis of a ZF criterion; transmission antennae are selected on the basis of a minimum error rate criterion; and finally, a base station finishes the sending of downward information. When the relativity of an antenna unit at the sending terminal exists, the number of the antennae is confirmed on the basis of the maximized capacity criterion, the transmission antenard information. When the relativity of an antenna unit at the sending terminal exists, the number of the antennae is confirmed on the basis of the maximized capacity criterion, the transmission antennae are selected on the basis of the minimum error rate criterion, and the adverse effect to MIMO and the precoding from the antenna relativity can be eliminated or weakened, thereby the capacity of tnae are selected on the basis of the minimum error rate criterion, and the adverse effect to MIMO and the precoding from the antenna relativity can be eliminated or weakened, thereby the capacity of the down-link of the TD-LTE system is increased, and the error ratio of the TD-LTE system is reduced; in addition, the number of radio frequency parts at the sending terminal of the TD-LTE system can bhe down-link of the TD-LTE system is increased, and the error ratio of the TD-LTE system is reduced; in addition, the number of radio frequency parts at the sending terminal of the TD-LTE system can be reduced, the dimensionality of the matrix which is involved in calculation during the treatment of MIMO can be ensured to be less, and the realization complexity and the cost are also reduced.e reduced, the dimensionality of the matrix which is involved in calculation during the treatment of MIMO can be ensured to be less, and the realization complexity and the cost are also reduced.

The invention provides a power grid state estimation method and device. A power grid comprises a first node, a second node and a third node, a PWU device is installed in the first node, and the method comprises the following steps: obtaining PMU measurement data of the first node in the power grid; calculating pseudo PMU measurement data of the second node satisfying a preset condition according to the PMU measurement data; obtaining SCADA measurement data of the third node in the power grid excluding the first node and the second node; and estimating a tidal current state of the power grid at present according to the PMU measurement data, the pseudo PMU measurement data and the SCADA measurement data. The technical scheme makes full use of the PMU measurement data and the SCADA measurement data, reduces the matrix dimensions in a power grid state estimation process, the modeling complexity and the solution difficulty through power grid decomposition and coordination calculation, and improves the power grid state estimation efficiency.

The invention discloses a fractional order system identification method based on Legendre wavelet multi-resolution analysis. The method comprises the following steps that firstly, Legendre wavelet orders are selected in the interval of [0,1]; secondly, an operation matrix of Legendre wavelet integration is derived by utilizing a block pulse function integral matrix; the input signal and the outputsignal of the system are expanded in the form of Legendre wavelet; the wavelet multi-fraction analysis characteristic is utilized, output signals are continuously decomposed, high-frequency information is discarded, the data length is reduced, and finally, system parameters and orders are solved in a cyclic fractional order integral mode by adopting a least square method and taking an IAE index as a target function. According to the method, the problems that parameter estimation is slow and a noise pollution system cannot be identified due to single dimension, high order and multiple coefficients of a traditional Legendre wavelet integral operation matrix identification method are effectively solved, the system identification speed is increased, and the influence of noise on the identification precision is reduced.

The invention discloses a target detection method of a broadband radar based on knowledge assistance. The main objective of the invention is to solve a problem of low target detection performance of a broadband radar in an uneven clutter background. The method comprises steps of 1) carrying out the Fourier transform on a time domain target echo signal so as to obtain a frequency domain target echo signal; 2) carrying out N point uniform sampling on each distance unit of the frequency domain target echo signal so as to obtain a discrete object signal; 3) equating a cluster model to a generalized balance non-relevant scattering WSSUS model so as to obtain spatial-time domain and frequency domain coherent parameters; 5) based on the spatial-time domain and frequency domain coherent parameters, obtaining a frequency domain covariance matrix RF and a spatial-time domain covariance matrix RT; 6) according to the RF and the RT, constructing a generalized likelihood ratio target detector; and 7), through the detector, detecting a target signal in the clutter background. According to the invention, a broadband radar target detector is designed by use of environmental prior information, so detection performance in the clutter background is improved and the method can be used for target identification in the uneven clutter background.

The invention provides an equivalent model method for performing electromagnetic transient simulation on an alternating current / direct current hybrid power grid. The method comprises the steps of (1) confirming an initial network, an internal network and an external network; (2) confirming an external equivalent network; (3) performing matrix calculation; (4) screening elements in node impedance matrixes; (5) modeling and checking the external equivalent network; (6) modeling and checking the internal network except the portion outside the direct current portion; (7) modeling and checking the direct current portion of the internal network. A direct current transmission line is completely retained, accurate research on performance of a relay protection device in the alternating current / direct current hybrid power grid is facilitated, and the influence of a direct current network to the relay protection device, the influence of an alternating current network to the relay protection device and mutual effects of the direct current network and the alternating current network under the breakdown condition in the alternating current / direct current hybrid power grid can be comprehensively analyzed.

The invention discloses a three-dimensional alternate iteration unconditional stability FDTD algorithm in the technical field of FDTD algorithms. The method aims to solve the technical problems that an FDTD algorithm in the prior art is not suitable for calculation with fine grids due to limitation of CFL stability conditions, and calculation efficiency is affected due to the fact that a large number of memory resources need to be occupied. According to the algorithm, an ADI algorithm is combined with an idea of utilizing Laguerre-based expansion; writing the Maxwell equation set of the Laguerre domain into a matrix form; through a method of actively adding error terms, calculation of electric and magnetic field components is decoupled, unknown field quantities in the constructed linear equation set are all located on the same straight line, the dimensionality of the linear equation set is geometrically reduced in series, the calculation efficiency is greatly improved, and memory consumption is reduced. And finally, the size of an error term is enabled to tend to zero through an alternating direction iteration method to obtain an accurate solution of the electromagnetic field component.

The present invention discloses a blind frequency offset estimation method for an OFDM system, aiming at the disadvantages of the existing blind frequency offset estimation method, such as the large amount of calculation of the singular value decomposition of the covariance matrix, and the existence of uncertainty in the frequency selective channel. , the method assigns virtual subcarriers at equal intervals at the transmitter, constructs a low-dimensional autocovariance matrix through undersampling at the receiver, and finally uses the subspace method to decompose the obtained covariance matrix to obtain the carrier frequency offset estimate. The method of the invention can overcome the uncertainty under the frequency selective channel, and at the same time can provide a closed analytical solution, not only greatly reducing the amount of calculation, but also significantly improving the estimation performance.

The invention discloses an electromagnetic transient simulation method and system for a power system comprising a power electronic switch, and the method comprises the steps: carrying out the switching function modeling of the power system comprising the power electronic switch, and obtaining an electromagnetic transient equivalent circuit; carrying out sub-net division on the electromagnetic transient equivalent circuit, and constructing a network equation of node injection current vectors between sub-nets and a connection line equation between the sub-nets according to power electronic switch states in the sub-nets; and according to the network equation of the connecting line current obtained by the network equation and the connecting line equation and the node injection current vector, obtaining the sub-net node voltage of the current switch simulation time step length. The power electronic switch elements are kept in the same topological structure in different states, and the problem that when multiple power electronic switch elements are simulated, an admittance matrix changes along with changes of the topology is solved.

The invention provides a method of enhancing the down-link performance of TD-LTE, which relates to a wireless communication system. The method comprises the following steps that the number of antennae to be selected is confirmed on the basis of a maximized capacity criterion; the antenna subset selection and the minimum singular value calculation are carried out; the precoding matrix and the signal-to-noise ratio are solved on the basis of a ZF criterion; transmission antennae are selected on the basis of a minimum error rate criterion; and finally, a base station finishes the sending of downward information. When the relativity of an antenna unit at the sending terminal exists, the number of the antennae is confirmed on the basis of the maximized capacity criterion, the transmission antennae are selected on the basis of the minimum error rate criterion, and the adverse effect to MIMO and the precoding from the antenna relativity can be eliminated or weakened, thereby the capacity of the down-link of the TD-LTE system is increased, and the error ratio of the TD-LTE system is reduced; in addition, the number of radio frequency parts at the sending terminal of the TD-LTE system can be reduced, the dimensionality of the matrix which is involved in calculation during the treatment of MIMO can be ensured to be less, and the realization complexity and the cost are also reduced.

The invention discloses a nonlinear equalization method based on weighted principal component analysis, which belongs to the field of optical fiber communication. The mode maps it nonlinearly into a Voltra series matrix; the first weight vector is used to calculate the weighted covariance matrix of the Voltra series matrix, and the weighted covariance matrix is ​​decomposed into eigenvalues, and the partial eigenvectors with the largest eigenvalues ​​are The principal component projection matrix is ​​formed, and the Waltra series matrix is ​​multiplied by the principal component projection matrix to obtain the principal component matrix; the second weight vector is used to weight and sum the principal components in the principal component matrix, and the nonlinear equilibrium is obtained. The output signal of ; the first weight vector is used to represent the weight of each Voltra kernel corresponding to the same sliding window, and the second weight vector is used to represent the weight of each principal component. The present invention can reduce complexity and avoid deterioration of equalization performance.

The invention discloses a named entity recognition method and system based on a flat grid enhanced linear converter, and belongs to the field of natural language processing named entity recognition. Firstly, the text sequence sample is obtained and the label category of the entity is marked, and the text sequence is converted into a grid structure; then the recognition model is trained by minimizing the negative log likelihood loss function to the named entity recognition model; in the process of named entity recognition, For the text sequence to be recognized, it is used as the input of the trained named entity recognition model after text preprocessing, and the recognition result is obtained according to the maximum prediction score. The invention introduces lexical information based on a more efficient flat grid structure for lexical enhancement, provides prior knowledge and entity lexical boundary information for the model, and improves the recognition accuracy of the model for entity boundaries and entity types. The context information is modeled by using a linear converter, which reduces the complexity of the model, significantly improves the efficiency of the model operation, and has higher practical value.

The invention provides a direct positioning method under an unknown mutual coupling condition, and the method comprises the steps: constructing a non-uniform grid, reducing the number of matrix dimensions in combination with singular value decomposition to reduce the calculation amount, carrying out the iteration of grid refinement and unknown mutual coupling estimation through an expectation maximization algorithm, updating the coordinates of grid points, and correcting the unknown mutual coupling. And array flow pattern errors caused by unknown mutual coupling are eliminated. According to the method, the calculation complexity is effectively reduced, and the problem of false peaks caused by strong correlation between lattice points is solved; through establishment of an off-grid model and updating of grid errors, the problem that the position of a radiation source is not on a grid point is effectively solved, and the positioning precision is improved; a signal subspace is used for replacing a received signal to reduce the matrix dimension so as to reduce the calculation amount; the band-shaped symmetric Toeplitz structure and the expectation maximization algorithm of the uniform linear array mutual coupling matrix are effectively utilized, unknown mutual coupling is updated and corrected, and the positioning precision when mutual coupling exists in the receiving array is improved.

The invention provides a signal collision separation method and apparatus. The signal collision separation method provided by the invention comprises the following steps: receiving an aliasing signal formed by signal collision; performing whitening processing on the aliasing signal to obtain the number of signal collision, and dividing a plurality of time periods according to the number of signal collision; and performing matrix transformation on a corresponding signal matrix within each time period to obtain a beam forming matrix, and obtaining separated signals according to the beam forming matrix. By adoption of the signal collision separation method and apparatus, the complexity of separation algorithm of multiple signal collision is reduced.

The invention discloses a personalized recommendation method based on the combination of content and collaborative filtering, which is divided into two stages: training and personalized recommendation: in the training stage, existing users-new item pre-recommendation matrix is ​​formed by collecting data information, and personalized recommendation The item recommendation for new users and the further recommendation after the user generates a score are completed in the stage. Compared with the traditional personalized recommendation method, the present invention can solve the cold start problem, predict newly added items, recommend them to users, use collaborative filtering recommendation after pre-recommendation, and improve the diversity and accuracy of recommended items. Moreover, generating virtual users through a clustering algorithm greatly reduces the dimension of the matrix, and at the same time reduces the complexity of calculation.

The invention discloses an FPGA (Field Programmable Gata Array)-based unscented kalman filter system mainly solving the problem that the traditional unscented kalman filter hardware has great implementation difficulty and poor instantaneity and comprising a covariance matrix Cholesky decomposition model A, a Sigma point generation module B, a time updating module C, an observation and prediction module D, a part-mean value and covariance matrix computation module E, a population mean value computation module F, a population covariance matrix computation module G, an observation and predictioncovariance matrix inversion module H, a gain computation module I and a state quantity and state covariance matrix estimating module J, wherein the module A generates K group of column vector to the module B and the B, C, D and E modules are connected in series and respectively comprise K submodules adopting a parallel arithmetic modular construction; the F and G modules receive and process the Kgroup of results of the module E and the processed results pass through the modules H, I and J in sequence to obtain the present result. The invention has the advantages of quick filter speed and easy hardware implementation and can be used for target tracking and parameter estimation.

The invention discloses a power consumption analysis method based on mean ridge regression. The method comprises the following steps: collecting power consumption data in the operation process of password equipment; calculating an intermediate value according to an algorithm structure; establishing a dynamic model for each hypothesis key based on association of a ridge regression thought and actually measured power consumption data, and performing equalization processing on a modeling result of each hypothesis key to establish a power consumption model; and calculating the correlation between the output value of the power consumption model and the actually measured power consumption data by adopting a Pearson's correlation coefficient, thereby realizing the analysis of internal secret information of the cryptographic algorithm and recovering the secret key information of the cryptographic equipment. According to the analysis method, the matrix dimension output by the power consumption model is effectively reduced through mean value modeling, and the calculation complexity of subsequent correlation analysis is effectively reduced. The analysis method has universality and is suitable for most block cipher algorithms.

The invention relates to a pseudo signal source identification method based on space projection, and belongs to the technical field of array signal processing. According to the method, under the condition that related information sources exist and the number of the information sources is unknown, when pseudo information source spectrum peaks exist in spatial spectrums of MUSIC, subspace fitting and other algorithms, the pseudo information sources can be identified, and therefore the accurate information source spatial distribution state is obtained. According to the method, an identification space and a confidence value calculation model are constructed by utilizing the uniqueness of a space complete base to vector description in a space, and a pseudo source spectrum peak is effectively identified by calculating a confidence value of each steering vector in the identification space, so that the method can be used as a post-processing means of spatial spectrum algorithms such as MUSIC and subspace fitting; high-precision direction finding under the condition that the number of information sources is unknown is achieved, and the engineering applicability of the algorithm is further improved.