18 results about "Signature vector" patented technology

The invention discloses an exoskeleton robot control parameter determination method and system, and particularly relates to the technical field of robot control, and the method comprises the steps: synchronously collecting rigid joint torque, flexible cable tension and inertia signals, generating a dual-channel response spectrum through Fourier transform, and obtaining a frequency ratio vector; a kinetic energy coupling index and a viscosity lag index are calculated, and a bandwidth difference index is generated through the cooperation phase model in combination with the inertia-damping signature vector; a rigid gain matrix and a flexible gain matrix are obtained in a self-adaptive mode under the energy consumption constraint through sparse Bayesian optimization; constructing a weighting function according to the motion prediction phase and the energy consumption constraint, and synthesizing the two gain matrixes into a composite control instruction stream; and expected and actually measured moment residual errors are monitored, a gain increment is generated according to gradient perception when the threshold is exceeded, and a coupling feature extraction unit is written back, so that millisecond-level gain correction is realized, and the problem that coupling oscillation is caused by unified gain mismatch due to the fact that energy mutual feedback and phase delay cannot be accurately quantified in the prior art is solved.

The present embodiments disclose devices, methods, apparatuses and computer readable storage media of low complexity beamforming. The method comprises determining, at a network device, a channel covariance matrix and at least one signature vector, the at least one signature vector characterizing features of a signal associated with the terminal device; constructing one or more beam sets based on the channel covariance matrix and the at least one signature vector; selecting a predefined number of beams from the one or more beam sets; and performing a transmission to the terminal device based on a target beamforming matrix determined at least based on the predefined number of beams.

The invention relates to an unmanned aerial vehicle cluster communication network full-end reliability prediction method based on GNN and network signature, and belongs to the technical field of communication networks, and adopts two-stage prediction: a first stage, determining topological structure characteristic variables of an unmanned aerial vehicle cluster communication network, generating a training data set based on simulation, and designing, training and optimizing a graph neural network model; the prediction of network signature vectors is realized, and the signature only depends on network structure characteristics and does not depend on link reliability parameters; and in the second stage, on the basis of the signature vector, the all-end reliability of the network in a link failure scene is calculated by combining link reliability parameters and utilizing a signature-based all-end reliability deterministic hybrid expression. According to the method, the sensitivity of the model to the reliability parameter change is reduced, and the generalization ability and applicability of the prediction method under different network scales and reliability conditions are improved.

Aspects of the disclosure relate to multi-factor authentication. A computing system may receive, from an authenticated entity, a plurality of authenticated inputs. Normalized authenticated inputs may be generated. An n-dimensional authenticated signature vector corresponding to the plurality of normalized authenticated inputs may be generated. A plurality of multi-factor authentication (MFA) inputs may be received. A plurality of normalized multi-factor authentication inputs may be generated. An n-dimensional multi-factor authentication vector corresponding to the plurality of normalized MFA input may be generated. There may be a determination of whether a distance between the n-dimensional authenticated signature vector and the multi-factor authentication vector not exceeding an authentication distance threshold. Based on the distance between the n-dimensional authenticated signature vector and the multi-factor authentication vector not exceeding the authentication distance threshold, an indication that the plurality of multi-factor authentication inputs have been authenticated may be generated.

The invention relates to the technical field of image registration, and discloses a laser speckle fundus image registration method and system, and the method comprises the steps: reading a laser speckle fundus video, dispersing the laser speckle fundus video into an original image sequence, determining a reference frame and a floating frame, generating an enhanced fundus image through spherical projection correction, and obtaining a multi-scale vascular skeleton hierarchical expression; constructing a local blood vessel tree signature vector based on multi-scale blood vessel skeleton hierarchical expression, and calculating a feature matching cost matrix; and finally, constructing a topological graph by using the candidate feature point set, screening a definite matching point pair set, calculating a spatial geometric transformation matrix, and outputting a high-signal-to-noise-ratio eye fundus image after floating frame registration is completed. According to the method, the feature point matching accuracy and the registration success rate of the multi-vessel cross point dense region can be improved, and the cascade registration failure phenomenon caused by wrong matching is eliminated.

The invention discloses a digital signature method based on an approximate trap door under an on-lattice standard model, and the method comprises the steps: generating a public parameter, a main public key and a main private key through a system parameter generation device according to a system safety parameter; generating a signature vector of the to-be-signed message through a signature generation device according to the main public key, the main private key and the to-be-signed message; verifying the validity of the signature through a signature verification device according to the main public key, the to-be-signed message and the signature vector; wherein in the signature generation process, an approximate trap door expansion generation device and an approximate trap door sampling device are adopted, and the approximate trap door sampling device is used for carrying out geometric compensation on distribution deviation caused by using an approximate trap door, so that output distribution of signature vectors and ideal spherical discrete Gaussian distribution are indistinguishable in statistics. According to the method, significant compression of the public key and the signature size is realized, and an efficient signature solution is provided for quantum cryptography.

The present embodiments disclose devices, methods, apparatuses and computer readable storage media of low complexity beamforming. The method comprises determining, at a network device, a channel covariance matrix and at least one signature vector, the at least one signature vector characterizing features of a signal associated with the terminal device; constructing one or more beam sets based on the channel covariance matrix and the at least one signature vector; selecting a predefined number of beams from the one or more beam sets; and performing a transmission to the terminal device based on a target beamforming matrix determined at least based on the predefined number of beams.

The application relates to a kind of GNN and network signature-based UAV cluster communication network full end reliability prediction method, belong to communication network technical field, adopt two-stage prediction: first stage, determine the topological structure characteristic variable of UAV cluster communication network, based on simulation generation training data set, design, train and optimize graph neural network model, realize the prediction of network signature vector, signature only relies on network structure characteristics, does not rely on link reliability parameter;Second stage, based on signature vector, combined with link reliability parameter, use the full end reliability certainty mixed expression based on signature, calculate the full end reliability of network under the link failure scene.The application reduces the sensitivity of model to the change of reliability parameter, improves the generalization ability and applicability of prediction method under different network scale and reliability conditions.

The application relates to the technical field of image registration, and discloses a laser speckle fundus image registration method and system, which comprises the following steps: reading a laser speckle fundus video and discretizing the laser speckle fundus video into a sequence of original images, determining a reference frame and a floating frame, generating an enhanced fundus image after spherical projection correction, and obtaining a multi-scale blood vessel skeleton hierarchical expression; then constructing a local blood vessel tree signature vector based on the multi-scale blood vessel skeleton hierarchical expression, calculating a feature matching cost matrix; finally, constructing a topological graph by using a candidate feature point set, screening a set of certain matching point pairs, calculating a spatial geometric transformation matrix, outputting a high signal-to-noise ratio fundus image after floating frame registration is completed, and improving the feature point matching accuracy and the registration success rate of a multi-blood vessel fork point dense area, and eliminating the cascade registration failure phenomenon caused by false matching.

Example embodiments of the present disclosure provide solutions for low-rank approximation based beamforming. In an example method, a device obtains low-rank channel information based on an original channel estimate matrix, wherein the original channel estimate matrix has a full dimension corresponding to a number of antennas, and the low-rank channel information has a reduced dimension. The device obtains at least one low-rank signature vector with the reduced dimension. The device determines a low-rank subspace beam set based on the low-rank channel information and the at least one low-rank signature vector. The device transforms at least one beam of the low-rank subspace beam set to the full dimension. The proposed solutions take advantages of both Krylov and Nyström and can achieve spectral efficiency performance of baseline methods while guaranteeing a low computational complexity.

Techniques are disclosed relating to automatically determining whether an entity is malicious. In some embodiments, a server computer system generates a feature vector for an unknown website, where generating the feature vector includes preprocessing a plurality of structural features of the unknown website. In some embodiments, the system inputs the feature vector for the unknown website into a trained neural network. In some embodiments, the system applies a clustering algorithm to a signature vector for the unknown website and signature vectors for respective ones of a plurality of known websites output by the trained neural network. In some embodiments, the system determines, based on results of the clustering algorithm indicating similarities between signature vectors for the unknown website and one or more of the signature vectors for the plurality of known websites, whether the unknown website is suspicious. Determining whether the entity is suspicious may advantageously prevent malicious (fraudulent) activity.

This application discloses a method for denoising and classifying high-frequency partial discharge signals from GIS equipment in substations, relating to the field of signal denoising. The method includes: sampling partial discharge signals from GIS equipment within the target substation; mapping the partial discharge signals to high-dimensional point cloud trajectories and converting these trajectories into signal trajectory signature vectors; performing sliding sampling of the signal trajectory signature vectors to obtain a set of signal trajectory points; quantizing the spatial quality density and performing weighted complex modeling on the signal trajectory point set to obtain a set of discharge signal skeleton points; constructing a probability density function and calculating the discharge signal score vector; extracting the discharge signal feature vector; constructing a discharge signal heterogeneity map and performing signal denoising on the partial discharge signals to obtain a low-noise discharge signal sequence; and generating a substation equipment fault report for the GIS equipment within the target substation. This application can effectively improve the accuracy of partial discharge signal denoising and classification.

Techniques of generating a lattice-based verification matrix and signature vector are disclosed. The method enables a generating device to sample a gadget matrix and then generate a reduced gadget matrix. The generating device may then sample a trapdoor matrix and use the trapdoor matrix and the reduced gadget matrix to generate a verification matrix. A sending device may receive the trapdoor matrix and the verification matrix from the generating device, in addition to receiving a message. The sending device may then use the trapdoor matrix and the verification matrix to generate a signature vector for the message. A verification device can receive the verification matrix, the message, and the signature vector. The verification device may use the verification matrix and the signature vector to verify the message.

Techniques are disclosed relating to automatically determining whether an entity is malicious. In some embodiments, a server computer system generates a feature vector for an unknown website, where generating the feature vector includes preprocessing a plurality of structural features of the unknown website. In some embodiments, the system inputs the feature vector for the unknown website into a trained neural network. In some embodiments, the system applies a clustering algorithm to a signature vector for the unknown website and signature vectors for respective ones of a plurality of known websites output by the trained neural network. In some embodiments, the system determines, based on results of the clustering algorithm indicating similarities between signature vectors for the unknown website and one or more of the signature vectors for the plurality of known websites, whether the unknown website is suspicious. Determining whether the entity is suspicious may advantageously prevent malicious (fraudulent) activity.

This invention discloses a lidar loop closure detection method based on hierarchical frequency domain feature extraction. The method includes: a robot collecting point cloud data of the current driving environment using an onboard lidar sensor during its movement, acquiring original lidar point clouds, with each lidar frame corresponding to one original lidar point cloud; dividing the original lidar point cloud into multiple layers according to height, and acquiring the frequency domain signature vectors of the layered sub-point clouds; integrating all frequency domain signature vectors of the lidar frames to obtain a frequency domain signature matrix; determining loop closure constraint factors and correcting the robot pose based on the frequency domain signature matrix corresponding to the current lidar frame acquired by the robot and the frequency domain signature matrices corresponding to the lidar frames to be matched in the historical trajectory; and correcting the robot's motion behavior based on the corrected robot pose. This invention has the advantages of high loop closure detection rate, high detection accuracy, strong detection robustness, and strong real-time performance.

The invention relates to an equipment abnormal vibration positioning method and system, equipment and a storage medium, and the positioning method comprises the following steps: obtaining historical vibration data when the equipment works normally; acquiring a baseline projection matrix according to the historical vibration data; acquiring real-time multi-channel sensor data; acquiring an initial projection matrix according to a multi-channel sensor; the initial projection matrix is optimized through a genetic algorithm, feature manifold data are obtained, and a fitness calculation formula in the genetic algorithm comprises a baseline projection matrix, a topological separation degree and a causal positioning confidence coefficient; calculating the topological characteristics of the characteristic manifold data to obtain a durability chart, calculating the topological separation degree, and endowing the topological separation degree in the fitness calculation formula with the value of the topological separation degree; obtaining an abnormal signature vector according to the topological features; and performing causal reasoning in the equipment knowledge graph according to the abnormal signature vector, and positioning an abnormal vibration source.

A method and system are configured to sample a plurality of strings from a data source, the data source having an unknown classification, construct an empirical Hankel matrix from the strings, generate an embedding of the data source in a k-dimensional space by extracting the top k singular values of the empirical Hankel matrix, repeat the sampling, constructing, and computing steps over n independent trials to produce n signature vectors, and aggregating the n signature vectors into a mean vector and a standard deviation vector to create a query point associated with the data source, query a memory with the query point, the memory storing known embeddings, each known embedding comprising a respective known mean vector and a respective known standard deviation vector having a known classification label, to estimate, for the query point, a similarity measure to the known embeddings, and generate the classification label as a proposed classification of the data source responsive to the similarity measure exceeding a predetermined threshold.