267 results about "Probability estimation" patented technology

A method for encoding a video frame or picture comprises dividing up the video frame or the picture in portions of a first type associated with a first scanning pattern, and portions of a second type associated with a second scanning pattern. Data corresponding to a predetermined of the portions of the video frame or picture is transformed into a two-dimensional transform coefficient array. Fhe transform coefficients are precoded in order to obtain transform data units. One of a first and a second set of context models is chosen depending on as to whether the predetermined portion is a portion of a type being associated with the first or the second scanning pattern. One context model of the chosen one of the first and the second set of context models is assigned to the transform data unit based on the scanning position assigned to the predetermined transform coefficient. Finally, the transform data unit or a sub-unit thereof is arithmetically encoded into a coded bit stream based on the probability estimation with which the assigned context model is associated.

A method for encoding a syntax element contained in a precoded video signal into a coded bit stream, the precoded video signal representing at least one video frame, the syntax element being associated with a predetermined portion of the video frame and being indicative of as to whether the predetermined portion of the video frame is precoded in a first or a second way into the precoded video signal is described. The method comprises investigating as to whether a neighboring portion of the video frame neighboring the predetermined portion is precoded in the first way or the second way, in order to obtain a binary value; assigning one of at least two context models to the predetermined portion of the video frame based on the binary value, wherein each context model is associated with a different probability estimation; and arithmetically encoding the syntax element into the coded bit stream based on the probability estimation with which the assigned context model is associated.

A method for encoding a syntax element contained in a precoded video signal into a coded bit stream, the precoded video signal representing at least one video frame, the syntax element being associated with a predetermined portion of the video frame and being indicative of as to whether the predetermined portion of the video frame is precoded in a first or a second way into the precoded video signal is described. The method comprises investigating as to whether a neighboring portion of the video frame neighboring the predetermined portion is precoded in the first way or the second way, in order to obtain a binary value; assigning one of at least two context models to the predetermined portion of the video frame based on the binary value, wherein each context model is associated with a different probability estimation; and arithmetically encoding the syntax element into the coded bit stream based on the probability estimation with which the assigned context model is associated.

A method of identifying and localizing objects belonging to one of three or more classes, includes deriving vectors, each being mapped to one of the objects, where each of the vectors is an element of an N-dimensional space. The method includes training an ensemble of binary classifiers with a CISS technique, using an ECOC technique. For each object corresponding to a class, the method includes calculating a probability that the associated vector belongs to a particular class, using an ECOC probability estimation technique. In another embodiment, increased detection accuracy is achieved by using images obtained with different contrast methods. A nonlinear dimensional reduction technique, Kernel PCA, was employed to extract features from the multi-contrast composite image. The Kernel PCA preprocessing shows improvements over traditional linear PCA preprocessing possibly due to its ability to capture high-order, nonlinear correlations in the high dimensional image space.

This invention concerns optimal ad selection for Web pages by selecting and updating an attribute set, obtaining and updating an ad-attribute profile, and optimally choosing the next ad. The present invention associates a set of attributes with each customer. The attributes reflect the customers' interests and they incorporate the characteristics that impact ad selection. Similarly, the present invention associates with each ad an ad-attribute profile in order to calculate a customer's estimated ad selection probability and measure the uncertainty in that estimate. An ad selection algorithm optimally selects which ad to show based on the click probability estimates and the uncertainties regarding these estimates.

A method (and structure) for developing a distribution function for the probability of winning a bid by a seller for a product or service, using the seller's own historical data for winning bids and lost bids, includes normalizing the data for winning bids and the data for lost bids and merging the normalized data into a single set of data.

Embodiments of a computer-implemented method of calculating estimates of a joint posterior probability of class membership given combinations of attribute values of a pair of attributes are disclosed. The calculating is performed on the basis of data representing a training set of a plurality of instances defined by attribute values for a plurality of attributes together with a class membership outcome. Embodiments of the method comprise calculating first and second estimates of a posterior probability of class membership given attribute values of a first and second attribute of the pair, respectively, and binning the first and second estimates into a respective plurality of first and second probability range bins. Instances of the training set are mapped to combinations of one of each of the first and second pluralities of probability range bins, and on the basis of the mapping, calculating estimates of a joint posterior probability of class membership.

A method of identifying and localizing objects belonging to one of three or more classes, includes deriving vectors, each being mapped to one of the objects, where each of the vectors is an element of an N-dimensional space. The method includes training an ensemble of binary classifiers with a CISS technique, using an ECOC technique. For each object corresponding to a class, the method includes calculating a probability that the associated vector belongs to a particular class, using an ECOC probability estimation technique. In another embodiment, increased detection accuracy is achieved by using images obtained with different contrast methods. A nonlinear dimensional reduction technique, Kernel PCA, was employed to extract features from the multi-contrast composite image. The Kernel PCA preprocessing shows improvements over traditional linear PCA preprocessing possibly due to its ability to capture high-order, nonlinear correlations in the high dimensional image space.

A quantum key distribution method according to the present invention includes an error probability estimation step of estimating error probabilities of transmission data and the received data, an error correcting step of correcting errors in the received data based on error correcting information, a matching determination step of determining whether the transmission data and the received data after correcting errors match, and an information amount estimating step of estimating an amount of information leaked to an adversary through a quantum communication path, and further compresses data based on the amount of information made public in a process of processing via a public communication path and an estimated value of the amount of information leaked to the adversary through the quantum communication path to make the data after compression a cryptographic key shaped by devices.

OCR errors are identified and corrected through learning. An error probability estimator is trained using ground truths to learn error probability estimation. Multiple OCR engines process a text image, and convert it into texts. The error probability estimator compares the outcomes of the multiple OCR engines for mismatches, and determines an error probability for each of the mismatches. If the error probability of a mismatch exceeds an error probability threshold, a suspect is generated and grouped together with similar suspects in a cluster. A question for the cluster is generated and rendered to a human operator for answering. The answer from the human operator is then applied to all suspects in the cluster to correct OCR errors in the resulting text. The answer is also used to further train the error probability estimator.

Methods, apparatus, and computer-readable storage media for dynamic color flow modeling. A dynamic color flow model that incorporates motion estimation in a probabilistic fashion is described. By automatically and adaptively changing model parameters based on the inferred local motion uncertainty, the dynamic color flow model accurately and reliably models the object appearance, and improves the foreground color probability estimation when compared to conventional methods. The dynamic color flow model may, for example, be applied to both foreground and background layers (or to additional layers) for video object segmentation, obtaining significantly improved results when compared to conventional methods.

The invention relates to a method for predicting dynamic risk and vulnerability at fine scale and belongs to the scientific field of global information. The method is mainly characterized in that an optimized Bayesian network is searched from multi-source heterogeneous spatiotemporal data on the basis of a grid format with certain resolution at fine scale; domain knowledge is combined to improve the network; therefore, the uncertain estimation of disaster risk and the vulnerability, namely probability estimation, is carried out. In the method, a nuclear density method is put forward to train a sample according to a sample derivative grid; an optimized discretization method is put forward to discretize continuous variables so as to provide discrete state space input for the network; a simulated annealing optimization algorithm is adopted to search an optimized network structure; and a method of accurate reasoning combined with approximate reasoning to predict the probabilities of risk and the vulnerability is adopted. The method provided by the invention can position the positions of the disaster risk and the vulnerability in real time at the fine spatial scale, estimate the spatial distribution of the risk probability and has important theoretical significance and practical value for improving the effects on the reduction and relief of disaster and building an intelligent public emergency pre-warning system by the state.

This invention provides a novel single-pass and multi-pass synchronized encoder and decoder, performing order(s) of magnitude faster data compression and decompression, at any compression ratio with the higher or the same perceived and measured decompressed image quality in comparison with the best state-of-the-art compression methods, using order(s) of magnitude less system resources (processor complexity, memory size, consumed power, bus bandwith, data latency). These features are achieved using novel direct and inverse non-stationary filters for the recusive octave direct and inverse subband transformation, novel simple context modeling and symbol probability estimation using a minimum number of histograms with the fast adaptation for the sign and the magnitude of the transformation coefficients, a novel accelerated range coder without division operations, and a novel synchronisation of the compressed data.

A multi-user turbo decoder combining multi-user detection and forward error correction decoding is disclosed that utilizes iterative decoding of received, interfering signals, and the construction of a decoding tree of the decoder is changed for each iteration of the decoding based on the previous conditional probability estimates of the value of the data bits of each signal making up the received, interfering signals. Before each iteration of multi-user decoding, a probability estimate is calculated that the value of the bit in a signal has a certain value for all of the data bits. Using the probability estimate a new decoding tree is constructed before each iteration of decoding such that the signal bit having the most reliable estimate is assigned to the lowest or root level of the tree. Using the probability estimate for the other signal bits, the signal bit having the next most reliable estimate is assigned to the second level of the tree, and so forth, with the signal bit having the least reliable estimate being assigned to the highest level of the tree adjacent the terminating nodes or leaves of the tree. By building the decoding tree in this manner for each iteration of symbol decoding, a reduced complexity search is more likely to include paths (and nodes) in the tree containing the correct value for the channel symbols.

The invention discloses a wireless sensor network energy conservation method based on redundancy controlling and clustering routing. The wireless sensor network energy conservation method combines a probability estimation based redundant node control algorithm with a node energy consumption balancing clustering routing protocol which considers multiple factors to realize energy conservation, wherein the probability estimation based redundant node control algorithm includes a method for defining and finding a redundant node and sleep of the redundant node, and the node energy consumption balancing clustering routing protocol considering multiple factors includes LEACH (low energy adaptive clustering hierarchy) based network clustering of introduced clustered factors, redundant data fusion, cluster data forwarding with distance and angle comprehensively considered, and confirmation of cluster rotating period. On the premise of meeting network connectivity degree and cover degree, the wireless sensor network energy conservation method based on redundancy controlling and clustering routing balances network load and improves energy efficiency of nodes to the greatest extent, so that the purpose for prolonging network service life is achieved.

The invention discloses an ultrasonic image automatic understanding and generation method, and relates to the technical field of computer information processing. The method comprises the steps of extracting an ultrasonic image eigenvector by a CNN model, performing word segmentation on a descriptive sentence corresponding to an ultrasonic image by a word segmentation tool, and obtaining a word sequence eigenvector; training an LSTM model by utilizing the ultrasonic image eigenvector and the corresponding word sequence eigenvector; and generating a plurality of proper sentences at a time step by the LSTM language model through utilizing a beam search method, and generating the sentence for describing the ultrasonic image according to probability estimation. The method can well solve the problem that adjustment cannot be carried out in time when a new combination situation occurs in conventional image topic description and automatic generation. The method can be widely used for understanding and statement description of the ultrasonic image.

The present invention discloses a Chinese homonym error auto-proofreading method. The method comprises: first, generating a confusion set of Chinese homonyms; through large-scale Web corpus training, collecting statistics on a left-adjacent binary model, a right-adjacent binary model, and an adjacent ternary model; obtaining a local adjacent NGram model by using the confusion set of Chinese homonyms and a probability estimation algorithm; by using a weighted combination method and by calculating a sentence context support degree of a word in a sentence and a sentence context support degree of a homonym in a homonym confusion set corresponding to the word, determining whether a homonym error exists; marking the homonym error and providing a correction suggestion list, so as to implement Chinese homonym auto-proofreading. The Chinese homonym error auto-proofreading method provided by the present invention is quick in system response and high in efficiency and accuracy, and meets precision requirements of actual applications.

A code volume control section refers to an input image and a reference image stored in a frame memory and thereby sets a quantization step for each block so that the volume of coded data generated for a frame will be a preset volume. A data loss probability estimation section estimates a data loss probability (the probability that data loss will occur to a target block due to transmission error) based on a code volume predicted value of the target block obtained by the code volume control section and the code volume from the latest synchronization code pattern to a block just before the target block. A degradation estimation calculation section calculates an estimate of degradation of the target block caused by errors based on image degradation power and the data loss probability. A mode selection section selects an optimum encoding mode for the target block based on the degradation estimate of the target block and an estimate of frame coding distortion. Forced refresh (intra-frame encoding of a block) is carried out properly and effectively, thereby image degradation is reduced without deteriorating coding efficiency.

Methods for generating more accurate location estimates are based on indicative factors. The factors are gleaned from wireless signals that were received by a wireless terminal and from other network-specific data. Each factor is probative of whether the wireless signals were received by the wireless terminal: (i) indoors versus outdoors, and / or (ii) “above-the-clutter” versus “below-the-clutter” of ambient wireless signals, and / or (iii) while moving at a certain speed classification, e.g., stationary / pedestrian versus vehicular speeds. Each factor tends to prove or disprove the particular characteristic. The result is an estimated probability, such as the probability that the wireless terminal received the signals indoors. The estimated probability is applied to the analysis of candidate locations, eliminating some candidates from further consideration. This enables computational resources to focus on the remaining higher-likelihood candidate locations and provides improved accuracy in the location estimate.

An in-line compensation method for the match error of multiple sensors integrated tracking system features that the maximal probability estimation method is used to determine the initial error, extracting the system error by use of the information measured by sensors, and the filtered tracking information, and performing in-line compensation.

The invention discloses a method and a device for providing taxi service. The method for providing taxi service comprises the following steps: acquiring driver information of a plurality of service taxies according to a current taxi request of a user, wherein the current taxi request comprises order information; estimating the order taking probability of each service taxi according to the order information and the driver information of each service taxi by using a pre-trained order taking prediction model. By adopting the method for providing taxi service, in the order taking probability estimation process, the order taking probability of each service taxi can be accurately estimated with the combination of the order information and the driver information.

The invention relates to a distributed cooperative caching method capable of realizing node and message state combined perception and belongs to the technical field of a distributed cooperative caching technology for opportunistic networks. The method aims to solve the problem of the lower node caching efficiency at present, dissemination states of messages are perceived, weight values of the messages are dynamically estimated, and the encounter probability of nodes and message target nodes is predicted with an encounter probability estimation method; then the messages are cached in a classified manner on the basis of differences between message source nodes, different caching priorities are given, and the messages in node caches are replaced in a cooperative manner with a cooperative partition cache replacement mechanism, so that the caching efficiency of the nodes is improved; finally, in order to solve the problem of decrease of the message delivery rate caused by unchecked deletion of messages in a traditional cache management mechanism, a distributed cooperative cache transfer mechanism is adopted, top-k cooperative node sets of the nodes are dynamically selected in advance, and the messages in the node caches are transferred to cooperative nodes in a communication range when the node caches are full, so that the delivery probability of the messages is increased.

Various methods and apparatus are described for a speech recognition system. In an embodiment, the statistical language model (SLM) provides probability estimates of how linguistically likely a sequence of linguistic items are to occur in that sequence based on an amount of times the sequence of linguistic items occurs in text and phrases in general use. The speech recognition decoder module requests a correction module for one or more corrected probability estimates P′(z|xy) of how likely a linguistic item z follows a given sequence of linguistic items x followed by y, where (x, y, and z) are three variable linguistic items supplied from the decoder module. The correction module is trained to linguistics of a specific domain, and is located in between the decoder module and the SLM in order to adapt the probability estimates supplied by the SLM to the specific domain when those probability estimates from the SLM significantly disagree with the linguistic probabilities in that domain.

The invention discloses a phonetic gain factor estimation device and method, which comprises the following parts: prior signal-to-noise ratio estimation unit, pure phonetic gain calculation unit, conditional phonetic existence probability estimation unit and gain factor calculation unit. The method comprises the following steps: calculating the prior signal-to-noise ratio of input signal; calculating the pure phonetic gain value according to the prior signal-to-noise ratio of input signal; calculating the existence phonetic probability of input signal; calculating the phonetic gain factor of input signal in connection with the self-keep noise signal reduction factor. The invention realizes the estimation of phonetic gain factor without integral and differential calculation, which decreases the demand of memory capacity and calculation quantity in the integer DSP chip operation.

The invention relates to a text classification method and a text classification device. The method comprises the following steps: an NLP (Natural Language Processing) pre-processing step, wherein analysis of a natural-language processing method is carried out on user dialogue text to obtain a word set and semantic labeling results about the user dialogue text; a multi-dimensional-feature selectionstep, wherein combination is carried out for the word set and the semantic labeling results according to a plurality of rules to obtain a vectorized characterization form of semantic information contained by the user dialogue text; and a classification step, wherein probability estimation values are calculated for user dialogue classes obtained by the multi-dimensional-feature selection step. According to the text classification method and the text classification system of the invention, the advantages of counting and a deep-learning method can be integrated, and a customer demand-oriented text classification solution can be realized through multi-dimensional-feature selection.

The entropy coding of a current part of a predetermined entropy slice is based on, not only, the respective probability estimations of the predetermined entropy slice as adapted using the previously coded part of the predetermined entropy slice, but also probability estimations as used in the entropy coding of a spatially neighboring, in entropy slice order preceding entropy slice at a neighboring part thereof. Thereby, the probability estimations used in entropy coding are adapted to the actual symbol statistics more closely, thereby lowering the coding efficiency decrease normally caused by lower-delay concepts. Temporal interrelationships are exploited additionally or alternatively.

The invention discloses a joint active user detection and channel estimation method of a dispatching-free transmission system. No time-frequency transmission resource needs to be requested via a physical layer control channel during uplink transmission, but signals are directly sent according to a synchronous time slot in a pre-allocated conflict domain. Considering scenes in which an initial active probability of a user is known and unknown, the joint active user detection and channel estimation method adopts three modules, namely, a channel estimation module, an active user detection moduleand an initial active probability estimation module, and a base station side obtains an active user set and channel response vector estimated values of all users in the set according to a received signal Y and a pilot matrix S via the iterative calculation of the three modules. Compared with the traditional block sparse signal reconstruction algorithm, the joint active user detection and channel estimation method disclosed by the invention has the advantages of significantly reducing the complexity of the algorithm and improving the active user leak detection performance and the channel estimation performance.

A method and system for authenticating IP source addresses by accessing one or more HTTP requests whose source client identifies itself as a legitimate web crawler. One or more IP addresses are detected from the one or more HTTP requests and each detected IP address is authenticated via a probability estimation regarding its association with a legitimate web crawler. A lookup table is preferably compiled for the authenticated IP addresses for reference, publication and authentication purposes.

The invention discloses an unsupervised model parameter migration rolling bearing life prediction method, and belongs to the technical field of rolling bearing state identification and residual life prediction. The method is provided for solving the problems that in practice, rolling bearing labeled vibration data under a certain working condition is difficult to obtain, health indexes are difficult to construct, and the service life prediction error is large. The method comprises the steps that firstly, extracting root-mean-square features from rolling bearing full-life-cycle vibration data,and introducing a new bottom-to-top time sequence segmentation algorithm to segment a feature sequence into three states composed of a normal period, a degradation period and a recession period; marking the state information of an amplitude sequence of the vibration signal subjected to fast Fourier transform, taking the amplitude sequence as an input of an improved full convolutional neural network, extracting deep features, and constructing a source-domain model and a state recognition model subjected to fine adjustment through training to realize rolling bearing multi-state recognition; andestablishing a rolling bearing life prediction model by using a state probability estimation method. Experiments prove that the method provided by the invention does not need to construct health indexes, can realize rolling bearing state identification and life prediction under different working conditions under an unsupervised condition, and obtains a better effect.