90 results about "Propagation of uncertainty" patented technology

An apparatus and method for monitoring a process involve development and application of a statistically qualified neuro-analytic (SQNA) model to accurately and reliably identify process change. The development of the SQNA model is accomplished in two stages: deterministic model adaption and stochastic model modification of the deterministic model adaptation. Deterministic model adaption involves formulating an analytic model of the process representing known process characteristics, augmenting the analytic model with a neural network that captures unknown process characteristics, and training the resulting neuro-analytic model by adjusting the neural network weights according to a unique scaled equation error minimization technique. Stochastic model modification involves qualifying any remaining uncertainty in the trained neuro-analytic model by formulating a likelihood function, given an error propagation equation, for computing the probability that the neuro-analytic model generates measured process output. Preferably, the developed SQNA model is validated using known sequential probability ratio tests and applied to the process as an on-line monitoring system. Illustrative of the method and apparatus, the method is applied to a peristaltic pump system.

This invention provides a system and method to propagate uncertainty information in a deep learning pipeline. It allows for the propagation of uncertainty information from one deep learning model to the next by fusing model uncertainty with the original imagery dataset. This approach results in a deep learning architecture where the output of the system contains not only the prediction, but also the model uncertainty information associated with that prediction. The embodiments herein improve upon existing deep learning-based models (CADe models) by providing the model with uncertainty/confidence information associated with (e.g. CADe) decisions. This uncertainty information can be employed in various ways, including (a) transmitting uncertainty from a first stage (or subsystem) of the machine learning system into a next (second) stage (or the next subsystem), and (b) providing uncertainty information to the end user in a manner that characterizes the uncertainty of the overall machine learning model.

The invention relates to a modeling approach and a modeling system of an acoustic model used in speech recognition. The modeling approach includes the steps of: S1, training an initial model, wherein a modeling unit is a tri-phone state which is clustered by a phoneme decision tree and a state transition probability is provided by the model, S2, obtaining state information of a frame level based on the fact that the initial model aligns the tri-phone state of phonetic features of training data compulsively, S3, pre-training a deep neural network to obtain initial weights of each hidden layer, S4, training the initialized network through error back propagation algorithm based on the obtained frame level state information and updating the weights. According to the modeling approach, a context relevant tri-phone state is used as the modeling unit, the model is established based on the deep neural network, weight of each hidden layer of the network is initialized through restricted Boltzmann algorithm, and the weights can be updated subsequently by means of error back propagation algorithm. Therefore, risk that the network is easy to get into local extremum in pre-training is relieved effectively, and modeling accuracy of the acoustic model is improved greatly.

A method and device for tracking error propagation and refreshing a video stream is provided. The proposed subject matter comprises of an error propagation tracking method that works in the sub-sampled domain to reduce computational cycles and memory bandwidth. Further, the tracking based update of the error propagation metric is done differently for static and non-static regions to avoid unnecessary refresh of static areas. Through suitable thresholding of the metric at a macroblock (MB) level, a set of refresh MBs are selected for each frame. These refresh MBs are coded either as an intra MB or as an inter MB that is predicted from one or more reliable reference frames (—frames that are known to be available at the decoder with negligible errors—). Such inter coding of refresh MBs improves the compression efficiency when compared to pure intra coding of refresh MBs. Further, variants to the threshold selection are presented that result in temporally uniform distribution of the number of refresh MBs and a strict refresh scheme wherein all MBs are guaranteed to be with negligible errors following a packet loss within a committed refresh period. In addition, to using the error propagation metric, spatial connectivity to already chosen refresh MBs is used in the selection of additional refresh MBs within a frame and across frames; this reduces the rate of error propagation due to part of a macroblock predicting from older, erroneous neighboring MBs and in turn requiring more refresh MBs on the average per frame.

The invention discloses a modeled clutter Doppler parameter estimation method of airborne radar. The clutter model provided by said invention is Doppler distribited clutter (DDC) model. The clutter covariance matrix described by said model is analyzed and expressed by four specific parameters of Doppler centre, Doppler spectrum-width extending coefficient, clutter roughness parameter and noise roughness parameter. The clutter parameter estimation method includes the following steps: (1) airborne radar can be used for transmitting pulse signal, and collecting echo data; 2). utilizing clutter data in collected signals to form itnerference covariance matrix of sample; and 3). according to the itnerference covariance matrix of sample combinatively estimating the unknown parameter in DDC clutter model. Besides, said invention also provides the advantage and application of said estimation method.

The invention relates to a long-endurance anti-jamming posture heading calibration method of an inertial satellite navigation integrated navigation system. The long-endurance anti-jamming posture heading calibration method is characterized in that in an inertial navigation system, inertial navigation resolving is performed to output posture heading, speed and position information; updating is performed; the position information of the satellite positioning navigation system is used as reference information; data fusion is performed with the position information of the inertial navigation system; according to fusion strategies, a mathematical model is determined according to the error spreading rule of the inertial navigation system; a kalman filter is used for performing online estimationand calibration on the error state of the navigation information of the inertial navigation system. The design is reasonable; the kalman filter is used for performing combined filtering on the navigation information of the inertial navigation system and the navigation information of the satellite navigation system; the optimum aerial carrier navigation information is obtained; in the severe environment, the fault of the satellite navigation system is subjected to diagnosis isolation; the posture angle error calibration problem of the inertial navigation system under the long-time work conditions can be solved; the accuracy and reliability of the posture angle error estimation can be ensured.

The invention discloses a global gravity field model inversion method. The method comprises the following steps: step 1: acquiring low-orbit gravity satellite data, and preprocessing the low-orbit gravity satellite data; step 2: calculating a pure Earth gravitation acceleration observation value by virtue of numerical differentiation; step 3: establishing a novel observation equation according toan observation value constant weight strategy; and step 4: estimating parameters, and obtaining an Earth gravity field model position coefficient. According to the acceleration domain residual error analysis of a reference model, a symmetric Toeplitz covariance matrix is established by directly adopting an autocorrelation sequence as a basic element, a complicated operation process of a high-dimensional huge matrix can be effectively avoided when the acceleration covariance matrix is calculated by an orbit prior dense array by virtue of an error propagation law, and the time for fixing the weight can be reduced, so that the calculation efficiency of the gravity field model can be improved. Compared with a classical method, an algorithm provided by the invention is simple and efficient in operation; and moreover, the stability and reliability of the inversion calculation efficiency of the gravity field model can be ensured.

A method of computer-assisted learning of control and/or feedback control of a technical system is provided. A statistical uncertainty of training data used during learning is suitably taken into account when learning control of the technical system. The statistical uncertainty of a quality function, which models an optimal operation of the technical system, is determined by uncertainty propagation and is incorporated during learning of an action-selecting rule. The uncertainty propagation uses a covariance matrix in which non-diagonal elements are ignored. The method can be used for learning control or feedback control of any desired technical systems. In a variant, the method is used for control or feedback control of an operation of a gas turbine. In another variant, the method is used for control or feedback control of a wind power plant.

The invention relates to a segmental counting-based relative radiation correction method comprising the following steps: acquiring original image data; determining segmental points and dividing the image into three brightness areas, including a low-brightness area, a medium-brightness area and a high-brightness area; performing histogram matching on the low-brightness area, the medium-brightness area and the high-brightness area respectively to obtain a corresponding brightness area lookup table of each probe element; and combining to obtain an integrated lookup table of each probe element and performing relative radiation correction on a satellite image. Aiming at the condition that integrating spheres in a lab are invalid, the method provides a method for effectively removing stripe noise of an image system and chromatic aberration of the system, so the lookup tables of different brightness areas are obtained through segmental counting according to the characteristic that the probe elements have different response models in different photoelectric response areas, error propagation in the counting process is reduced, the precision of the lookup tables is improved, the imaging quality of the image can be improved well, the algorithm has high efficiency, and the quasi-real time requirement of image production can be guaranteed timely.

The invention relates to an initial value correcting method for space target track improvement, which relates to the space technical field. In the invention, a nonlinear spreading model of prediction error coefficient related to along-track/normal direction prediction errors is established based on the prediction error spreading regulation of a space target track; nonlinear equation sets related to the prediction error coefficient of the along-track and normal direction prediction errors are respectively established based on multiple sets of observation datum acquired in different times; the evaluation to the along-track and normal direction prediction error coefficient is completed by fitting the along-track and normal direction errors; and based on the relationship between the prediction error coefficient and the track initial value error, the track initial value correcting amount is acquired, thus realizing the correction to the track initial value. Through repeated iteration, the accuracy of the track improvement initial value is improved; the track improvement process is promoted to be reliable; and the convergence is rapid.

The invention discloses a deep learning method of an attribute emotion word vector. The deep learning method comprises the following steps: a) using a label of 'common impression' in comments on e-commerce website products as an attribute emotion label to automatically mark a product comment data set, acquiring pre-training word vector dictionary through language model training; b) putting forward a deep learning model ASWV to infuse attribute emotion information contained in the comments into a word vector training process, obtaining an attribute emotion word vector; c) if iteration stop conditions are not satisfied, propagating back an attribute emotion error to update the word vector; and d) outputting the attribute emotion word vector. The deep learning method provided by the invention infuses attribute and emotion information of products into automatic learning of word vector characteristics by adopting a deep learning technology, automatically obtains attribute emotion word vector characteristics of the products without using a method for artificially designing characteristics in traditional natural language processing, avoids error propagation and superposition caused by stepwise strategy, and also overcomes a difficulty that emotion words can only have single emotion polarity constraint.

A replacement sensor model for optimally exploiting imaging data obtained by an image sensor. The replacement sensor model includes a ground-to-image function such as a polynomial for converting three-dimensional ground point data to two-dimensional image (pixel) data. Coefficients of the polynomial can be determined from a rigorous sensor model of the image sensor, which reflects physical characteristics of the particular image sensor. The polynomial can be adjusted by an adjustment vector. The adjustment vector can be determined from a pre-determined list of possible adjustment components having particular characteristics relating to the image sensor and reflecting errors in the physical characteristics of the image sensor as included in the rigorous sensor model. Finally, an error covariance associated with the adjustment vector can be determined from an error covariance associated with the rigorous sensor model. The adjusted polynomial and error covariance of the replacement sensor model permit optimal image exploitation, including error propagation and geopositioning.

The invention discloses a method for eliminating an underwater acoustic channel interference in underwater acoustic communication. The method comprises: step one, establishing an impulse response timedomain model and an impulse response frequency domain model of an underwater acoustic channel; step two, carrying out training by using a training sequence according to the impulse response time domain model of the underwater acoustic channel to obtain a channel impulse response matrix; step three, eliminating a channel interference of an information sequence by the training sequence by using thechannel impulse response matrix, thereby obtaining an information sequence after the interface elimination; step four, carrying out balancing between channels on the information sequence after the interface elimination to obtain estimation of the information sequence after channel balancing; and step five, carrying out soft-decision decoding on the estimation of the information sequence to obtaina final hard decision bit. According to the method disclosed by the invention, an accurate feedback symbol can be obtained; the inter-symbol interference is eliminated well; and the error propagationeffect is improved.

The present invention relates to an asteroid attachment trajectory robust optimization method, belonging to the technical field of planetary landing. The method comprises a step of using a linear covariance method to deduce the error spreading equation of an asteroid attachment process according to the mathematical model of uncertain factors of an initial state, thrust direction and size, a target celestial parameter and the like, a step of achieving the purposes of suppressing error spreading and improving landing trajectory robustness are achieved through introducing a tracking guide strategy in the trajectory optimization process. According to the method, the fuel consumption in a lander attaching process can be saved, at the same time, the influence of parameter uncertainties on trajectory can be reduced. According to the asteroid attachment robust optimization method, the linear covariance method is used to quantitatively describe the influence of random variables on a landing process, a tracking guide strategy is considered in the trajectory optimization process, thus the spread of various errors can be inhibited in the tracking process, and the purpose of improving the landing precision is achieved.

The invention relates to a digital signal processing method. The method solves the problem of unable to obtain the accurate wideband harmonic phase by using the existed technique. Based on the NTN correct technique, this invention obtains a serious wideband impulses, corrects orderly their time base aberration, uses average signal shift to eliminate co-modulate interference, corrects mismatch aberration, and signal dithering. Then phase-spreading and eliminating linearization against the obtained phase frequency response function obtains the harmonic phase. Finally, the phase uncertainty can be obtained by means of introducing the error spread formulation and assumption verification estimation in the complex frequency area. The wideband phase obtained by this invention can be used in the phase correction in large-signal network analyzer and hi-speed sampling oscilloscope.

Improved decision directed adaptation and decision feedback equalizers are provided in a block coded digital communication system. The performance of a receiver is significantly improved by allowing the decision feedback equalizer to perform time-tracking and residual frequency offset compensation during the data portion of a frame. This is accomplished by capitalizing on the inherent correlation among the chips of a code word in a block coded digital communication system to identify certain instances where more reliable symbol estimates can be derived from a sliced chip without introduction the delay inherent in decoding. As the more reliable symbol estimates are fed back into the chip slicer, the total efficiency of the decision feedback equalizer is improved and the more reliable symbol estimates can be used to replace older content in the feedback filter to further improve the accuracy of the modified slicer input and further decrease the effects of error propagation by the decision feedback equalizer.

Linear predictive system with classification of LP residual Fourier coefficients into two or more overlapping classes, and each class has its own vector quantization codebook(s). The use of strong and weak predictors minimizes codebook size by only quantizing the difference between Fourier coefficients of a frame and the Fourier coefficients predicted from a prior frame. The choice of using either a strong or weak predictor adapts to the prior choice of predictor so that a strong predictor following a weak predictor is changed to a weak predictor to insure attenuation of error propagation as arise from frame erasures.

A registration solution between two or more sets of point cloud data is provided to register a first target set of point cloud data with a second source set of point cloud data. Through error propagation, degrees of freedom masking and transformation filtering, a system may provide an intelligent method of enabling and disabling degrees of freedom in a real-time registration application. A system may adapt to varying environments (feature-rich and featureless) ensuring a high-integrity registration estimate through providing estimates of the observable parameters. By optionally and iteratively masking zero or more degrees of freedom from the registration estimate, a system may locate desirable data to use in a registration estimate. Then, the system may blend updates from the registration estimates to achieve the best available registration solution applied to the original second source set as well as additional sets of point cloud data.

The invention provides an analyzing method of free phase ballistic error propagation under nonspherical perturbation effect of the earth. For a typical ballistic missile, firstly, under a condition of considering the nonspherical perturbation effect of the earth, a high precisely simplified model applicable to free phase motion analysis and calculation is built; then according to small error assumption and linear system theories, an equiangular/equal geocentric distance/isochronous propagation model and the state transition matrix analytical solution of the equiangular/equal geocentric distance/isochronous propagation model are exported; further through decoupling lengthways and lateral propagation models, lateral errors resulting from higher order terms are corrected based on a least square fitting method; and the balance between the speediness and the accuracy in calculating the free phase ballistic characteristic under nonspherical perturbation effect of the earth is realized. The method provided by the invention has the advantages of rapid calculation speed, wide application area and accurate correction model and satisfying demands for rapid mobile lunch and on-missile real time calculation.

The invention provides a long-range rocket launching initial state error spreading estimation method, and relates to the long-range rocket launching initial state error spreading estimation method. The method aims at solving the problems that the calculation efficiency of burnout point position deviation, speed deviation, landing point longitudinal deviation and transverse deviation is low due to the existing launching initial state error, so that the spreading mechanism of the launching initial state error in the trajectory design process cannot be sufficiently analyzed. The goal of the method is achieved through the following technical scheme that the method comprises the following steps of: 1, building a kinetics perturbation equation; 2, solving long-range rocket thrust acceleration deviation, pneumatic acceleration deviation, normal gravitational acceleration deviation, Coriolis acceleration deviation and centrifugal acceleration deviation; and 3, obtaining similar analytical solutions of the burnout point position deviation and the speed deviation due to the long-range rocket launching initial state error and similar analytical solutions of the landing point longitudinal deviation and transverse deviation according to the first step and the second step. The method is applied to the field of long-range rocket or carrier rocket flight dynamics.

A decision feedback-based communication system comprising a unit for computing an estimated Error Propagation Probability (EPP) for a plurality of received signals, and a unit for receiving the EPP estimation and improving detected signal quality based thereon. The system can be a coded system or an uncoded system.

A method of line coding is disclosed that limits error propagation in a decision feedback equalizer (DFE) of a receiving device. A communications device receives a set of bits to be transmitted over a channel and divides the set of bits into a plurality of blocks based, at least in part, on a line coding scheme. The device then encodes each of the blocks of bits into a corresponding block of symbols based on the line coding scheme. Specifically, the line coding scheme has a non-uniform coding efficiency, wherein a first bit or a last bit of each block of bits is mapped to a single data symbol. For some embodiments, the line coding scheme may be a ternary line coding scheme that maps a block of 3k+1 bits to a corresponding block of 2k+1 symbols, where k is an integer greater than 1.