135 results about "Bayesian probability" patented technology

An “unstructured event parser” analyzes an event that is in unstructured form and generates an event that is in structured form. A mapping phase determines, for a given event token, possible fields of the structured event schema to which the token could be mapped and the probabilities that the token should be mapped to those fields. Particular tokens are then mapped to particular fields of the structured event schema. By using the Naïve Bayesian probability model, a “probabilistic mapper” determines, for a particular token and a particular field, the probability that that token maps to that field. The probabilistic mapper can also be used in a “regular expression creator” that generates a regex that matches an unstructured event and a “parameter file creator” that helps a user create a parameter file for use with a parameterized normalized event generator to generate a normalized event based on an unstructured event.

The present invention is uniquely designed to interact with web retailer's customers with real agent reaction times as they give astute answers directly concerning web retailer's products and goals. The entire process is manageable through a third-party website which includes scripting, settings and other parameters selected by the web retailer. An artificial intelligence engine uses the combination of Bayesian probability keyword selection, natural language parsing and regular expression processing. The technology updates its response database with every client interaction-learning always takes place as it simulates a live agent, in real time. Every client interaction is recorded and analyzed, and as a result of the analysis the changes in the answer database are made.

A computer-implemented method for discovering network attack paths is provided. The method includes a computer generating scoring system results based on analysis of vulnerabilities of nodes in a network configuration. The method also includes the computer applying Bayesian probability to the scoring system results and selected qualitative risk attributes wherein output accounts for dependencies between vulnerabilities of the nodes. The method also includes the computer applying a weighted-average algorithm to the output yielding at least one ranking of nodes in order of likelihood of targeting by an external attacker.

The computational cost of many statistical modeling algorithms is affected by the input / output (I / O) cost of accessing out-of-core training data. This is an important challenge for emerging data mining applications, where the amount of training data can be potentially enormous. A heuristic approach to this problem is described. This approach is based on constructing a simple probability model from the large training data set, and using this model to generate simulated pseudo data for some aspects of the statistical modeling procedure. This approach is illustrated in the context of building a Naive Bayes probability model with feature selection. Here, the usual algorithms would require numerous data scans over the massive training data set, but our heuristic obtains models of comparable accuracy with just two data scans.

The invention provides an intelligent fault diagnosis method of a numerical control machine tool. The intelligent fault diagnosis method uses fault tree diagnosis as a base, faults are diagnosed by combining a Bayes probability and rule reasoning method, fault cause is found, and maintenance advices are proposed. The intelligent fault diagnosis method includes: firstly extracting fault information from historical maintenance records and machine tool operation instructions, and establishing a numerical control machine tool fault information knowledge base; searching fault trees belonging to a fault position in the fault information knowledge base according to fault characteristic information, calculating probability of occurrence of each fault tree under the current condition if a plurality of fault trees fit conditions, and determining sequence for diagnosing the fault trees according to size of the probability; and finally judging whether all sub-events of top events of the fault trees occur, returning to a fault treatment method corresponding to the events and serving as a solution if the sub-evens are bottom events of the fault trees, otherwise searching sub-events of the events continuously until finding all bottom events fit the fault characteristic information. The intelligent fault diagnosis method of the numerical control machine tool is rapid in diagnosis speed and accurate and reasonable in diagnosis result.

One disclosed embodiment is a method based on an iteratively employed Bayesian-probability-based pixel classification, used to refine an initial feature mask that specifies those pixels in a region of interest, including and surrounding a feature in the scanned image of a microarray, that together compose a pixel-based image of the feature within the region of interest. In a described embodiment, a feature mask is prepared using only the pixel-based intensity data for a region of interest, a putative position and size of the feature within the region of interest, and mathematical models of the probability distribution of background-pixel and feature-pixel signal noise and mathematical models of the probabilities of finding feature pixels and background pixels at various distances from the putative feature position. In a described embodiment, preparation of a feature mask allows a feature-extraction system to display feature sizes and locations to a user prior to undertaking the computationally intensive and time-consuming task of feature-signal extraction from the pixel-based intensity data obtained by scanning a microarray.

The present invention is uniquely designed to interact with web retailer's customers with real agent reaction times as they give astute answers directly concerning web retailer's products and goals. The entire process is manageable through a third-party website which includes scripting, settings and other parameters selected by the web retailer. An artificial intelligence engine uses the combination of Bayesian probability keyword selection, natural language parsing and regular expression processing. The technology updates its response database with every client interaction-learning always takes place as it simulates a live agent, in real time. Every client interaction is recorded and analyzed, and as a result of the analysis the changes in the answer database are made.

The invention provides a method, device and terminal for determining an identity identification of a human face in a human face image. The method includes the step of obtaining an original characteristic vector of the human face image, selecting k candidate vectors from a human face image database according to the original characteristic vector, choosing a matching vector of the original characteristic vector from the k candidate vectors, and determining an identity identification, recorded in the human face image database, of the matching vector as the identity identification of the human face in the human face image according to the matching vector of the original characteristic vector. According to the method, device and terminal, low-lever human face characteristic vectors and middle-level characteristic vectors formed by interacting with self-correlation and mutual-correlation submatrixes in a combined Bayesian probability matrix are stored in the human face image database, the middle-level characteristic vectors contain characteristic vectors of human faces and information interacting with the self-correlation and mutual-correlation submatrixes in the combined Bayesian probability matrix, and efficiency and accuracy of human face identification can be improved.

The invention relates to a fabric defect detection and classification method based on unsupervised segmentation and ELM (Extreme Learning Machines), and belongs to the fields of computer vision, pattern recognition and image application. Fabric defect images are input, and fabric defect image segmentation is realized to obtain segmented fabric defect images; shape and texture features of the segmented defect images and the original defect images are extracted to realize feature extraction of the fabric defect images; a feature set and labels of the to-be-classified fabric defect images are used as a training set to train an ELM classifier to obtain ELM classifier parameters; and fabric defect image classification is realized through the trained classifier according to Bayesian probability fusion, and a fabric defect image classification result is output. The detection and classification method provided by the invention has a higher accuracy rate.

The invention provides a real-time visual object tracking method having shielding sensing mechanism, tracking results based on filter tracking algorithm based on correlation and Bayesian probability model tracking algorithm based on color statistical information are fused, position prediction of a target is performed, one-dimensional scale pyramid is independently trained to perform target scale estimation, in addition, a shielding sensing mechanism is designed, the mechanism can judge whether a target is partly or completely shielded, then an update strategy of the model and which kind of model utilized to track new frame are determined, tracker drifting caused by pollution of background information is avoided, and according to the above steps, the tracking algorithm is used for tracking the target persistently and iteratively. The above algorithm can be used for accurately tracking the target in real time, and a plurality of problems including deformation, rotation, shielding, size variation, illumination variation, motion blur and the like can be effectively handled.

The invention discloses a scene text identification method based on a Bayesian probability frame, and relates to vision and mode identification of a computer. The scene text identification method based on the Bayesian probability frame comprises the steps that S1, scene image text is input; S2, character detection and recognition are carried out; S3, a detection-identification candidate grid is built, particularly, a candidate character area, a corresponding character class and a corresponding identification score are stored in the detection-identification candidate grid, each detection-identification route in the detection-identification candidate grid corresponds one text detection and identification result, and a route evaluation function is designed to evaluate each detection-identification route in the candidate grid; S4, the best detection-identification route is searched for from the candidate grid through a dynamic planning algorithm, that is to say, the identification result is obtained; S5, the text identification result is output. The scene text identification method based on the Bayesian probability frame achieves probability modeling and parameter learning of scene text identification integrating detection and identification.

A controller receives outputs form a plurality of sensors such as a camera, LIDAR sensor, RADAR sensor, and ultrasound sensor. Sensor outputs corresponding to an object are assigned to a tracklet. Subsequent outputs by any of the sensors corresponding to that object are also assigned to the tracklet. A trajectory of the object is calculated from the sensor outputs assigned to the tracklet, such as by means of Kalman filtering. For each sensor output assigned to the tracklet, a probability is updated, such as using a Bayesian probability update. When the probability meets a threshold condition, the object is determined to be present and an alert is generated or autonomous obstacle avoidance is performed with respect to an expected location of the object.

The invention provides an optimal protection policy analysis system and method based on an improved ant colony algorithm. The system comprises a network risk assessment subsystem, an attack path prediction subsystem and a security policy management subsystem. The method comprises the following steps: 1) importing a Bayesian probabilistic model to describe the complex relationship between multiple steps of atomic attacks, and constructing an attack diagram by using the Bayesian belief network to evaluate a network security risk; 2) simulating a decision making process of an attacker in combination with the quantitative indexes of the protection cost and the gain, and predicting a possible attack path; and 3) using the improved ant colony algorithm capable of calculating the quasi-optimal solution within a polynomial time to solve the security policy selection problem on the attack path. By adoption of the optimal protection policy analysis system and method, a risk solution can be provided for a network security administrator in time, the attack path prediction and protection strategy selection problems in network security protection are solved, and the security of the whole network under the condition of limited protection resources is effectively protected.

A speech recognition system comprises exactly two automated speech recognition (ASR) engines connected to receive the same inputs. Each engine produces a recognition output, a hypothesis. The system implements one of two (or both) methods for combining the output of the two engines. In one method, a confusion matrix statistically generated for each speech recognition engine is converted into an alternatives matrix in which every column is ordered by highest-to-lowest probability. A program loop is set up in which the recognition outputs of the speech recognition engines are cross-compared with the alternatives matrices. If the output from the first ASR engine matches an alternative, its output is adopted as the final output. If the vectors provided by the alternatives matrices are exhausted without finding a match, the output from the first speech recognition engine is adopted as the final output. In a second method, the confusion matrix for each ASR engine is converted into Bayesian probability matrix.

The invention relates to the field of social network and data mining, and provides a community discovery method capable of combining links to attribute information mainly comprises the following steps: inputting social network data with links and attribute information, and constructing an adjacent matrix which takes a link relation as a basis and an attribute matrix which takes the attribute information as a basis; constructing a combined Bayesian probability model according to two data matrixes; calculating the maximum attribution ratio of nodes by using a non-negative matrix decomposition method to obtain preliminary community division; and calculating absolute attribution ratio of the nodes according to attribution conditions of the nodes to obtain an overlapped community structure. Link information in the social network is combined to the attribute information, the using ratio of data in community detection is increased favorably, the accuracy and efficiency of community discoverycan be improved, and the community discovery method is suitable for being applied to discovering a topical community with attributes and link information.

The invention discloses a multi-document automatic abstract generation method based on phrase subject modeling. Multiple sample documents are subjected to word segmentation to obtain phrases and frequency of occurrence of the phrases, and the documents are expressed in the form of a phrase bag; joint probability distribution of the documents is calculated on the basis of an LDA subject model, the LDA subject model is converted into a phrase subject model, then a Gibbs sampling algorithm is used for estimating implicit parameters in the phrase subject model according to Bayesian probability, and lastly probability distribution of the subject in words is obtained; the tested documents are subjected to word segmentation, the subject weight and word frequency weight of obtained sentences are calculated and obtained, the final weight of the sentences is obtained by means of weighting calculation, and abstract content is generated according to the final weight. The method is more standard and precise, the relationship between different words is taken into consideration, the subject weight of the sentences is introduced, and the generation result better conforms to the practical essay abstract writing conditions of people after the subject weight of the sentences is introduced.

The invention discloses a method and a system for deducing a sudden event situation based on a case. The method comprises the steps of generating a case script of the case, abstracting the case script, extracting a case event link and storing the case event link in a server database; by a client, obtaining the case event link from the server database, analyzing the case event link, generating a logic relationship and specific information of the event in the case, generating a scalable vector diagram of the event based on the logic relationship and the specific information and adding the scalable vector diagram to a page of a webpage; and adjusting the scalable vector diagram based on the logic relationship, finishing the showing of the sudden event situation by combining with a visualization technology, and finishing deduction of the event automatically by adopting a Bayesian probability statistics principle and by combining with the internal logic probability of the event. According to the method and the system for deducing the sudden event situation based on the case, a visual showing and editing function of the event link is provided, and assistance is offered for non-specialized persons to edit and drill the script.

The invention relates to an industrial process fault detection method based on correction type independent element analysis and Bayes probability fusion. A conventional fault detection method based on the correction type independent element analysis requires selection of a non-quadratic function to measure non-gaussianity. Different industrial process data or objects can cause a fact that enough experiential knowledge for guiding the selection of the non-quadratic function is hard to acquire in an actual application. By aiming at different non-quadratic functions, the normal data training of the industrial process is adopted to acquire different correction-type independent element models. The Bayes probability fusion method is used to integrate decision results of a plurality of fault detection models together to acquire a final probability type monitoring index. Compared with the prior art, the industrial process fault detection method is used to solve a model uncertainty problem caused by non-quadratic function diversity, and a plurality of model possibilities are fully considered, and the reliability of the fault detection models is enhanced to a great extent.

The invention provides a coherent signal DOA (Direction-of-Arrival) estimation method based on sparse Bayesian learning. The method comprises the following steps that: obtaining the output signal of areceiving array, carrying out meshing on an observation space, constructing a super complete array manifold, combining with a sparse representation thought to convert a DOA estimation problem into aspare signal reconstruction problem, solving the following sparse matrix equations, establishing a spare Bayesian probability model, adopting a variation Bayesian deduction method to calculate the approximate posterior distribution of each hidden variable, and calculating the DOA estimation value of an incidence signal. By use of the estimation method, the number of incidence signals does not needto be estimated in advance, the estimation method does not relate to a decorrelation operation, the DOA estimation of a coherent signal is effectively realized, and the DOA estimation method which isput forward has good angle resolving power by aiming at the coherent signal. Compared with other existing DOA estimation methods, the method exhibits better direction-finding accuracy.

A method and system for truth discovery may implement a methodology that accounts for accuracy of sources and dependency between sources. The methodology may be based on Bayesian probability calculus for determining which data object values published by sources are likely to be true. The method may be recursive with respect to dependency, accuracy, and actual truth discovery for a plurality of sources.

A barcode reader and processing circuitry are disclosed for identifying a position of a cylinder rod, for example, along a path of movement. The barcode is marked on a surface of the rod and constitutes a series of markings extending along a portion of the length of the rod. The markings represent encoded information based on numbers calculated with a random number generator, and are grouped into subsets, each of which corresponds to a particular rod position. A portion of the rod including a subset of barcode markings is exposed to light emitted from the reader. Light reflected off the barcode is sensed with an array of photosensors provided in the reader. The photosensors generate electrical signals, and using Bayesian probability, a likely subset of markings is determined corresponding to the received electrical signals. Once identified, the markings subset is associated with a location corresponding to the position of the rod. A barcode reader is also described that is relatively easy to assemble and provides uniform illumination of the barcode.

The invention discloses a personalized information retrieval method and device. The method comprises generating retrieval results according to an assigned user's retrieval task; subjecting each query term in the user's retrieval task and each query word in each query term to Bayesian probability matrix decomposition to obtain query-document relevancy; acquiring the user's long-term behaviors and short-term behaviors, and acquiring user-document attention according to the user's long-term behaviors, short-term behaviors and behavior weighted parameters; generating the user's sorting weighted parameters, and determining relevancy of each document in the retrieval results according to the query-document relevancy, user-document attention and sorting weighted parameters; sorting the documents in the retrieval results according to the relevancies of the documents in the retrieval results, and returning to the user. The personalized information retrieval method and device have the advantages that the accuracy of a personalized information retrieval sorting model can be improved and retrieving may be independent of specific contents of query and documents.

The invention discloses a vehicle re-recognition method based on space-time constraint model optimization. The method comprises the following steps: 1) obtaining a to-be-queried vehicle image; 2) fora given vehicle query image and a plurality of candidate pictures, extracting vehicle attitude features through a vehicle attitude classifier and outputting a vehicle attitude category; 3) fusing thevehicle attitude feature and the fine-grained identity feature of the vehicle to obtain a fusion feature of the vehicle based on visual information, and obtaining a visual matching probability; 4) estimating the relative driving direction of the vehicle, and establishing a vehicle space-time transfer model; 5) obtaining a vehicle space-time matching probability; 6) based on the Bayesian probability model, combining the visual matching probability and the space-time matching probability of the vehicle to obtain a final vehicle matching joint probability; and 7) arranging the joint probabilitiesof the queried vehicle and all candidate vehicles in a descending order to obtain a vehicle re-recognition sorting table. The method provided by the invention greatly reduces the false recognition rate of vehicles and improves the accuracy of a final recognition result.

A barcode reader and processing circuitry are disclosed for identifying a position of a cylinder rod, for example, along a path of movement. The barcode is marked on a surface of the rod and constitutes a series of markings extending along a portion of the length of the rod. The markings represent encoded information based on numbers calculated with a random number generator, and are grouped into subsets, each of which corresponds to a particular rod position. A portion of the rod including a subset of barcode markings is exposed to light emitted from the reader. Light reflected off the barcode is sensed with an array of photosensors provided in the reader. The photosensors generate electrical signals, and using Bayesian probability, a likely subset of markings is determined corresponding to the received electrical signals. Once identified, the markings subset is associated with a location corresponding to the position of the rod. A barcode reader is also described that is relatively easy to assemble and provides uniform illumination of the barcode.

A system for processing audio data comprising a linear demixing system configured to receive a plurality of sub-band audio channels and to generate an audio output and a noise output. A spatial likelihood system coupled to the linear demixing system, the spatial likelihood system configured to receive the audio output and the noise output and to generate a spatial likelihood function. A sequential Gaussian mixture model system coupled to the spatial likelihood system, the sequential Gaussian mixture model system configured to generate a plurality of model parameters. A Bayesian probability estimator system configured to receive the plurality of model parameters and a speech / noise presence probability and to generate a noise power spectral density and spectral gains. A spectral filtering system configured to receive the spectral gains and to apply the spectral gains to noisy input mixtures.