47 results about "Probability matching" patented technology

Embodiments of the invention provide a system and method of automatically generating weights for matching data records. Each field of a record may be compared by an exact match and / or close matches and each comparison can result in a mathematical score which is the sum of the field comparisons. To sum up the field scores accurately, the automatic weight generation process comprises an iterative process. In one embodiment, initial weights are computed based upon unmatched-set probabilities and default discrepancy weights associated with attributes in the comparison algorithm. A bulk cross-match is performed across the records using the initial weights and a candidate matched set is computed for updating the discrepancy probabilities. New weights are computed based upon the unmatched probabilities and the updated discrepancy probabilities. Test for convergence between the new weights and the old weights. Repeat with the new weight table until the weights converge to their final value.

In embodiments of the present invention improved capabilities are described for identifying a first classification scheme associated with product attributes of a first grouping of products, identifying a second classification scheme associated with product attributes of a second grouping of products, and receiving a record of data relating to an item, the classification of which is uncertain. It may also involve receiving a dictionary of attributes associated with products and assigning the item to at least one of the classification schemes based on probabilistic matching among the attributes in the classification schemes, the attributes in the dictionary of attributes and the known attributes of the item.

A method of detecting, comparing, blocking, and eliminating spam emails sent through email servers of Internet service providers (ISPs) or to email users' email-boxes. The method includes the steps of generating a spam decipher signature for each email in an ISP's mail server or a user's email-box, comparing newly generated spam decipher signatures to a server or user database containing spam decipher signatures of known spam emails to detect spam emails when there is a probability match at a pre-determined high percentage, and preventing the spam emails from going through the ISP's mail server or to the email user as non-spam emails. The method also includes the steps of updating a master spam decipher signature database by comparing spam decipher signatures in a new signature database with existing spam decipher signatures in the master database, incrementing a counter value of a matching spam decipher signature by the number of matches, and adding all new spam decipher signatures that have counter values reaching or exceeding a pre-set threshold and therefore are considered spam to the master spam decipher signature database. The method further includes the steps of initially loading the master spam decipher signature database to the ISP email server or the user's computer to establish the server or user database, and updating the server or user database with the master spam decipher signature database.

The present invention relates to techniques for contextual filtering for improving an output of a speech recognizer. The techniques comprise receiving a representation of a speech utterance into a shallow parser component in the form of a first word lattice including a set of potential language matches for the utterance, where the language matches include at least one word; the parser component operative for receiving and analyzing the first word lattice to produce a second word lattice, and including a filter that assigns a probability match for a potential word match in relation to another potential word match in the utterance based on a particular filter-specific criteria; and outputting the probability matches from the shallow parser component as a portion of a second word lattice for further processing to determine most likely combinations of words present in the speech utterance.

In embodiments of the present invention improved capabilities are described for identifying a classification scheme associated with product attributes of a grouping of products of an entity, receiving a record of data relating to an item of a competitor to the entity, the classification of which is uncertain, receiving a dictionary of attributes associated with products, and assigning a product code to the item, based on probabilistic matching among the attributes in the classification scheme, the attributes in the dictionary of attributes and at least one known attribute of the item.

In embodiments of the present invention improved capabilities are described for identifying a classification scheme associated with product attributes of a grouping of products of an entity, receiving a record of data relating to an item of a competitor to the entity, the classification of which is uncertain, receiving a dictionary of attributes associated with products, and assigning a product code to the item, based on probabilistic matching among the attributes in the classification scheme, the attributes in the dictionary of attributes and at least one known attribute of the item.

A system and method for dialog state tracking employ an ontology in which a set of values are indexed by slot. A segment of a dialog is processed to detect mentions. Candidate slot values are extracted from the ontology, based on the detected mentions. The candidate slot values are ranked. A dialog state is updated, based on the ranking of the candidate slot values, which may be conditioned on the output of a temporal model, which predicts whether the value of the slot has been instantiated, modified, or is unchanged.

The present invention involves an enhanced method for operating a speech recognition system in which sequential vocabularies are loaded for comparison to the input speech. Form each sequential vocabulary a subset of candidate vocabulary elements are selected for matching, the probability match scores of each of the selected candidate vocabulary elements is weighted by a weighting factor that is dependent on the particular vocabulary from which the vocabulary elements are selected. As each set of candidate vocabulary elements is selected from the next sequential vocabulary, the selected set is combined with the set of previously selected and weighted vocabulary elements, which is then further weighted according to the weighting criteria of the then active vocabulary. At the end of the sequential vocabulary selection grouping process, the final, sequentially weighted match scores for the candidate vocabulary elements are arranged and an appropriate match to the input speech is presented.

The invention discloses a CSI-MIMO indoor positioning method based on feature value extraction. The method comprises: an offline stage: a wireless receiver (for example, the Intel wireless network card 5300 network card) obtains a reference point CSI matrix, and processes the CSI matrix through a feature value extraction method to establish an offline fingerprint database; and an online matching stage: feature value extraction is performed on the CSI matrix of a to-be-measured point, and estimated position information is obtained through a probability matching algorithm based on KNN. The CSI-MIMO indoor positioning system based on feature value extraction disclosed by the invention uses a WIFI wireless router as a signal transmitter, and uses a mobile terminal provided with a wireless receiver (for example, the Intel wireless network card 5300 network card) to receive electromagnetic signals. The fingerprint information is extracted via a feature value, thereby reducing the computational complexity and the correlation of information, and improving the positioning performance of the indoor position.

This twenty-step business process uses a clinical trials matching system to perform scalable searches of derived patient data records for patients that may be eligible for clinical trials. The patent-bearing business contracts with healthcare providers of patients found in derived patient data records to review the likely matches of those patients to clinical trials. Contracted healthcare providers also provide sufficient data for the patent-bearing business to determine if their patients qualify for clinical trials. Probabilistic matching techniques are used for this determination. The healthcare provider solicits qualified patients for clinical trials and interacts to enroll patients in trials. Solicitation is active in that the patent bearing business need not wait for a patient to initiate contact in order to seek determination of a patient's potential eligibility to participate in a clinical trial. Patient privacy is protected by limiting patient contact to be only through the patient's healthcare provider.

Techniques for remote authentication using reconfigurable boson samplers are provided. In one aspect, a method for remote authentication includes the steps of: providing an input photon configuration for an optical transmission network; receiving a response including measured output quantum photon coincidence frequencies from the optical transmission network based on the input photon configuration; comparing the measured output quantum photon coincidence frequencies to output quantum photon coincidence probabilities calculated for the optical transmission network; and verifying the response if the measured output quantum photon coincidence frequencies matches the output quantum photon coincidence probabilities calculated for the optical transmission network with less than a predetermined level of error, otherwise un-verifying the response. A verification system including an optical transmission network is also provided.

The invention relates to a multi-model self-adaptive fusion filtering method of a ship dynamic positioning system and belongs to the technical field of ship dynamic positioning. The method includes the steps: (1), building a ship three-degree-of-freedom low-frequency and high-frequency motion model, and acquiring a filter state formula and a measuring formula; (2), utilizing a differential global positioning system and a platform compass to measure position information and a heading angle, and collecting information in real time; (3), utilizing prior information and posterior information to initialize input of a model-based filter; (4), on the basis of a system model, utilizing a strong tracking filter and a Sage-Husa filter for parallel filtering; (5), subjecting the model to probability updating, and utilizing residual covariance output by the filters to calculate model probability matched with the model; (6), according to the model probability, acquiring fusion output of multi-model state estimation, namely ship position and heading information. The multi-model self-adaptive fusion filtering method has the advantages of strong robustness, high accuracy in Sage-Husa filter state estimation, stable system, high positioning accuracy and the like.

The invention provides a method of recognizing large-sized vehicles on an expressway based on GMM-HMM (gaussian mixture model and hidden Markov model) and relates to the technical field of intelligenttraffic. The method includes extracting MFCC (Mel frequency cepstral coefficient) features of an audio signal, analyzing a specific distribution of the features of the audio signal in a feature space, fitting a feature vector through Gaussian probability-density function to obtain an GMM-HMM, training the GMM in a training phase through EM (expectation maximization) algorithm to estimate model parameters, selecting a training sample via K-means algorithm, training GMM-HMM parameters through Baum-Welch algorithm in combination with an observation sequence probability distribution fitted via the Gaussian probability-density function in order to generate a training model, extracting MFCC feature parameters of audio data to be recognized in a recognition phase, and subjecting the MFCC featureparameters to feature probability matching with various models in a model base via Viterbi algorithm, so that the model having maximum matching probability is the recognition result. The method herein is suitable for accurately recognizing large-sized vehicles on the expressway.

The invention discloses an online single-shot thrust fault diagnosis method for a carrier rocket. The method comprises the following steps: S100, establishing a carrier rocket state equation based onthe attitude kinematics and kinetic equation of the carrier rocket, widening the band identification quantity into the state quantity, and establishing a corresponding number of system state equationsand measurement equations for fault diagnosis of the single-shot thrust of the carrier rocket according to the actual number of the single-shot thrust of the carrier rocket; S200, in the real-time flight process of the carrier rocket / missile, estimating the single-shot thrust of the carrier rocket by using unscented Kalman filtering, and obtaining an estimation error and a covariance matrix; andS300, performing probability matching calculation by using the error and the covariance matrix to realize thrust fault positioning and obtain the output thrust of a faulted engine. The method has theadvantages of being simple in structure, simple in design process, high in convergence speed and high in identification precision, and therefore the method has wide application prospects in carrier rocket single-shot thrust fault diagnosis.

The invention relates to target detection and tracking of digital images, providing a quick target detection and tracking method which is based on DSP and is easy to realize, as well as a digital image processing system for realizing the method. The method comprises the following steps: according to the Y component of an image, extracting an image edge to determine a target window; according to the Y component, the U component and the V component of the target window, extracting the characteristic value of the target window; and calculating the probability of the characteristic value of the target window in an initial frame and the probability of the characteristic value of the target window in a prediction frame, thus obtaining the pixel coordinate of the target window in the optimal prediction frame by matching the probability of the characteristic value of the target window in an initial frame with the probability of the characteristic value of the target window in a prediction frame. YUV space has favorable clustering characteristics; image information output by a DSP video acquisition module does not need format conversion; YUV space is adopted to carry out target tracking of Mean shift to ensure more accurate result and lower operation complexity.

The method involves analyzing a voice signal to detect voice information, converting a detection result with the highest probability back into a voice signal and outputting it to the user for verification and / or correction. A number of alternative recognition results is generated during the analysis with the next highest probability of a match to the detected voice signal. These can be used by the user to correct the output recognition result. AN Independent claim is also included for the following: (a) a voice recognition system (b) and a computer program product for implementing the inventive method.

The invention provides a vehicle integrated positioning system and method based on map matching. The system comprises a mobile phone communication module, an inertial navigation unit, a vehicle-mounted speedometer and a controller, wherein the mobile phone communication module is used for obtaining coarse position information of a vehicle, determining the range of a candidate map according to the position and extracting a corresponding road net map; the inertial navigation unit is used for acquiring a change in the course angle of the vehicle; the vehicle-mounted speedometer is used for obtaining a traveling distance of the vehicle; and the controller is used for matching a sensor input with the road net map, thereby outputting the position and orientation of the vehicle. The method is based on probability matching of the road net map, and the map is stored off line in a computing device or acquired on line; the result of the method is hardly influenced by the environment, and the environment does not need to be reconstructed; and the method can be used for replacing a GPS (global positioning system) to realize wide-range real-time vehicle positioning.

Some embodiments may be directed to matching users, such as employees, with insurance records in an integrated database, wherein the integrated database includes a plurality of insurance records, each record being associated with a unique employee and having a plurality of fields. According to some embodiments, new employee information may be received and it may be determined that the new employee information does not qualify as a strong match with any insurance record in the integrated database. Moreover, it may be determined, based on a probabilistic pattern match of the new employee information with values in the fields of the integrated database, that the new employee information qualifies as a weak match with a particular insurance record in the integrated database. Subsequent to the determination of a weak match, supplemental employee information may be received, and, responsive to the supplemental employee information, the match between the new employee information and the particular insurance record in the integrated database may be upgraded from a weak match to a strong match.

The embodiment of the invention provides a method and device for identity recognition, and relates to the technical field of televisions. The method and device are used for improving the accuracy of user identity recognition. The method comprises the steps that a voice signal of a user is received; a first parameter of the voice signal is extracted; first model matching is respectively carried out on the first parameter and a prestored first model parameter, and the maximum first probability matched with a first model is determined; second model matching is respectively carried out on the first parameter and a prestored second model parameter, and the maximum second probability matched with a second model is determined; the maximum first probability is compared with a first threshold value, and if the maximum first probability is not smaller than the first threshold value, the maximum second probability and the maximum first probability are operated; an operation result is compared with a second threshold value, and the user is determined as a user corresponding to the maximum second probability if the operation result is not smaller than the second threshold value.

The invention discloses a camera scene change detecting method based on the image angular point set characteristic. The camera scene change detecting method based on the image angular point set characteristic comprises the following steps that firstly, image angular point set information is obtained; secondly, an angular point set matching template is trained; thirdly, whether scene change occurs is judged. According to the camera scene change detecting method based on the image angular point set characteristic, the image angular point set information characteristic and the mathematical statistic probability template are combined to be used for judging whether scene change occurs within the monitoring range of a camera, and the image angular point set information is adopted ruing processing; compared with an existing method, only angular point information is adopted, the interference of pseudo angular points caused by accidental factors can be reduced, and the subsequent angular point matching is facilitated due to the fact that the angular point set information has the statistical probability stability; the overall angular point set probability matching template which is trained in advance is essentially manufactured according to statistical probability events in an angular point appearing area; in this way, the influence of the external accidental factors can be eliminated to a great extent, the method can be adapted to the practical engineering application environment, and the accurate detection effect can be achieved.

The invention relates to a novel photon probability mapping method based on terrace forming, which belongs to the technical field of mapping method. The method comprises the following steps: matchingaccording to channel conditions and generating an optimal Gaussian beam by a laser; performing probability matching by a probability matcher, and changing the shape of the Gaussian beam by a spatial phase array to obtain a novel beam formed in terrace; performing the 64QAM projection mapping on the data by the novel beam; concentrating and converging the 64 constellation point optical signals; performing pre-emphasis; amplifying the optical signals after passing through a channel, detecting by a photodetector, performing low-pass filter, orthogonalizing and normalizing, and performing data convergence and equalization; matching probabilistic solutions, removing step noise by wavelet transform analysis, performing constellation demapping; estimating channel conditions according to the received data, feeding back channel state information to the probability matcher, and performing real-time dynamic regulation on the probability matcher. The novel photon probability mapping method based on the terrace forming can effectively reduce the signal transmission power, reduces the bit error rate, improves the frequency band utilization, and lengthens the transmission distance.

The invention relates to the field of rainfall forecasting, in particular to a multi-mode integrated rainfall forecasting method and device. According to the method and the device, the forecasting performance of each forecasting mode is evaluated, and the forecasting effect of each mode at different precipitation levels and the traditional integrated forecasting effect are analyzed. Then, the integration method is improved, and a sliding window method is used. A prediction mode with an optimal precipitation prediction effect in the preset time period is screened out; the forecasting modes areintegrated to generate an integrated forecasting mode; calculating an integrated forecast by using probability matching set average optimization; compared with the difference of forecasting effects oftraditional integrated forecasting in each precipitation level, the rainfall values corresponding to the magnitudes of the modes better than the traditional integrated forecasting effect are converted into the integrated forecasting step by step, so that the phenomenon that the forecasting effect of the traditional integrated forecasting is poor in some magnitudes can be overcome, and the optimalintegrated forecasting integrating all single optimal forecasting modes is formed.

The invention discloses a surgical navigation method and system based on augmented reality assistance. The method comprises the following steps: (1) performing three-dimensional CT scanning on a patient; (2) performing segmentation and three-dimensional reconstruction on the bile duct to obtain a three-dimensional model of the bile duct, importing the three-dimensional model into a storage unit in a computer, extracting a center line of the three-dimensional model of the bile duct through a data processing unit, and recording three-dimensional coordinate information of each point on the center line; (3) selecting a characteristic marker with a proper size; (4) deploying an augmented reality aided navigation system; (5) displaying the virtual bile duct model by identifying the feature marker; and (6) performing perspective imaging on the patient by using X-rays, acquiring two-dimensional coordinates of a tip in each frame of image through a surgical instrument tip detection algorithm in the data processing unit, matching the tip coordinates with three-dimensional bile duct center line point set coordinates according to a 2D-3D registration algorithm, deducing optimal probability matching mapping, and determining the optimal probability matching mapping according to the optimal probability matching mapping. The specific position of the tip of the surgical instrument in the three-dimensional bile duct model is judged, and coordinates are recorded.

Systems and methods for adaptive assessment may utilize test deliver policies derived from a multi-armed bandit approach in combination with an item response theory model. A monotonic policy may involve defining a difficulty range according to which test items are selected for delivery during an assessment, where the difficulty range is updated following delivery of each test item based on test-taker performance. A multi-stage policy may implement several stages for test item delivery, with a different initial difficulty range for test item selection being defined for each stage and the difficulty ranges being updated based on test-taker performance within each respective stage. A probability matching policy may involve defining an item difficulty probability distribution according to which test items are selected for delivery to a test taker, where the probability distribution is initialized based on test taker skill level and updated based on test taker performance.

The invention discloses an RGB-D point cloud splicing method and a system based on 2D-3D weak feature 3D neighborhood probability matching. The method mainly comprises the following steps: calculatingweak feature points of an RGB image corresponding to a point cloud, calculating 2D-3D descriptors of the weak feature points, and performing feature matching based on the 2D-3D descriptors to obtaininitial matching point pairs; screening matching point pairs according to a 3D neighborhood probability method; and performing point cloud splicing and the like. The method has the advantages of highmatching precision, high speed, simple method, easy implementation and the like.