136 results about "A priori probability" patented technology

To appropriately express an erasure position of a code by a small-scale, simple-structured circuit, a soft-output decoding circuit (90) in each element decoder includes a received value and a priori probability information selection circuit (154) to select an input to-be-decoded received value TSR and extrinsic information or interleaved data TEXT, whichever is necessary for soft-output decoding. Based on inner erasure position information IERS supplied from an inner erasure information generating circuit (152), the received value and a priori probability information selection circuit (154) replaces a position where no coded output exists due to puncture or the like with a symbol whose likelihood is “0”. That is, the received value and a priori probability information selection circuit (154) outputs information which assures a probability in which a bit corresponding to a position where there is no coded output is “0” or “1” to be “½”.

The invention relates to a webpage training method and system and a webpage prediction method and system. The webpage training method comprises obtaining a prior probability table of classified keywords according to existing data which are associated with the classified keywords; preprocessing a webpage to be trained to obtain a webpage text to be trained; extracting features in the webpage text to be trained according to the prior probability table to obtain an association relation feature vector representation F1 between the webpage to be trained and a specified category; performing model training on the association relation feature vector representation F1 to obtain a classification result of the webpage to be trained. According to the webpage training method and system and the webpage prediction method and system, category systems which are strong in heterogeneity can be simultaneously processed, the large category systems can be processed through less training data, and the problem of data sparseness is largely solved due to the fact that the browsing and search behavior of a user on the whole Internet not just a website is collected.

Techniques are disclosed for analyzing and learning behavior in an acquired stream of video frames. In one embodiment, a trajectory analyzer clusters trajectories of objects depicted in video frames and builds a trajectory model including the trajectory clusters, a prior probability of assigning a trajectory to each cluster, and an intra-cluster probability distribution indicating the probability that a trajectory mapping to each cluster is least various distances away from the cluster. Given a new trajectory, a score indicating how unusual the trajectory is may be computed based on the product of the probability of the trajectory mapping to a particular cluster and the intra-cluster probability of the trajectory being a computed distance from the cluster. The distance used to match the trajectory to the cluster and determine intra-cluster probability is computed using a parallel Needleman-Wunsch algorithm, with cells in antidiagonals of a matrix and connected sub-matrices being computed in parallel.

A likelihood indicating the distribution of the frequency of use of each specification of the existing device is calculated for each version of a software component used in the control software of the existing device, and a prior probability indicating the distribution of the frequency of use of each version is calculated for each software component used in the control software of the existing device. A posterior probability indicating the reusability of each version of the existing software component is calculated for each specification of a device to be developed, on the basis of the likelihood and the prior probability.

A computer implemented method and system is provided for classifying a document. A classifier is trained using training documents. A list of first words is obtained from the training documents. A prior probability is determined for each class of multiple classes. Conditional probabilities are calculated for the first words for each class. Confidence thresholds are determined. Confidence grades are defined for the classes using the confidence thresholds. A list of second words is obtained from the document. Conditional probabilities for the list of second words are determined from the calculated conditional probabilities for the list of first words. A posterior probability is calculated for each of the classes and compared with the determined confidence thresholds. Each class is assigned to one of the defined confidence grades based on the comparison. The document is assigned to one of the classes based on the posterior probability and the assigned confidence grades.

The invention relates to a decision making technique for the events, in order to solve the problem of decision-making simply and rapidly under the condition of a large amount of knowledge. The invention, for each event, can group each condition and become one group for the condition with identical prior probability and identical ' and , or ' relation to form one condition group, the corresponding relation table can be established according to the mapping relation between events, conditions and condition group, then the knowledge base for decision making is established based on the information in the relation table. A decision making application system for the events, which includes a server providing with the knowledge base and a computer connected with the server, can be established based on the knowledge base; the computer may receive one or more conditions inputted by user, the server, for every condition received, can carry out special ergodic process in the knowledge base, and look up out all condition groups with the condition, and the relative decision probability for one or more events can be figured out based on the information such as maximum probability or value of condition group, the satisfied condition number.

A decoder that receives, as input, probability information AMP / CR×yt. This probability information is obtained by dividing a channel value obtained by multiplication of received value yt and a predetermined coefficient AMP by the first additive coefficient CR for regulating the amplitude of the received value yt and the probability information 1 / CA×APPt obtained by multiplying the a priori probability information APPt by the reciprocal of the second additive coefficient CA for regulating the amplitude of the a priori probability information APPt to a soft-output decoding circuit. The soft-output decoding circuit, which may be a large scale intergrated circuit, generates log soft-output CI×Iλt and / or external information 1 / CA×EXt using additive coefficients for regulating the amplitude of arithmetic operations in the inside of the soft-output decoding circuit.

Methods and systems for determining a sequence of energy interactions in a detector. A plurality of discrete energy interactions is received in a plurality of detector voxels. A plurality of possible sequences of interaction is formed based on the received plurality of discrete energy interactions. For each of the plurality of possible sequences of interaction, an a posteriori probability is computed, where the a posteriori probability is based on a likelihood that the possible sequence of interaction is consistent with the received plurality of discrete energy interactions. Additionally or alternatively, the a posteriori probability may be based on an a priori probability. One of the formed plurality of possible sequences of interaction is selected based on the computed a posteriori probability.

A probabilistic data signal processor used to determine health of a system is described. Initial probability distribution functions are input to a dynamic state-space model, which iteratively operates on probability distribution functions, such as state and model probability distribution functions, to generate a prior probability distribution function, which is input to a probabilistic updater. The probabilistic updater integrates sensor data with the prior to generate a posterior probability distribution function passed to a probabilistic sampler, which estimates one or more parameters using the posterior, which is output or re-sampled and used as an input to the dynamic state-space model in the iterative algorithm. In various embodiments, the probabilistic data signal processor is used to filter output from any mechanical device using appropriate physical models, which optionally include chemical, electrical, optical, mechanical, or fluid based models. Examples to valve bearings and pipe systems are provided.

The invention relates to a multiple transform domain based super-resolution reconstruction method and belongs to the technical field of image super-resolution. The method comprises the following steps: performing nonrelated transform domain analysis on K input low-resolution images separately; according to the sizes of a target super-resolution image and the input low-resolution images, determining a down-sampling matrix D; and reconstructing the super-resolution image by adopting an optimization algorithm for L1-norm minimization in multiple transform domains through the constraint of a cost function min lambda1||W1(z1)||1+lambda2||W2(z2)||2+......+lambdaj||Wj(zj)||1 by z=z1+z2+...+zj and ||DHiMiz-gi||2<=ni (i=1,2,...K). Compared with the prior art, the method solves the problem in selection of a priori probability model in super-resolution reconstruction, namely, multiple transform domain based remote sensing image sparsity expression; and meanwhile, the method solves the problems of irregular geometric distortion and dislocation of low-resolution remote sensing images caused by load platform shake and atmospheric disturbance, and reconstructs the super-resolution image with a good effect.

An apparatus and a method for computing a Speech Absence Probability (SAP), and an apparatus and a method for removing noise by using the SAP computing device and method are provided. The provided SAP computing device for computing the SAP indicating probability that speech is absent in a mth frame, from a first through Ncth posteriori (Nc means the total number of channels) Signal to Noise Ratios (SNR) calculated with regard to the mth frame of a speech signal and a first through Ncth predicted SNRs predicted with regard to the mth frame, includes: a first through Ncth likelihood ratio generators for generating a first through Ncth likelihood ratios from the first through Ncth posterior SNRs and the first through Ncth predicted SNRs, and outputting them; a first multiplying unit for multiplying the first through Ncth likelihood ratios by a predetermined a priori probability, and outputting the multiplication results; an adding unit for adding each of the multiplication results received from the first multiplying unit to a predetermined value, and outputting the added results; a second multiplying unit for multiplying the added results received from the adding unit and outputting the multiplication result; and a inverse number calculator for calculating inverse number of the multiplication result received from the second multiplying unit and outputting the calculated inverse number as the SAP. Therefore, since the accuracy of the calculated SAP is high, noise can be efficiently removed from the speech signal that may have noise and an enhanced speech signal with an enhanced quality can be provided.

Connected components of dark pixels are clustered from across the image. A “most likely” representative image for each cluster of images is determined, with likelihood determined by a probabilistic model of the image capturing process. An a priori (prior) probability distributions on bitmaps may be used to determine the most likely representative images. For example, a priori probability distributions based on so-called chain codes are implemented. The representative images are used to cluster connected components. Clustering may be repeated. The output page is assembled by replacing each member of a cluster of images by that cluster's representative image.

The invention provides a method for extracting phrases of statistical machine translation, which comprises the following steps of: 1) acquiring a plurality of aligned sentence pair combinations from a bilingual language material from two directions, and calculating the priori probability of the plurality of aligned sentence pair combinations; 2) calculating the alignment probability of word pairsaccording to the sum of the priori probabilities of the word pairs of the plurality of aligned sentence pair combinations, and forming an alignment matrix by using the alignment probability of the word pairs; 3) calculating the frequency of phrase alignment according to the alignment matrix; and 4) calculating the relative frequency and the lexicalization probability of the phrase alignment according to the frequency of the phrase alignment. The method can effectively express all probable aligned phrase combinations, and improves the quality of phrase extraction, thereby being capable of improving the quality of translation which is performed according to the extracted phrases.

The invention discloses an image recognition method and a device, which relate to the technical field of detection and recognition. In order to solve the problem in the prior art that the accuracy ofimage recognition decreases when the distribution of the training set and the test set differs greatly, the image recognition method includes: acquiring a sample image (101); determining a priori probability distribution (102) of each category image occurring in a current scene; and recognizing the sample image according to a priori probability distribution to obtain a recognition result (103). The method is applied in the process of image recognition.

The invention provides a smoothing constraint unscented Kalman filtering method. The smoothing constraint unscented Kalman filtering method comprises the following steps: 1, acquiring an original prior probability density function of a target state at a current target observation moment according to unscented transformation; 2, calculating a mean value and a variance of an original prior probability through numerical expectation; 3, introducing noise constraint information, and calculating the center of the approximate feasible region to obtain corrected prior probability density; 4, seeking aGaussian distribution mean value and a variance meeting the constraint condition through posteriori iterative optimization, and generating a new Gaussian sigma point meeting the constraint condition;and 5, carrying out weighted calculation on the Gaussian sigma points to complete a filtering process. The smoothing constraint unscented Kalman filtering method has the advantages in the aspects ofaccuracy and robustness, and meanwhile, the real-time performance of the smoothing constraint unscented Kalman filtering method is superior to that of a particle filtering algorithm. Correspondingly,the invention further provides a target tracking method.

The invention discloses an image registration method based on a convolution neural network. The method includes the steps: using a VGG-16 convolution network to extract feature points from that reference image and the move image respectively, thereby generating a reference feature point set and a moving feature point set; When the distance matrix of the feature points simultaneously satisfies thefirst and second constraint conditions, performing a pre-matching operation, that is, a feature point x in the reference feature point set and a feature point y in the moving feature point set are matching points; setting a certain threshold, combining with iteration to select the dynamic interior points of the pre-matched feature points, screening out the final feature points, and obtaining a priori probability matrix; according to a priori probability matrix and EM algorithm, searching for the optimal parameters and realizing image registration. The invention dynamically increases the interior point step by step through the dynamic interior point selection when the characteristic point is matched, and improves the registration accuracy rate.

The invention is a method applicable to oil and gas exploration and exploitation for automatically detecting geological objects belonging to a given type of geological object in a seismic image, on a basis of a priori probabilities of belonging to a type of geological object assigned to each of samples of the image to be interpreted. The image is transformed into seismic attributes applied beforehand, followed by a classification method. For each of the classes, an a posteriori probability of belonging to a type of geological object is determined for each of the samples of the class according to the a priori probabilities, of the class, of belonging, and according to a parameter α describing a confidence in the a priori probabilities of belonging. Based on the class of the sample, the determined a posteriori probability of belonging to a type of geological object is assigned for the samples of the class. The geological objects belonging to the type of geological object are detected based on determined of the a posteriori probabilities of belonging to the type of geological object for each of the samples of the image to be interpreted.

The invention discloses an MEC random task migration method based on a Bayesian network. The method comprises the following steps of converting an application into a directed graph containing a plurality of sub-tasks; using a probability calculation method of a sub-node in the Bayesian network to calculate a priori probability of a current sub-task migration decision; generating a set of scheduling strategies to minimize the energy consumption of mobile devices according to the probability; using the weak exhaustion algorithm to adjust the generated scheduling strategy so as to obtain the optimal computing task migration strategy. The technical proposal of the invention solves the problem of random task scheduling under the MEC scene.

The invention discloses a defect recognition method based on a power equipment defect recognition learning model. The method is based on multi-layer label labeling and identification; According to thenovel learning model integrating feature extraction, boundary frame regression and classifier technology set, a priori probability is replaced by a clustering algorithm in the model, a training set is generated by training power equipment samples, the type and the defect of the power equipment can be detected in real time, the accuracy and the high efficiency of defect recognition are ensured, and the safe operation of the power equipment is ensured. The automation degree of power equipment state monitoring and defect recognition is improved, a power grid big data analysis and processing system is researched and developed, the power equipment defect recognition technical concept is innovated, and technical development in the field of power systems is promoted. Advanced technologies are applied to the field of power systems, manpower resources and time management cost are saved, and field development is promoted.

The invention discloses a posterior probability distribution calculation method and a system for the hill fire disaster fault of a power network. The method comprises the following steps: the hill fire tripping data of a transmission line under different hill fire disaster conditions are counted; the quantitative relationship between the key factors of transmission line fault and hill fire tripping is analyzed; fault probability distribution functions of transmission lines related to multiple fault probability distribution parameters are constructed; a priori probability distribution of each fault probability distribution parameter is obtained, and random values are extracted from the priori probability distribution repeatedly and independently for verification to train the model; and iterate to the model stationary; the posterior probability distribution of hill fire disaster is obtained by taking several groups of stochastic values as samples after the model is stationary. The invention can quickly obtain the posterior distribution of the hill fire fault probability of the transmission line, which plays an important role in guiding the risk analysis of large-scale mountain fire disaster.

A method for recovering a sparse signal of a finite field may include: updating discrete probability information of a target signal element of the finite field and discrete probability information of a measurement signal element of the finite field by exchanging the discrete probability information of the target signal element with the discrete probability information of the measurement signal element a predetermined number of times, wherein the target signal element and the measurement signal element are related to each other; calculating a final posteriori probability based on a priori probability of the target signal element and the discrete probability information of the measurement signal element, acquired as the exchange result; and recovering the target signal by performing maximum posteriori estimation to maximize the final posteriori probability.

The embodiment of the invention provides a vegetation coverage estimation method and device for surface heterogeneity area and an apparatus. The method comprises: acquiring a priori probability of vegetation coverage and a likelihood probability of vegetation coverage at a predicted time based on a low spatial resolution vegetation coverage product corresponding to each time point of a preset timeseries and high spatial resolution remote sensing data; based on the a priori probability of vegetation coverage and the likelihood probability of vegetation coverage, using a dynamic Bayesian network to obtain a high spatial resolution vegetation coverage estimation result at the prediction time. The embodiment of the invention overcomes the problem of low precision of the estimation result whenthe real surface condition is complex and heterogeneous, can obtain accurate estimation result, and can reflect the dynamic change situation of the vegetation coverage degree, and the texture of theregional vegetation coverage estimation result map is clear, which can reflect the vegetation coverage situation of the surface with strong heterogeneity.

The invention discloses a single-frame image-based high-dynamic range image generation method. According to the method, an edge image block set is found through edge detection, and the total distance of the edge image block set is calculated under a camera response inverse function; a corresponding conditional probability model is constructed; a sampling Gaussian mixture model is constructed according to 201 camera response curves in a DoRF database so as to describe a priori probability model; a maximum posteriori probability model is obtained by using a Bayesian framework; new camera response inverse functions are obtained under a condition that the maximum posteriori probability model is maximum; a Levenberg-Marquardt method is adopted to iteratively solve the new camera response inverse functions to obtain an optimal camera response inverse function; and reconstruction is performed according to the optimal camera response inverse function, so that a high-dynamic range image can be obtained. Image dynamic range expansion is performed in a high-brightness area and a low-brightness area simultaneously, and therefore, the problem of artifacts and details loss at the dark area of the image caused by a method only performing image dynamic range expansion in the high-brightness area can be avoided.

The invention discloses a text-based key personal name extraction method and system. The method comprises the steps of 1, executing a word segmentation operation on a target text, and extracting target words of which part of speech is a personal name; 2, performing statistics on an occurrence frequency of each target word in the target text, and setting a weight of the target word according to the occurrence frequency; 3, adjusting the weight of the target word according to an occurrence probability of the target word serving as the personal name and recorded in an ambiguous personal name priori probability dictionary; and 4, selecting the target word with large weight as a key personal name. Through the method, the extraction of figures related to specific events, the extraction of the key personal name in the text and the extraction of important spreading users, event development node users, public pointing users and information source users can be realized, and the accuracy and validity of figure extraction can be improved.

Systems and methods of precoded faster than Nyquist (FTN) signalling are provided. In the transmitter, Tomlinson-Harashima Preceding (THP) is applied to produce precoded symbols. The THP is based on inter-symbol interference (ISI) due to using faster than Nyquist (FTN) signalling. An inverse modulo operation is not performed in the receiver. Instead, in the receiver, FTN processing is performed based on a matched filter output by determining log a-posteriori probability ratio LAPPR values computed for an nth bit bn of a kth received symbol and pre-computed a-priori probabilities of an extended constellation for a given pulse shape h(t) and FTN acceleration factor combination.

The invention discloses an adaptive incremental particle filtering method for a Mars atmosphere entry section. The method comprises the following steps of: (1) initializing, namely sampling to obtain particles and endowing the same weight by employing a priori probability density function; (2) updating the particles and the weight by utilizing a likelihood probability density function of the particles under a motion equation and an incremental measurement equation; (3) comparing all weights calculated in the step (2), obtaining and storing the maximum weight, the minimum weight and corresponding serial numbers, and solving measurement information to calculate the Euclidean distance between the maximum and the minimum and the Euclidean distance from each particle to a particle which corresponds to the minimum weight according to the incremental measurement equation and an incremental measurement value corresponding to the current moment; (4) determining an adaptive coefficient value and recalculating the weight; (5) normalizing the weight calculated in the step (4), and obtaining a novel weight; (6) resampling; and (7) returning to the step (2) until the time is ended. According to the method, an unknown system error in the measurement system can be eliminated.

The invention discloses a model updating method, device and equipment. According to the scheme provided by the embodiment of the invention, the method includes under the condition that a shadow set Sand a target sample are given, respectively calculating prior probabilities alpha of training samples contained in the shadow set; and sampling according to the shadow set S, and training the samplingtraining parameter distribution of the sampling model obtained when the model is trained, calculating the posterior probability P of the target sample z in the shadow set according to the training parameter distribution of the given model and the characteristic value of the target sample; evaluating whether the privacy leakage degree of the trained model for the training sample set is qualified or not according to the difference value of the posterior probability P and the prior probability alpha; and if not, changing the affiliation relationship between the target sample and the training sample set, thereby obtaining a new training sample set and performing model adjustment to avoid privacy data leakage.

The invention provides a joint modulation coding method for deep-space link residual frequency offset. The joint modulation coding method for the deep-space link residual frequency offset is characterized by comprising the following steps, wherein a step A is that low density parity check (LDPC) coding is carried out to information bits at a transmitting terminal, and at the same time, the order of a coding sequence is changed through an intermediate conversion matrix so that the information bits are enabled to be located at positions where phase deviation is small; a step B is that Feher-patented quadrature phase shift keying (FQPSK) modulation is carried out to the coding sequence, and transmitting signals generated through the modulation are transmitted into an additive white gaussian noise (AWGN) channel; a step C is that the transmitted signals are received at a receiving terminal and demodulated by an MAP to obtain a bit-by-bit logarithmic likelihood ratio LLR, and then the obtained LLR is utilized for carrying out compensation and estimation for phase offset and frequency offset to codes; and a step D is that soft information is sent into a decoder eventually, a posteriori probability of the demodulation is used as a priori probability of decoding, and a belief propagation decoding algorithm based on the soft information is adopted to achieve a decoding process. The joint modulation coding method for the deep-space link residual frequency offset effectively assists in frequency offset compensation, and improves the frequency deviation resistant performance of a system.

This invention has as its object to implement the constitution of a receiver that receives signals sent by performing multi-valued pulse modulation and performs iterative decoding. The constitution includes: (1) a bank of pulse correlators that achieves correlation with all predetermined sent pulse waveforms, (2) a pulse demapper that calculates the log likelihood ratio for each bit of the interleaved code word from said pulse correlator outputs and a priori information for each bit, (3) a deinterleaver that performs deinterleaving on the output from said pulse demapper, (4) a decoder that calculates likelihood information for the deinterleaved code word bits and information bits, respectively, (5) an interleaver that interleaves the output of the decoder in the same manner as on the sending side, and (6) a feedback circuit that provides feedback of the output of said interleaver as a priori probability to the pulse demapper.