36 results about "Mean square error minimization" patented technology

The waveform equalizing device includes: an FIR filter for generating an equalized signal pattern y(i, n) on the basis of equalization of a waveform of the reproduced signal pattern u(i, n); a Viterbi decoding circuit for detecting a path metric difference s(n) between a correct path determined as a survivor path and an error path which fails to survive the correct path in Viterbi decoding based on the equalized signal pattern y(i, n); a target register for setting a target value ds for the path metric difference s(n); and a tap coefficients update circuit for adapting the equalization according to an error of the detected path metric difference from the target value. The tap coefficients update circuit adapts equalization properties so that the mean square error is minimized, thereby achieving a satisfactory result in lowering the error rate.

The invention provides a flow rate level prediction method based on convolutional neural network deep learning. The method comprises the following steps that: selecting an impact factor which is potentially related to the inflow flow rate of a reservoir as an input set; carrying out sample set classification and original input dataset construction; carrying out standardized processing on the original input dataset in the sample set; building a multilayer convolutional neural network; taking mean square error minimization as a loss function to determine prediction accuracy; carrying out networkparameter training; carrying out network performance testing; checking prediction accuracy; carrying out the rolling learning training on model parameters; automatically saving a learning training achievement, and automatically updating the knowledge record of a real-time library; and through a network model, carrying out calculation to obtain a final flow rate level prediction result. By use ofthe method, low-layer characteristics are combined to form high-layer characteristic fusion, so that an objective can be subjected to advanced abstract description, a data input pattern and a spatialand temporal distribution rule can be found through automatic learning, and therefore, the method can be effectively applied to the field of drainage basin water regimen forecasting.

The invention provides a Massive Multiple-Input Multiple-Output (Massive MIMO) pre-coding method, which aims to solve the problems of high error rate and low energy efficiency of a Massive MIMO system in the prior art. The method comprises the following steps: obtaining a mathematical optimization model of the Massive MIMO system; designing a receiving end base band decoding matrix; designing a sending end base band pre-coding matrix; designing a sending end radio frequency pre-coding matrix; and outputting the base band and radio frequency hybrid pre-coding matrixes. The Massive MIMO pre-coding method provided by the invention has the benefits that the base band and radio frequency hybrid pre-coding method adopts the principle of mean square error minimization, and combines the convex optimization method with the Particle Swarm Ant Colony Optimization (PSACO) method, so as to achieve a high iteration rate and obtain the hybrid pre-coding matrixes; compared with the traditional pre-coding method, the method can better reduce the error rate of the system; meanwhile, compared with the full-digital base band pre-coding, the method has the advantages of low hardware implementation cost and high energy efficiency.

The invention provides an underwater terrain-aided navigation method based on an adaptive sampling particle filter. The method comprises the steps that a state space model based on an inertial navigation position error and a measurement model based multi-beam depth-sounding sonar are built; one-step predicted particle updating is conducted through initial state distribution and the state space model, the number of particles is adjusted according to predictive state distribution by adopting a KLD sampling technology, and a predicted particle set is obtained; when a multi-beam measured value isreached, the depth, the inertial navigation guiding position and an underwater reference digital map interpolation function are measured in combination with a pressure-depth meter, and particle measurement updating is conducted through the measurement model; finally, by means of the particle set and weight which are obtained after measurement updating is conducted, aircraft position error estimation is conducted by adopting mean squared error minimization rules, and the estimated error is used for correcting the inertial navigation guiding position. Accordingly, the navigation real-time property can be improved while the underwater terrain-aided navigation precision is guaranteed.

The invention discloses a low delay filter design method for sampling rate conversion in an electronic transformer. The method comprises the steps of first adopting cascade connection of an interpolator and an extractor to achieve conversion of random fraction ratio sampling frequency, enabling an anti-image filter in the interpolator and an anti-aliasing filter in the extractor to be combined into a low pass filter; then adopting a mean square error minimization standard to solve a coefficient vector of the filter, enabling a filter transmission band amplitude and a stop band amplitude to serve as constraint conditions, and enabling mean square error minimization to serve as an optimized target; and finally enabling a solving process of the filter coefficient vector based on constraint least square method design to be converted into solving of a positive definite quadratic programming problem so that the filter coefficient vector can be obtained by direct solving. According to the method, the problems that large output delay is caused in the direct linear convolution filtering process and the higher the filter order, the larger the group delay are solved, and rapidity of protection action is greatly improved.

Electrical connectors capable of being mounted on circuit substrates by BGA techniques are disclosed. Also, disclosed is a method of manufacturing such connectors. There is at least one recess on the exterior side of the connector elements. A conductive contact extends from adjacent the interior side into the recess on the exterior side of the housing. A controlled volume of solder paste is introduced into the recess. A fusible conductive element, in the form of solder balls is positioned in the recess. The connector is subjected to a reflow process to fuse the solder ball to the portions of the contact extending into said recess. Contacts are secured in the insulative housing of the connector by deformable sections that minimize stress imposed on the central portions of the contacts to promote uniformity of solder volume.

The invention discloses a FIR filter design method based on particle swarm optimization. The method is characterized by using the unit impulse response of a FIR filter as the particle position vector of the particle swarm optimization; designing the fitness function of the particle swarm optimization on the basis of a signal-to-noise ratio of a signal filtered by the FIR filter and the mean square error of the FIR filter and obtaining the FIR filter by iteration. The method designs the fitness function of the particle swarm optimization based on minimized signal-to-noise ratio and mean square error, and designs the FIR filter based on particle swarm optimization so as to achieve the FIR filter taking account of signal-to-noise ratio performance.

The invention provides an image de-noising method which comprises the steps of initializing a preset dictionary for obtaining an initial dictionary, and initializing a to-be-processed image to an image matrix which corresponds with the initial dictionary; wherein useful information in the to-be-processed image has a specific structure and sparse representation of the useful information in the to-be-processed image can be performed in the initial dictionary, and a noise in the to-be-processed image has no structure characteristic and sparse representation of the noise cannot be performed in the initial dictionary, updating the initial dictionary and a sparse coefficient matrix so that a reconstruction mean square error of the image matrix is minimized, reconstructing a de-noised image block by means of an updated dictionary and an updated sparse coefficient matrix, and generating a new to-be-processed image through combining the de-noised image block and the to-be-processed image for updating the dictionary and solving the sparse coefficient matrix, so that the updated dictionary ion better matches the structure characteristic information of the original to-be-processed image, thereby facilitating filtering of the noise component in the original to-be-processed image and improving image de-noising precision.

The present invention discloses an image registration and splicing method on the basis of a Lucas-Kanade algorithm. The method provided by the invention comprises: pretreatment of an original image, wherein unmatched parameters between two images are generated because of the differences such as shooting angles, shooting distances, shooting illumination and the like, and the pretreatment of images is carried out prior to the image registration; image registration, wherein the image registration is taken as a key step in the image splicing technology, the splicing quality of the images is largely determined by the precision of the image registration; and image fusion, wherein the objective of the image fusion is to convert registered images to be spliced to the same coordinate system for fusion of pixel levels through a coordinate conversion relation so as to more comprehensively display scene information. According to the invention, the problem is solved that the accuracy of a homography matrix estimation is low caused by few feature points and noise jamming, and an LK image aligning algorithm and the changed algorithm thereof are creatively provided for optimizing the homography matrix estimation. Through iteration optimization of a whole image, the mean square error of the image is minimized, and the accuracy of the homography matrix estimation is enhanced, so that the fusion effect of registration and splicing fusion of the image is improved.

An equalization method and apparatus for a downlink channel in a telecommunication system transmitting MC-CDMA symbols on a plurality of carrier frequencies, in which a plurality of links are established between a transmitter and a plurality of receivers in order to transmit a plurality of coded signals, each link using a distinct access code amongst N possible codes, there being P links for P users where P is less than N. The equalization including the formation and use of a matrix of reduced size based upon the number of links being P by eliminating matrix elements relating to the unused codes and effecting an estimation of the said transmitted symbols according to a mean square error minimization criterion.

The invention, which relates to the field of power grid planning and scheduling operation, provides a daily on-off type peak load operation method of a gas engine set. The power grid and the difference between the peak and valley increase rapidly, so that the peak regulation pressure of the system increases continuously; and the peak regulation pressure of the power grid increases further because of the influence of large-scale feeding of the extra-high-voltage direct-current power. The scheduled electric quantity is used as the single-machine control object; the start-up time range is determined by considering the upper limit and the lower limit of the output; and a feasible continuous start-up time set is searched and determined by using a load value as heuristic information. With consideration of the adjustable output lower limit, the high-peak output process of each start-up scheme is determined in an optimized manner by using an improved load shedding strategy; and an optimal start-up operation scheme and a set generating plane are selected based on a criterion of residual load mean square error minimization. According to the method provided by the invention, the peak regulation space of the gas engine set at the peak load time is released fully, so that the tracking and response capability of the grid peak load is improved effectively; and the grid peak regulation demand and the gas engine set operation restriction are considered simultaneously.

The invention discloses a precoding and equalization joint optimization method based on rate matching, which comprises the following steps: establishing a mean square error minimization model, iteratively solving the optimization model through a step-by-step iterative method, splitting the optimization model in the iterative process, firstly solving the optimal data stream number of each user, andthen solving the optimal data stream number of each user; fixing an equalization matrix solving and coding matrix, then taking the precoding matrix as a known quantity to solve the equalization matrix, then iteratively solving the precoding matrix and the equalization matrix, and terminating iteration when a convergence condition is met. According to the method, the number of the user data streams and the precoding and equalization matrix are adaptively selected according to the channel state information (CSI) of different users in the multi-user multi-input multi-output MIMO VLC system, so that the system bit error rate in the same type of scheme is effectively reduced. The system performance is improved.

The invention discloses an equipment state big data computing method based on a Goldstein-BP algorithm. The method comprises the steps that S101, historical data and current data of grid equipment state evaluation indexes and grid equipment historical state evaluation results are acquired; S102, a corresponding BP neural network model is constructed according to equipment state index historical data characteristics; S103, a mean square error function is constructed according to the historical data of the grid equipment state evaluation indexes and the grid equipment historical state evaluationresults; S104, a Goldstein method is adopted to obtain a minimum mean square error, a learning rate in the BP algorithm is solved, and BP neural network parameters are computed through the BP algorithm; S105, the BP neural network parameters are substituted to obtain a BP neural network model; and S106, current equipment state evaluation index data is used as input of the BP neural network modelto obtain an output result. Through the technical scheme, the Goldstein method is adopted to solve the optimal learning rate of the mean square error function in the BP algorithm, and therefore the precision level of grid equipment management is raised.

The invention aims to provide a design method suitable for a filter coefficient in a filter-and-forward relay network, relates to the technical field of wireless communication relay transmission, and aims to obtain optimal design of the filter coefficient in the filter-and-forward relay network by following a channel capacity maximization principle. The solution of a channel capacity maximization problem is transformed into the solution of a weighted mean square error minimization problem of a receiver in a virtual channel by constructing the virtual channel between a transmitter and the receiver, so that optimal design of the filter coefficient in the whole filter-and-forward relay network is obtained. A good theoretical basis is laid for the design of the filter coefficient, the processing of data and the allocation of resources in the filter-and-forward relay network.

The invention discloses a large-scale MIMO low-complexity channel estimation method based on weighted Kapetyn grade number expansion. Firstly a Kapetyn grade number expansion method is utilized for performing approximate expansion on a channel covariance inverse matrix in a Bayesian-MMSE channel estimation expression. A matrix inversion operation is converted to matrix multiplication and matrix addition operations. Then a weighting manner is performed on each coefficient of a polynomial for optimizing polynomial expansion, establishing a model for solving weighting coefficient vectors alpha and beta for minimizing an estimated mean square error, and estimating the channel matrix by means of solving results of alpha and beta. Experiment results represent a fact that an MSE which is obtained through the channel estimation method based on a weighted Kapetyn grade number expansion is convergent to an MMSE method along with order number increase of the polynomial, and furthermore calculation complexity of the channel estimation method is lower than that of the MMSE method. Compared with a traditional Taylor-MMSE and Kapetyn grade number expansion channel estimation method, the channel estimation method based on the weighted Kapetyn grade number has higher convergence speed to the MMSE method.

The invention discloses a digital all-pass filter design method based on a hybrid particle swarm algorithm, which comprises the following steps: constructing a phase-frequency response of the digitalall-pass filter based on second-order cascade connection according to design requirements; then based on target phase-frequency response, building a cost function via an error value between the phase-frequency response of the digital all-pass filter and the target phase-frequency response; and finally, searching a global optimal solution in a value domain range by using a global search particle swarm algorithm to obtain a rough solution, and triggering accurate search performed by using a Levenberg Marquardt algorithm from the rough solution, thereby optimizing the positions of poles of all second-order all-pass filters, so that the mean square error minimization of the designed all-pass filter phase frequency response and the target phase frequency response is realized, and the design ofthe filter coefficient is realized.

A complex adaptive phase detector (PD), as proposed in certain aspects of the present disclosure, is an adaptive filter that accurately estimates the instantaneous frequency of a dynamic complex signal. In dynamic complex signal environments, PD can demonstrate accurate instantaneous frequency estimation and fast convergence even when the input frequency changes rapidly. The direct PD structure can iteratively adjust the phase of the complex exponential by minimizing the mean square error of the complex error signal, thereby estimating the instantaneous frequency of the complex primary signal. The instantaneous frequency can be estimated directly from the phase of the complex exponential. In contrast, the indirect PD structure can iteratively adjust the complex coefficients by minimizing the mean square error of the complex error signal, thereby estimating the instantaneous frequency of the complex primary signal. The instantaneous frequency can be estimated indirectly by extracting the phase of the complex coefficients.

The invention discloses a forced integer precoding method of a visible light communication system. Firstly, a mean square error minimization model is established, then the mean square error minimization model is converted into iterative solution of a non-convex problem, in one iteration, the mean square error minimization model is split to obtain a model for optimizing a transmitting precoding matrix and a model for optimizing a forced integer matrix respectively, the transmitting precoding matrix is solved by using a convex approximation iteration method, a compulsory integer matrix is solvedby using a greedy method, and iteration is stopped when a convergence condition is met, so that the solved precoding matrix and the compulsory integer matrix are final results, otherwise, the next iteration is continued. According to the invention, a smaller mean square error is obtained under the same optical power constraint condition.

The invention discloses a method for estimating a 2x-mode channel coefficient of an 802.11 ax protocol, and the method comprises the steps: correcting an unknown channel coefficient in a 2x mode through employing a feedback method on the basis of a linear interpolation method, and achieving the minimization of the mean square error of an estimation coefficient value. Compared with a one-time linear interpolation / polynomial interpolation algorithm, the method has the advantages that the channel estimation precision in the 11ax system is improved through multiple iterations, and the signal receiving requirement is better met.

The invention discloses a large-scale MIMO low-complexity channel estimation method based on weighted Kapetyn grade number expansion. Firstly a Kapetyn grade number expansion method is utilized for performing approximate expansion on a channel covariance inverse matrix in a Bayesian-MMSE channel estimation expression. A matrix inversion operation is converted to matrix multiplication and matrix addition operations. Then a weighting manner is performed on each coefficient of a polynomial for optimizing polynomial expansion, establishing a model for solving weighting coefficient vectors alpha and beta for minimizing an estimated mean square error, and estimating the channel matrix by means of solving results of alpha and beta. Experiment results represent a fact that an MSE which is obtained through the channel estimation method based on a weighted Kapetyn grade number expansion is convergent to an MMSE method along with order number increase of the polynomial, and furthermore calculation complexity of the channel estimation method is lower than that of the MMSE method. Compared with a traditional Taylor-MMSE and Kapetyn grade number expansion channel estimation method, the channel estimation method based on the weighted Kapetyn grade number has higher convergence speed to the MMSE method.