169 results about "Vector group" patented technology

The invention provides a dialogue data interaction processing method based on a recurrent neural network. The method comprises the following steps: receiving a dialogue input statement of a user; carrying out knowledge base matching calculation, and judging whether a knowledge base has a problem statement, the matching degree between which and the dialogue input statement reaches a preset value; if not, requesting a dialogue generation model to give an answer to the dialogue input statement, wherein a coding layer of the dialogue generation model is constructed into the recurrent neural network, analyzing the dialogue input statement in the recurrent neural network to obtain a middle vector expressing problem meanings, and a decoding layer of the dialogue generation model is also constructed into the recurrent neural network, analyzing the middle vector in the recurrent neural network to obtain an answer vector group expressing answer meanings; and taking the answer vector group as an answer output statement and outputting the answer output statement. The method can enable human-machine interaction to be smoother; and answers given by the dialogue generation model can further enable the knowledge base to be expanded and updated.

A method for classifying documents to a browser is disclosed, the method comprising establishing communication with a selected network address; displaying at least one document with the browser; moving the at least one document inside the browser; and generating a document vector grouping a plurality of documents.

The invention discloses a fuzzy multi-keyword retrieval method of encrypted data in cloud environment. A file is subjected to set encryption by a data owner to generate a ciphertext file; keywords are extracted from each file; the keywords are subjected to binary segmentation and vectorization to obtain a binary vector group; the binary vector group is subjected to dimensionality reduction and is then inserted into a counting type bloom filter to generate index vectors; each index vector is encrypted to obtain a security index; the ciphertext file and the security indexes are sent to a cloud server; a pre-authorized data user or the data owner extracts the keywords from query data; binary segmentation and vectorization are performed to generate a query vector; encryption is performed to obtain a trap door; the trap door is sent to the cloud server; the cloud server obtains a certain number of files with the highest relevancy degree through query according to the trap door and the security index; after sorting, the files are returned to the data owner. The large data volume of ciphertext multi-keyword retrieval is supported; compared with the prior art, the method has the advantages that the index building and query processes are more efficient; the sorting result is more accurate; the data privacy is effectively protected.

Provided is an apparatus and method for providing to a differential relay an operational vector-group compensation setting pair that automatically provides correction for a phase shift occurring between currents of at least two windings of a power transformer. The method includes calculating a first and second plurality of phasors using secondary currents derived from respective first and second winding of the at least two windings. The method also includes selecting one pair of vector-group compensation settings based on operate current values calculated using different pair combinations of the vector-group compensation settings applied to the first and second plurality of phasors. The method further includes comparing phase angles of like-phase phasors to determine if the phase angles are within a pre-determined angle range, and enabling application of the selected pair of vector-group compensation settings as the operational vector-group compensation setting pair if the phase angles are within the pre-determined angle range.

An image processing device calculates, from a registration image representing a photographed object and three-dimensional shape data in which respective points of a three-dimensional shape of the object are correlated with pixels of the registration image, by assuming uniform albedo, a shadow base vector group having components from which an image under an arbitrary illumination condition can be generated through linear combination. A shadow in the registration image is estimated with using the vector group. A perfect diffuse component image including the shadow is generated, and based on the image a highlight removal image is generated in which a specular reflection component is removed from the registration image. Thus, an image recognition system generates illumination base vectors from the highlight removal image and thereby can obtain the illumination base vectors based on which an accurate image recognition process can be carried out without influence of a specular reflection.

The invention discloses a transformer partial-discharging mode recognition method based on a singular value decomposition algorithm, and the transformer partial-discharging mode recognition method comprises training model and classification recognition process, and the method comprises the steps of firstly establishing an artificial defect experimental environment, collecting data samples, calculating statistic characteristic parameter of each sample to form a data sample matrix; conducting singular value decomposition for the sample matrix, determining an order of an optimum reserved matrix by judging whether the characteristic of the reserved matrix is obvious or not, and obtaining a type characteristic space description matrix after the dimensionality reduction and a class center description vector group; preprocessing the sample to be recognized to obtain a sample vector, utilizing the type characteristic space description matrix to linearly convert the sample vector to obtain the sample description space vector after the dimensionality reduction, and then calculating the similarity of the vector with each vector in the type vector group to obtain a classification judgment result. The algorithm is simple and high efficient, reliability for distinguishing an interference signal and a discharging signal in the partial-discharging detection can be realized, and the accuracy for diagnosing the partial-discharging mode can be improved.

The invention discloses a position and attitude measurement method based on time of flight (TOF) scanning-free three-dimensional imaging, which is characterized by comprising the following steps of: establishing a three-dimensional coordinate information database for a target object, taking a TOF camera as an imaging and distance data acquisition device, selecting three identifiable objects in a photographed picture as mark points, acquiring coordinate information of the mark points in the database under a target body coordinate system, and establishing a vector group; obtaining a distance from an optical center to the mark points by a method for calculating a distance between two points in a non-iterative three-dimensional space through the data information acquired by the TOF camera, and calculating coordinates of the mark points under a camera coordinate system; and establishing another vector group, and calculating a rotation matrix and a translation matrix through a relationship between the two vector groups so as to acquire an attitude angle and translation quantity, namely a relative attitude of the target object. By the method, a large number of iterative algorithms are avoided, position and attitude can be quickly solved, and the requirement of position and attitude parameter determination accuracy can be met.

Provided is an apparatus and method for providing to a differential relay an operational vector-group compensation setting pair that automatically provides correction for a phase shift occurring between currents of at least two windings of a power transformer. The method includes calculating a first and second plurality of phasors using secondary currents derived from respective first and second winding of the at least two windings. The method also includes selecting one pair of vector-group compensation settings based on operate current values calculated using different pair combinations of the vector-group compensation settings applied to the first and second plurality of phasors. The method further includes comparing phase angles of like-phase phasors to determine if the phase angles are within a pre-determined angle range, and enabling application of the selected pair of vector-group compensation settings as the operational vector-group compensation setting pair if the phase angles are within the pre-determined angle range.

The invention discloses a velocity ambiguity resolving method based on a coherent frequency agile radar. Through the method, the problem that target detection is not accurate due to distance-velocityambiguity in pulse radar system signal processing is solved. According to the implementation method, a radar pulse echo signal is received; a baseband pulse echo signal sampling matrix is obtained; pulse compression is performed; a radar Doppler vector group is constructed; Doppler offset frequency is determined, and the velocity of a motion target is calculated; velocity ambiguity resolving is performed on the radar pulse echo signal; and target detection is performed through sparse recovery. According to the method, the accurate velocity of the motion target is calculated through the Dopplervector group, velocity phase compensation is performed on the radar pulse echo signal to realize velocity ambiguity resolving, and the target is detected through sparse recovery. The method is low ininterception and resistant to interference, the velocity of the motion target can be accurately calculated, complexity is low, and the problem of distance-velocity ambiguity in pulse radar system signal processing can be solved. The method is applied to the field of radar target detection.

A method for recognizing a transformer partial discharge pattern based on a singular value decomposition (SVD) algorithm includes a training model and a classification recognizing process, comprising: firstly setting up an experimental environment having artificial defects, collecting at least one datum sample, and calculating statistical feature parameters of each datum sample to form a datum sample matrix; performing singular value decomposition on the datum sample matrix and determining an order of an optimal retention matrix by judging whether a feature of a retention matrix is clear, so as to obtain a type feature description matrix and a class-center description vector group after dimensionality reduction; preprocessing samples to be recognized to obtain a sample vector, and performing linear transformation on the sample vector utilizing a type space description matrix.

The invention relates to a human body behavior recognition method based on global characteristics and sparse representation classification. The method comprises the following steps: performing Gaussian kernel convolutional filtering preprocessing on a video frame, and extracting a moving foreground pixel by using a differential method; sampling a pixel value according to a time space dimension ofa parameter, determining a moving area, adjusting the size of the video frame, performing primary dimension reduction, splicing video frames in rows to form a vector group, and acquiring characteristic vectors; splicing the characteristic vectors in rows to form a characteristic matrix, performing secondary dimension reduction, calculating a primary characteristic dictionary of the characteristicmatrix, initializing the dictionary, after dictionary initialization, performing dictionary learning by using a class accordant K-time matrix singular value decomposition method, calculating an inputsignal sparse code according to the dictionary, inputting the code into a classifier, and outputting a behavior type; and counting dictionary learning parameters, and performing behavior recognition in real time. By adopting the method, dictionaries and linear classifiers with both reconstitution functions and classification functions are acquired, human body behavior recognition efficiency is improved, and the method is applicable to scientific fields such as security monitoring, video search based on contents and virtual reality.

The invention discloses a face identification algorithm under different illumination conditions, comprises the steps that: (1) the inputted face original image F(x, y) is subjected to Gaussian filter so as to eliminate noise point in the image and smooth the image, thereby obtaining a new image F'(x, y); (2) the F' (x, y) is subjected to feature extraction by using a Weber local operator description; (3) an edge is extracted from a new image WF(x, y) by LoG operator operation; (4) the inputted image is converted to a principal component space by using fast PCA algorithm; (5) in the PCA transform space, a inter-class scatter matrix is calculated (img file='DDA0000447013750000011.TIF' wi ='45' he='65' / ); (6) the obtained new image feature WF'(x, y) is projected on a feature projection vector group Phi; (7) finally, a nearest neighbor classifier carries out classification. The invention carries out feature extraction of the image operated by the LoG operator, more complete identification information of the original imagely can be extracted, the data from the above experiment shows that the new algorithm can obtain the prominent face identification rate from all three face databases, and also completely meet the real-time requirements in the speed.

A value bill true or false identification method comprises: 1, using the light of at least two different wavelengths to respectively scan the surface of an object value bill, using a photoelectric sensor to receive and convert reflected light into analogue electric signals reflecting the distribution characteristic of special material in bill paper and / or surface ink, processing the analogue electric signals to obtain initial data corresponding to different wavelengths; 2, preprocessing, comparing and analyzing the initial data of at least two kinds corresponding to different wavelengths to form at least one characteristic vector group; 3, comparing at least one characteristic vector group with the corresponding characteristic vectors of a true bill, to identify true or false. A system thereof comprises a light source module, a photoelectric sensor module, a signal preprocessing module, an automatic calibration control module and a bill recognition platform. The invention adopts photoelectric colorimetric method to qualitatively and quantitatively analyze bills, to accurately identify true and false bills, thereby improving identification precision. The invention has wide application for identifying true and false bills, having strong generality.

The invention relates to a method for distinguishing speakers based on a protective kernel Fisher distinguishing method. The method comprises steps as follows: (1) pretreating voice signals; (2) extracting characteristic parameters: after framing and end point detection of voice signals, extracting Mel frequency cepstrum coefficients as characteristic vectors of speakers; (3) creating a speaker distinguishing model; (4) calculating model optimal projection vector: by using optimal solution of LWFD method, calculating to obtain an optimal projection vector group; (5) distinguishing speakers: projecting original data xi to yi belonging to R<r>( r is more than or equal to 1 and less than or equal to d) according to optimal projection classification vector phi, wherein r is cut dimensionality;the optimal projection classification dimensionality of original c type data space is c-1, then solving a central value of data of each type after injection and normalizing; after projecting data tobe classified to a sub space and normalizing, calculating Euclidean distance from the normalized protecting data to the central point of each type of data in the sub space, and judging the nearest tobe a distinguishing result. The invention has high distinguishing rate, simple model construction and favorable rapidity.

The invention discloses a mechanical arm grabbing control method based on machine vision and depth learning. The method comprises the steps that an operation scene image under the current state of a mechanical arm is obtained, and a motion instruction vector group is generated according to a sampling mean value and an initial variance of motion instruction vectors; the motion instruction vectors are combined with operation scene pictures to obtain possibility prediction values corresponding to the motion instruction vectors; the multiple possibility prediction values corresponding to the motion instruction vectors are sequenced according to the value, and at least one optimal motion instruction vector corresponding to the maximum possibility prediction value is obtained; and the possibility prediction value of the grabbed object under the current state of the mechanical arm is compared with the possibility prediction value of the optimal motion instruction vector to determine a grabbing motion decision. The invention further discloses a mechanical arm grabbing control system based on machine vision and depth learning. The technical scheme can be applied to many mechanical arm application fields such as industrial mechanical arm sorting, feeding, service mechanical arm grabbing and the like, and an intelligent and stable grabbing effect is provided.

The present invention provides a ground PM2.5 concentration feature vector spatial filter value modeling method based on remote sensing data. The method comprises: obtaining data, selecting a model variable, carrying out data processing and matching, constructing a spatial adjacency matrix from the location of the national control point of a study area, carrying out centralization, calculating thematrix eigenvalues and eigenvectors, and extracting the appropriate eigenvectors from the vector group as the spatial influence factor of the PM2.5 concentration; and obtaining an eigenvector spatialfilter regression model of the PM2.5 concentration, interpolating the extracted eigenvectors raster images with the same spatial resolution as the AOD, and bringing the raster images into the eigenvector spatial filter regression model for raster calculation to obtain the continuous spatial distribution model of the PM2.5 concentration in the study area. According to the method provided by the present invention, for the problem that the number of ground control points is small and the ground control points are unevenly distributed, the remote sensing data with high resolution and continuous distribution is selected to perform the inversion of the ground PM2.5 concentration for study on a wide range of PM2.5 spatiotemporal features.

The present invention relates generally to data communications, and more particularly to techniques based on the G.fast protocol for managing operation around potentially degrading un-cancellable crosstalk among separate vector groups implemented in a single G.fast based box located at a network distribution point, referred to as a Distribution Point Unit (DPU). In embodiments, techniques according to the invention configure transmission of signals from the different vector groups so as to avoid or prevent transmission of signals, either in the frequency domain or time domain or a combination of the two, from causing severe degradation in performance due to un-cancelled crosstalk among the separate groups.

The invention discloses a literature entity relationship discovery method and system based on a knowledge graph. The method comprises the following steps of constructing the knowledge graph accordingto the entities in the literature contents and the relationships among the entities; extracting the RDF data in the knowledge graph, and performing vectorization processing on the RDF data to obtain the vector data; obtaining a native entity relationship vector group and a non-connected entity relationship vector group according to the association relationship of the vector data; carrying out thevector matching degree calculation on the unconnected entity relationship vector groups, and screening out the unconnected entity relationship vector groups with the vector matching degrees greater than a preset threshold value, or sorting the unconnected entity relationship vector groups according to the vector matching degrees obtained through calculation. According to the method, the knowledgegraph is constructed by taking the entity in the literature content as the granularity; and the entity relationship matching is carried out through the vector calculation based on the knowledge graph,and the potential entity relationships among literature contents can be deeply mined, so that an innovative research method is provided to find the potential knowledge which is not found by human beings, and the development process of human science is accelerated.

The present invention is intended to provide quick search of a plurality if any of sets subject to search for desired members. A common representative point setting block sets a common representative point common to sets on the basis of a feature vector stored in a vector information storage block and stores the set point into a common representative point information storage block. A neighborhood table group creation block creates a neighborhood table for each set on the basis of the vector information and common representative point information and stores the created table in a neighborhood table group storage block. A nearest neighborhood vector group search block, upon acquisition of an input vector, searches each set for a nearest neighborhood vector of the input vector on the basis of the vector information, the common representative point information, and the neighborhood table group and outputs the retrieved nearest neighborhood vector as a comparison result. The present invention is applicable to image processing apparatuses.

The invention discloses a method and a system for intelligent retrieval of series public security cases. The method comprises the steps of obtaining case retrieval information and extracting characteristics from the case retrieval information, respectively, to form a multi-dimensional case characteristic vector group, calculating the coefficient of association of the multi-dimensional case characteristic vector group and the multi-dimensional historical case characteristic vector group of each historical public security case in a public security case database, and outputting the information of the public security cases of which the coefficients of association exceed a preset coefficient threshold of association. According to the method for intelligent retrieval of series public security cases, the serial-parallel analysis of the series cases can be executed quickly and the efficiency is improved, and therefore, the public security organization or the investigation personnel can be helped to quickly detect cases.

The invention discloses a self-adaption anti-coherent interference technology based on characteristic component rejection. Firstly, characteristic decomposition is carried out on an antenna multichannel receiving data covariance matrix to judge the number of a big characteristic value; then, the characteristic vector corresponding to signals and coherent interference is searched in a characteristic vector group corresponding to the big characteristic value so as to subtract the characteristic vector from the covariance matrix; finally, diagonal loading is carried out on the covariance matrix removing signals and coherent interference components, and the loaded matrix is used for calculating a self-adaption weight, data is weighed, and specific steps are disclosed in the drawing. The method of the invention avoids cancellation of desired signals while inhibiting incoherent interference, does not have array aperture loss and does not need to master the direction priori information of coherent interference. The method only relates to characteristic decomposition and inversion operation but does not relate to high-order cumulant operation, so that the method has simple steps and small calculation amount; the device is simple and has low cost. In addition, the method of the invention receives and utilizes the coherent interference as a wanted signal so as to improve the receiving gain of a target signal, thus owning better receiving performance. The method of the invention can be realized only by downloading a program to a general signal processing board, is easy to popularize and only needs to programme on a programmable signal processing board; thus, the system is convenient to upgrade while the system structure is not changed. The method of the invention can be widely applied to a system with various kinds of receiving channel structures and has popularization and application value.

The invention relates to a new generation method of initial codebook of quantization of vector, which pertains to the technology field of the process of digital images and digital audio. The beginning is the initialization of parameter, and the following process is a circulation process: first, calculating distance threshold; writing a first cluster exercising vector into a cluster vector group, and then calculating a second exercising vector and its distance d1; If the distance d1 is bigger than d0, putting the second exercising vector into the cluster vector group; if not, including the second exercising vector in the cell of the first cluster vector and calculating the vector number of the cell; second, calculating sizplus; third, judging, if at this time the number of vectors of the codebook is siz book, requirement is meet and the calculation finishes; if the codebook increment brought by circulation plus the current codebook size can not meet the codebook size set by design mark, deleting all the vectors in each marked cell collectively from the exercising vectors and switching to the first step.

An image processing device calculates, from a registration image representing a photographed object and three-dimensional shape data in which respective points of a three-dimensional shape of the object are correlated with pixels of the registration image, by assuming uniform albedo, a shadow base vector group having components from which an image under an arbitrary illumination condition can be generated through linear combination. A shadow in the registration image is estimated with using the vector group. A perfect diffuse component image including the shadow is generated, and based on the image a highlight removal image is generated in which a specular reflection component is removed from the registration image. Thus, an image recognition system generates illumination base vectors from the highlight removal image and thereby can obtain the illumination base vectors based on which an accurate image recognition process can be carried out without influence of a specular reflection.

The invention discloses a machine self-learning construction knowledge atlas training method based on a neural network. The method comprises the following steps of acquiring a statement based on a natural situation which is sent by a user, using a voice noise reduction algorithm to carry out filtering noise reduction on an input statement, and determining a matched feedback statement; if the matched feedback statement does not exist, according to a neural network dialogue model, giving an answer of the statement sent by the user, which includes the following steps of constructing a coding layer of a user sending statement model to be a first neural network; analyzing a user sending statement in the first neural network and acquiring a first intermediate vector used for expressing a user sending statement meaning; and constructing a decoding layer of a dialogue generation model to be a second neural network, analyzing the intermediate vector in the second neural network, and acquiring a vector group used for expressing a statement answer. In the invention, a threshold voice noise reduction algorithm is used to acquire a small mean square error and a signal to noise ratio of a reconstruction voice signal is increased.

The invention discloses a crossroad-traffic-signal-lamp control algorithm based on a neural network. The crossroad-traffic-signal-lamp control algorithm includes the steps that driving speeds and accelerated speeds when vehicles pass through crossings are detected, and single-vehicle delay time is estimated according to the quantity of traffic flow; according to the estimated single-vehicle delaytime, evaluation indexes of a whole road section are obtained; road evaluation indexes are defined according to urban road grades; a neural network model is established, and a signal lamp control strategy according to crossing traffic flow is given, wherein the delay time, the evaluation indexes of the whole road section, road evaluation indexes and walker passing time serve as an input layer of input variables; a first neural network is established, the vehicle delay time and the evaluation indexes are analyzed in the first neural network, and a vector group expressing the signal lamp controlstrategy is obtained; the vector group expressing the signal lamp strategy serves as control strategy to be output, and the congestion problem of the tide traffic flow phenomenon is effectively solved.

The invention discloses a compressed sensing image reconstruction method based on relevance vector grouping, which mainly solves the problems of inaccuracy and low robustness of compressed sensing image reconstruction. The realization process is as follows: 1) receiving an observation matrix and an observation vector; 2) obtaining an initial relevance vector by the observation vector and a sending matrix; 3) dividing the relevance vector into sub-relevance vectors according to the spatial neighbourhood relationship of wavelet coefficients; 4) adding a component in each sub-relevance vector and sequencing the components; 5) updating the reconstructed wavelet high-frequency coefficients and observation vectors on the basis of a Bayesian framework according to the sequencing order; 6) carrying out invert wavelet transform on the reserved low-frequency wavelet decomposition coefficients and the reconstructed high-frequency wavelet coefficients to obtain a reconstructed image. Compared with OMP and BEPA methods, the compressed sensing image reconstruction method based on relevance vector grouping disclosed by the invention has the advantages of high quality and good robustness of the reconstructed image, and can be used for reconstruction for natural images and medical images.

An apparatus and method of motion estimation and mode decision includes a candidate vector group generation unit generating a candidate vector group including a motion vector of an adjacent image crossing a first macro block, where the first macro block is a macro block of an image before scaling, and the first macro block and the adjacent image exist in a same sub-group of a picture (sub-GOP). A motion vector estimation unit estimates a base vector of a second macro block from the motion vector and the candidate vector group of the first macro block overlapped with the second macro block, where the second macro block is a macro block of an image after scaling. A mode decision unit determines a mode of the second macro block from the estimated base vector and a mode of the first macro block.

The invention discloses a data security protection method and device. The method comprises the following steps that factors including the data asset Z, the data threat T and the data vulnerability V which are involved in data security risks are determined; a data risk degree matrix R is determined according to the data asset Z, the data threat level T and the data vulnerability level V; a weight vector group G related to the data asset Z, the data threat T and the data vulnerability V is determined; the data risk level membership level Y is determined according to the data risk matrix and theweight vector group, wherein Y=G*R; and a corresponding data protection scheme is determined according to the risk level membership degree. According to the technical scheme, the problems that in an existing network security risk assessment method, the assessment index is single, and the assessment process is unreasonable are well solved, and therefore the more targeted data protection scheme is conveniently adopted.

The invention provides an object name resolution system and method in the internet of things. The method comprises: a normal node sends an object query name resolution request to a super node in a vector group corresponding to the super node; the super node determines the grouping of the second level of Chord in the first level of Chord according to the identification code of manufacturers; the super node inquires a URL (uniform resource locator) corresponding to the product identification code in the grouping; and a destination node corresponding to the URL returns the object name service address to users. The first level of Chord is the network constituted by the super node groups of each grouping based on a distributed hash table; the second level of Chord is constituted by each grouping, and each grouping is responsible for the object name service of an organization; and the first level of Chord and the second level of Chord are interconnected through the super node group in each grouping. The object name resolution system in the invention can be used for solving the problems of performance bottleneck in existing high levels, low ability in resisting attacks, frequent configuration errors and unbalanced loading and the like, and supporting multiple encoding methods.