115 results about "Group model" patented technology

A speech recognition method and apparatus perform speaker clustering and speaker adaptation using average model variation information over speakers while analyzing the quantity variation amount and the directional variation amount. In the speaker clustering method, a speaker group model variation is generated based on the model variation between a speaker-independent model and a training speaker ML model. In the speaker adaptation method, the model in which the model variation between a test speaker ML model and a speaker group ML model to which the test speaker belongs which is most similar to a training speaker group model variation is found, and speaker adaptation is performed on the found model. Herein, the model variation in the speaker clustering and the speaker adaptation are calculated while analyzing both the quantity variation amount and the directional variation amount. The present invention may be applied to any speaker adaptation algorithm of MLLR and MAP.

The invention discloses a game theory-based multi-micro-grid interconnection running optimization method. The method includes the following steps: analyzing load demand response characteristics of allmicro-grids, and constructing a micro-grid group model; then initializing a system, establishing a day-ahead game model, and calculating a net load of the i-th micro-grid in a j-th period and a totalnet load amount of the system by day-ahead forecast data; obtaining respective optimal strategies by solving by all the micro-grids according to initial status, carrying out information exchange in amicro-grid group, sharing respectively obtained best strategy information, and updating status information; judging, by a system control center, whether Nash equilibrium is reached, and if the systemmeets an equilibrium condition, ending a game, and outputting a final optimization strategy set; and otherwise, turning ahead to carry out optimization again according to updated status information.The method can give consideration to both the stability and the economy of the system, and realize individual benefit maximization and optimal group running of the micro-grids at the same time.

The invention provides a driving track real-time risk assessment framework based on internet of vehicle big data. A track scoring system is established by using the vehicle feature information and theclaim information stored in a database and the driving track information collected from a vehicle-mounted sensor in real time, wherein the track scoring system includes a vehicle grouping model, a track clustering model and a behavior claim model so that real-time scoring of the driving tracks can be realized, the reference is provided for vehicle insurance premium pricing, insurance premium adjustment, insurance premium prediction and creating the new vehicle insurance, and thus the insurance premium price preference is given to the tracks of great driving environment and operation behaviorsand the insurance premium price penalty is given to the tracks of poor driving environment and operation behaviors.

A state based adaptive feedback / feedforward PID controller includes a model set component, communicatively coupled to a process input, having a state variable defining a number of process regions, and a number of models grouped into the process regions. Each of the grouped models includes a plurality of parameters having a value selected from a set of predetermined initial values assigned to the respective parameter. The adaptive controller further includes an error generator communicatively coupled to the model set component and a process output. The error generator configured to generate a model error signal representative of the difference between a model output signal and a process output signal. The error generator, communicatively coupled to a model evaluation component, is configured to compute a model squared error corresponding to a model and correlating the model squared error to parameter values represented in the model. The adaptive controller further includes a parameter interpolator communicatively coupled to the model evaluation component for calculating a respective adaptive parameter value for parameters represented in the model and a controller update component, communicatively coupled to the parameter interpolator, for updating the controller in response to adaptive parameter values upon conclusion of an adaptation cycle.

Websites operated by relationship building service providers collect information from users for use in identifying candidate matches for a user. At least a portion of the information may include freeform text entered by the user to identify the user's interests. The freeform text entered by users may be preprocessed and a user interest group model formed using a generative Natural Language Process that is applied to determine user interest groups from the freeform text. Such user interest groups identify related words based on the frequency of their occurrence and relationship within the freeform text. Individual user interest group membership vectors for each user included in a number of users may be determined using the user interest group model. At least one compatibility aspect between users may then be assessed based on the similarity or distance between the user interest group membership vectors associated with each user.

An affine transformation analysis system and method is provided for matching two images. The novel systolic array image affine transformation analysis system comprising a linear rf-processing means, an affine parameter incremental updating means, and a least square error fitting means is based on a Lie transformation group model of cortical visual motion and stereo processing. Image data is provided to a plurality of component linear rf-processing means each comprising a Gabor receptive field, a dynamical Gabor receptive field, and six Lie germs. The Gabor coefficients of images and affine Lie derivatives are extracted from responses of linear receptive fields, respectively. The differences and affine Lie-derivatives of these Gabor coefficients obtained from each parallel pipelined linear rf-processing components are then input to a least square error fitting means, a systolic array comprising a QR decomposition means and a backward substitution means. The output signal from the least square error fitting means then be used to updating of the affine parameters until the difference signals between the Gabor coefficients from the static and dynamical Gabor receptive fields are substantially reduced.

A method and system for detecting a damaged machine component during operation of the machine are provided. The damaged machine component detection system includes one or more processors, one or more memory devices communicatively coupled to the one or more processors, an image capture device configured to generate a stream of temporally-spaced images of a scene including a machine component of interest, the images generated in real-time, and a grouping model configured to compare features of a current image of the stream of images to features of previously captured images sorted into a plurality of groups of images having similar features, generate an alert if predetermined features of the current image deviate from corresponding features of the grouped images by a predetermined amount, and update the groups with the current image if the predetermined features of the current image are similar to corresponding features of the grouped images by a predetermined amount. The system also includes an output device configured to output at least one of the alert and the current image.

The invention discloses an outsourcing encryption and decryption CP-ABE method capable of achieving user revocation, and relates to the technical field of data encryption and decryption in a mobile cloud environment. The method further achieves the outsourcing encryption operation on the basis of conventional outsourcing attribute encryption, and reduces the encryption calculation cost of a local client. The method achieves the access control of fine granularity of a file on a cloud server, simplifies the secret key management operation for a user, and achieves the revocation of the access authority of an unauthorized user. When the user is revoked, a legal user does not need to update a secret key, and alleviates the burden of an authorization mechanism. The local user just needs to carry out the simple XOR operation to update a cryptograph. The security proof indicates that the method provided by the invention has the non-adaptive selection plaintext safety under a common dual linear group model. The performance analysis indicates that the method reduces the encryption calculation cost of mobile equipment more effectively, can achieve the revocation of an illegal user efficiently, and achieves the control of the access authority of the user effectively.

A conception of object dimension is introduced into the area, where the administrated objects (AO) are rather concentrative. The definition of AO is finely compartmentalized. The AO are compartmentalized into manifold groups under different dimensions. Nesting between groups is allowed. A multi-layer tree is formed. The different dimension represents different sorting modes of AO. The system supports several sorting modes co-existed. The system authorities are split into data authority and operate authority. For the data authority, all data entity in this system controlled by the data authority must possess corresponding safety administration objects (SAO). Associating these SAO with different Role, the invention can complete the data authorities splitting. All SAO in this model are grouped. Multi-grouping all SAO in the system in the same running are supported. Each grouping model is named a 'dimension'. Each value on the dimension is corresponding to a certain group.

The invention relates to a city building complex and rendering method considering city image. According to the method, a fine continuous facet element in a building group model is abstracted to be an intuitive two-dimensional shape, by means of a multi-resolution expression binary tree (EBT), the city image is rendered, wherein the level of detail (LOD) of the city image is more flexible than that defined by the city geography markup language (GML) standard, and the readability and the verisimilitude of the city image are simultaneously guaranteed. According to the city building complex and rendering method considering the city image, an automatic road analysis algorithm is designed for extracting relevant space elements of the city image so as to assist grouping and combining of buildings, and by means of an indexing mechanism, an abstract edge Fedge, a building group example and the EBT are stored to serve as city building complex results so that the complex results are subjected to flexible pre-processing before the LOD scene of the city image is rendered and model updating is facilitated. The city building complex and rendering method considering the city image real-timely subjects a three-dimensional digital city model to visualized rendering and has a wide application prospect in aspects of city navigation, spatial position services and the like.

A speech recognition method and apparatus perform speaker clustering and speaker adaptation using average model variation information over speakers while analyzing the quantity variation amount and the directional variation amount. In the speaker clustering method, a speaker group model variation is generated based on the model variation between a speaker-independent model and a training speaker ML model. In the speaker adaptation method, the model in which the model variation between a test speaker ML model and a speaker group ML model to which the test speaker belongs which is most similar to a training speaker group model variation is found, and speaker adaptation is performed on the found model. Herein, the model variation in the speaker clustering and the speaker adaptation are calculated while analyzing both the quantity variation amount and the directional variation amount. The present invention may be applied to any speaker adaptation algorithm of MLLR and MAP.

The invention discloses a new energy active power control method for a power generation right transaction in an electricity market environment. The new energy active power control method comprises the steps of: regarding the tracking of real-time lower spinning reserve of a thermal power generating unit as a general objective of new energy control, and calculating new energy power generation indexes; establishing a control group model according to power generation right transaction components, wherein power generation priority of each control group conforms to the priority of the corresponding component electric quantity; initializing grouping according to the power generation right component with the highest priority in each new energy field stations; carrying out intergroup distribution on the new energy power generation indexes according to the priorities of the control groups; performing precise control on the new energy field stations in the control groups after considering section constraint conditions and actual power generation capacities of different levels; and counting currently participating transaction component electric quantity of each new energy field station in real time, and automatically switching the control groups to realize hierarchical control of the power generation right after the component electric quantity index is completed until all power generation right electric quantity is completed. The new energy active power control method is of great importance in motivating new energy power generation enterprises to participate in the power generation right transaction and enhancing the level of new energy consumption.

The invention provides a content push method and device, and a computer device. The content push method includes the steps that user characteristics of an exposure application program and click behavior of a corresponding application program are acquired; a sample is generated according to the user characteristics and the click behavior of the application program; the sample is inputted into a grouping model to determine a group to which the sample belongs in a grouping model; the sample is inputted into a corresponding click rate prediction model of the group, and the estimated click rate ofthe application program by a user is obtained; the content is pushed to the user according to the estimated click rate of the application program. After the group to which the user belongs is determined by the content push method, the estimated click rate is obtained according to the click rate prediction model of the group, the accuracy of the estimated click rate is improved, and the content ispushed to the user accurately.

The invention provides a Chinese reading ability evaluation method, which comprises the following steps of: obtaining user information of a user, and calling an initial test question group corresponding to the user information in an evaluation question bank for the user to test; Calculating an initial capability value of the user according to a test result, obtaining an evaluation test paper corresponding to the initial capability value, and sending the evaluation test paper to a terminal where the user is located for the user to perform evaluation, the evaluation test paper comprising test question groups of multiple question types of continuous texts and discontinuous texts; And based on the project response theory, calculating a reading capability value of the user for answering the evaluation test paper by utilizing a preset high-order question group model. The invention also provides a Chinese reading ability evaluation device and a readable storage medium. According to the methodand the system, the user is tested according to the test papers of various question types of the continuous text and the discontinuous text, the item response of the user for answering the test papers is evaluated through the high-order question group model, and the question types have diversity and can more accurately and comprehensively inspect the Chinese reading capability of the user.

The invention discloses a pattern recognition classification method expressed based on grouping sparsity, comprising the steps of: obtaining an initial expression of a sample to be recognized by solving a least square solution of a linear equation; compensating a smaller grouping coefficient in the solution space of the linear equation, gradually enhancing the sparsity of solution vectors in the meaning of a grouping sparse model, and carrying out repeated iteration until constringency to obtain the grouping sparse expression of the sample; and judging the classification of the sample to be recognized as the largest grouping of the corresponding coefficient according to the obtained sparsity, and balancing the confidence coefficient by the concentration degree of the distribution in each group of the coefficient with the sparsity. The grouping model adopted by the invention is more suitable for the requirement on the classification, and improves the recognition capability. The sparsity of the solution is improved by combining the method of compensating the coefficient in the solution space, and the calculation amount is reduced. The method is not only suitable for the classification of pattern recognition, but also can be used in the fields of compressed sensing, and the like, and has wide application prospect.

The invention discloses a speech recognition system and a recognition method thereof. The system comprises a speech acquiring module, a speech pretreatment module, a speech feature extraction module, a grouping judgment module and a speech recognition module, wherein the grouping judgment module is used for grouping speeches in a clustering mode; the speech acquiring module is connected with the speech pretreatment module, the speech pretreatment module is connected with the speech feature extraction module, the speech feature extraction module is connected with the grouping judgment module, and the grouping judgment module is connected with the speech recognition module; the grouping judgment module comprises a grouping judgment unit and at least two grouping models; the speech feature extraction module is connected with the grouping judgment unit, and the grouping judgment unit is respectively connected with the at least two grouping models; and the grouping models are connected with the speech recognition module.

The invention discloses a method and device for determining the temperature of cable group cable cores based on distributed optical fiber temperature measurement, and an installation method of a distributed optical fiber temperature measuring device. The method for determining the temperature of a cable group comprises the steps of measuring parameters of the environment where the cable group is laid, and parameters of cables in the cable group; establishing a cable group model according to the parameters of the environment and the parameters of the cables; reading a current of a conductor of the cable group, and according to the cable group model, conducting simulation computation on the skin temperature of the cables to obtain the estimated skin temperature Tc of the first cables in the cable group; measuring the actual skin temperature Tb of the first cables in the cable group in the axial direction; according to the estimated skin temperature Tc and the actual skin temperature Tb, computing a conversion coefficient alpha; according to the conversion coefficient alpha, correcting the estimated skin temperature of the cables, and computing actual skin temperature of each cable in the cable group. According to the method and device, the computational accuracy of the temperature of the cable group can be improved, and the temperature computation range of the cable group can be increased.

A method and system for detecting a damaged machine component during operation of the machine are provided. The damaged machine component detection system includes one or more processors, one or more memory devices communicatively coupled to the one or more processors, an image capture device configured to generate a stream of temporally-spaced images of a scene including a machine component of interest, the images generated in real-time, and a grouping model configured to compare features of a current image of the stream of images to features of previously captured images sorted into a plurality of groups of images having similar features, generate an alert if predetermined features of the current image deviate from corresponding features of the grouped images by a predetermined amount, and update the groups with the current image if the predetermined features of the current image are similar to corresponding features of the grouped images by a predetermined amount. The system also includes an output device configured to output at least one of the alert and the current image.

The invention discloses an energy internet management method based on intraday rolling optimization. An energy local area network group model is constructed, a system is initialized and initial data is required; and then, intraday rolling optimization is carried out, errors of actual wind and light output, load fluctuation and initial prediction data are corrected, the purpose of correcting net load fluctuation deviation is achieved, and each ELN sub-network decides an optimal strategy of itself according to the updated internal electricity price calculated by the network group control centerand corrects each variable. After the correction is finished, the current optimization result is taken as an actual scheduling scheme, the arrival of a next optimization period is waited for, and theintraday rolling optimization process is repeated. According to the invention, the new energy consumption capability of the energy Internet and the system reliability in fault can be effectively improved, and the economic benefit of the energy Internet is increased to a certain extent.

In a graphical modeling environment supporting a model having at least two different analysis frameworks operating therein, a system and corresponding method of processing the graphical model modify the model to group model portions together for processing in the same analysis framework. Model parts are identified and associated with the analysis framework in which they operate. Model parts are then grouped based on their association with their analysis framework to form model portions that operate in one of the different analysis frameworks. In instances where topological separation of model portions operating in the same analysis framework occurs, the system and method reconfigure intervening model portions to be amenable with operation in the analysis framework of the surrounding model portions to improve processing efficiency.

The invention relates to an image fusion method based on a convolution neural network, belonging to the field of information fusion and image processing. The invention obtains a fusion picture by training and using a convolution neural network. Convolution neural network is trained in advance by selecting training set according to the pictures to be fused, and the whole training process belongs tosupervised training. In the process of training, it involves the analysis and synthesis of pictures. Then the analysis of the trained two-group model weight feedforward network and the synthesis of the feedback network are used for the fusion model of the deep neural network. The fusion rules in the training and fusion process adopt the sigmoid function fusion rules based on depth learning. The invention avoids the omission of redundant information and related information in the fusion process.

A client device can receive information about a population to which a user belongs. During operation, the client device determines information about a user, determines a group identifier for the user, and communicates the determined information about the local user and the group identifier to a group-modeling server. The client device then receives a group-activity model that corresponds to the group identifier, and generates a user-activity model for the local user based on the group-activity model and the determined information about the local user. The client device uses the user-activity model to compute an activity probability for a corresponding target activity. The group-modeling server receives user information from a plurality of client devices of a group, and generates a group-activity model for the group based on the user information. The server then sends the group-activity model to users of the identified group.

The invention discloses a load control method based on an electric water heater load group model. The method comprises the following steps that 1), an electric water heater model with water flowing ina water tank considered is established; 2), clustering is carried out on electric water heater indexes in a load dispatching center control area on the basis of probability distribution, and an electric water heater load group probability model is established; 3), a heat storage state variable of the electric water heater and a calculation method of the heat storage state variable are defined according to the water temperature in the electric water heater; and 4), controllable subsets of the water heater load group are selected according to the extent of the heat storage state of the water heater in the load group, the controllable subsets of the load group of the water heater are sequenced on the basis of a sequencing method taking the comfort level of a user into consideration, and theload control is completed according to the sequencing result. According to the method, due to the fact that the temperature distribution and the natural convection process in the electric water heaterare considered, the model is objective and accurate; and the heat storage state is defined, so that the comfort level of the user is guaranteed to a certain extent, and a great assistant effect is applied for propelling the temperature control to reach and participate in the power grid frequency control.

The invention discloses a crowd motion simulation method and system. The method is completed in a simulation server and the specific process comprises the following steps: receiving a scene model and crowd individual information, and respectively initializing the scene and the crowd individual information; performing grouping by using the K-Medoids algorithm with multiple influence factors according to the distances and relationships among individuals in the crowd to obtain a grouping condition of the crowd; and using the grouping condition of the crowd as an initial state of crowd motion, performing local path planning by using a companion group model based on affinity to realize group behaviors and collision prevention, and figuring out the motion path of each individual to arrive at a selected exit; and rendering the motion path and displaying a simulation result in real time.

The invention provides an HMM voiceprint recognition signing-in method and system based on grouping statistics. According to the method, at first, a signing-in client side collects voiceprint signals of signers, pre-emphasis, framing, windowing and end point detecting are sequentially carried out on the voiceprint signals, the voiceprint signals are transmitted to a server through a network, and then a signing-in server side carries out voiceprint characteristic parameter extraction on the voiceprint signals to generate signer grouping characteristic parameters and generate grouping characteristic parameters of a grouping model; According to the grouping characteristic parameters of each group of the grouping model and the grouping characteristic parameters of signer voiceprints, whether the signers are members of a certain group is judged; finally, the voiceprints of the signers are judged. According to the method, under the circumstance that the number of the signers is large, real-time efficient voiceprint recognition signing is achieved, investment on public equipment is reduced, use is convenient and efficient, by means of the combination of the system and the method, voiceprint recognition signing is achieved with high recognition efficiency and high real-time property.

The invention discloses a rapid crowd evacuation simulating method and device based on geographic information. The method includes the following steps of obtaining a two-dimensional map of a scene, extracting the geographic information based on the two-dimensional map, reestablishing a road model and a building group model, establishing a route semantic topological map according to the topologicalrelation between road framework data and a road, conducting crowd movement calculation based on the normal distribution relative speed obstruction method and scene semantics, rendering the crowd movement calculation result, the road and the building group model in combination with a realistic rendering technology to obtain crowd evacuation simulating animation. The method is low in complexity, and crowd evacuation simulating can be efficiently conducted.

The invention discloses a method for realizing a personal virtual application firewall based on Openstack. The method comprises the following operation steps of selecting a protocol analysis engine; expanding Neutron security group rules; designing a Neutron application security group model; increasing models and tables for personal application firewall rules; expanding a Neutron client part and increasing definition of personal application firewall rule group management interfaces; expanding a Neutron server part and increasing storage and issuance of a personal application firewall rule group; increasing the query response of ovs-agent to the security group rules; expanding an agent end and realizing the effect of the personal application firewall through iptables+ndpi+tc. The method has the beneficial effects that the Openstack realizes the security group, the security group is equivalent to a personal state firewall of a PC end in a physical network, through expansion of the security group, the fined control for an application is realized, and the security of network boundaries is facilitated.

The invention discloses a method for establishing a contact group in a mobile terminal and the mobile terminal. The method includes the steps that the mobile terminal receives condition attribute information of establishing a first contact group through a man-machine interface; the mobile terminal generates a first group model according to the condition attribute information; the mobile terminal matches attribute information of contacts stored in the mobile terminal according to the first group model to generate a first contact group. Due to the application of the method for establishing the contact group in the mobile terminal and the mobile terminal, the problem that in relative technologies, the only group can not meet requirements of users on different contacts as the establishing conditions of contact groups are single can be solved; and further users can establish contact groups in user-defined mode, multi-dimensionality of contact groups is realized, and improve systematical performance and user experience are improved.

The invention discloses a particle swarm algorithm-based neuron swarm model parameter adaptive optimization method. The method comprises the steps of S1, initializing a particle swarm, setting basic parameters of the particle swarm algorithm, setting a parameter combination search range of each particle, and setting basic parameters of a neuron swarm model; S2, calculating a fitness value of eachparticle in the initial particle swarm, and initializing a particle individual extremum and a global extremum according to the fitness value of the particle; S3, updating the particle speed and position; s4, calculating a new particle fitness value, and updating a particle individual extreme value and a global extreme value; s5, judging whether the maximum number of iterations is met or not, if yes, outputting globally optimal particles, and otherwise, returning to the step S3; and S6, obtaining an optimal parameter combination of the electroencephalogram frequency band according to the globally optimal particles output in the step S5. The invention provides a convenient and efficient neuron group model parameter adjustment method, the parameter identification accuracy is improved, and theadjustment time is shortened.