79 results about "Laplace distribution" patented technology

The disclosure relates to a method for performing context based binary arithmetic coding with a stochastic bit-reshuffling scheme in order to improve MPEG-4 fine granularity scalability (FGS) based bit-plane coding. The method comprises steps of: replacing 8×8 DCT with 4×4 integer transform coefficient in MPEG-4 AVC (Advance Video-Coding); partitioning each transform coefficient into significant bit and refinement bit; setting up significant bit context based on energy distribution within a transform block and spatial correlation in adjacent blocks; using an estimated Laplacian distribution to derive coding probability for the refinement bit; and using the context across bit-planes to partition each significant bit-plane for saving side information bit.

The invention discloses a scanning radar angle super-resolution imaging method. The scanning radar angle super-resolution imaging method comprises the following steps that a backward model for real beam scanning radar angle super-resolution imaging is established according to the Bayesian theory, the likelihood function relationship between a target and an echo and prior information of the target are respectively expressed through the Poisson distribution and the Laplace distribution in the model, radar angle super-resolution imaging is expressed as a posterior probability problem existing between the target and the echo, finally, according to the convex optimization theory, based on the nonlinear optimization method and approximate treatment, the target information corresponding to the maximum posterior probability is solved, and radar angle super-resolution imaging is achieved through target information reconstruction. By the adoption of the scanning radar angle super-resolution imaging method, super resolution of multiple targets in a real beam can be achieved.

A method for discriminating noise from signal in a noise-contaminated signal involves decomposing a frame of samples of the signal into decorrelated components, and using a difference between probability distributions of the noise contributions and the signal contributions to identify signal and noise. A Gaussian distribution is used to determine whether the components are only noise whereas a Laplacian distribution is used to determine whether the components contain the signal. Such discrimination may be used in speech enhancement or voice activity detection apparatus.

The invention discloses an airborne scanning radar imaging method in an iteration compression mode. The airborne scanning radar imaging method is characterized in that an azimuth-direction super-resolution imaging back model is built under the Bayesian theory framework, and the model built under the Bayesian theory can effectively fuse prior information of a target. When the super-resolution imaging back model is built, the assumption of a radar imaging basic idea on the noise statistical property is followed, and the Gaussian distribution function is adopted in the model to describe the noise statistical property. Due to the fact that target scattering has sparsity relative to an imaging background, the prior information of target scattering is described through the Laplace distribution function, radar super-resolution imaging is precisely converted into the maximum posterior probability problem in mathematics, corresponding target information when the posterior probability is maximum is solved, then the target information is reconstituted, and therefore the radar super-resolution imaging is achieved.

The invention provides a detection method for secondary compression of a JPEG image. By using the characteristic that an AC coefficient histogram in a DCT coefficient of primary JPEG compression accords with Laplace distribution, a Laplace distribution function of the AC coefficient of the primary JPEG compression of an image to be detected is estimated in a fit mode according to the AC coefficient distribution of the JPEG image to be detected; the statistic is detected through a difference structure of normalized histogram distribution of the fit distribution function and an actual AC coefficient; and whether the JPEG image is subjected to the secondary compression is judged by using SVM classification. The method has high detection accuracy and wide applicability.

The invention discloses a face cheating detection method based on domain adaptive learning and domain generalization. The face cheating detection method mainly comprises the following steps: constructing an encoder based on a deep residual network; constructing a classifier for detecting face cheating; constructing a discriminator which is used for guiding characteristics to accord with Laplace distribution; forming a training network by using the three parts; constructing a loss function of network training; setting a model optimization algorithm; processing the training data set sample imageto change the size; training and optimizing network parameters; processing the test image to change the size; and carrying out face cheating detection by using the trained encoder and classifier. According to the invention, common features of source domain training data are extracted through a maximum mean difference MMD training encoder; and meanwhile, by combining AAE technology of an anti-autoencoder. The characteristics conform to Laplace distribution, the generalization performance of the detection method is further improved, and the detection performance of the method on face spoofing attacks under complex conditions in practical application is effectively improved.

The invention is applicable to the technical field of privacy protection, and provides a spectral clustering method based on differential privacy protection. The method includes the steps of pre-processing sample data; calculating a similarity matrix; based on a k-near-value, simplifying the similarity matrix; adding the random noise satisfying Laplace distribution to the similarity matrix; constructing an adjacent matrix and a degree matrix based on the similarity matrix after random noise perturbation; obtaining the Laplace matrix based on adjacency matrix and degree matrix; obtaining the first m large eigenvalues and corresponding eigenvectors of Laplace matrices; normalizing the eigenvector to form eigenmatrix; using k-means clustering method to cluster the feature matrix to get the label of clustering. A spectral clustering algorithm is used to calculate the sample similarity between the sample data as the weight value between the data points, and then differential privacy algorithm is used to add random noise of Laplace distribution to the weight value to interfere with the weight value to achieve the purpose of privacy preservation. The interfered data can not only achieve privacy preservation but also ensure the effectiveness of clustering.

The invention relates to a likelihood-function-particle-filter-based power battery state-of-charge (SOC) estimation method and system. A state equation and a measurement equation are obtained by a battery Thevenin model; after parameter initialization, state prediction is carried out and a mean value and a covariance of a state prediction value are calculated; and sampling is carried out again and a sampling distribution function is reconstructed. A voltage prediction value of a battery end is calculated, a particle weight is calculated, weight normalization is carried out, and an effective particle number is calculated. The effective particle number Neff is compared with an effective particle number threshold value; when the Neff is less than Nthr, a Laplacian distribution is used as a likelihood function and a variance regulatory factor and a working condition adaptation factor are introduced, thereby modifying a variance of the likelihood function in an adaptive mode and thus realizing adaptation to different working conditions of the power battery. And then an updated SOC estimation value and covariance are obtained finally. According to the system, a microcontroller is connected with a voltage sensor and a current sensor; and all program execution modules are arranged in the microcontroller. According to the invention, the effective particle number is increased; overcorrection of the variance is effectively avoided; and the estimation precision is high.

The invention discloses an electric system active power real-time dispatching method considering wind power prediction error. The electric system active power real-time dispatching method comprises the following steps of describing random distribution character of the wind power prediction error by using Laplace distribution; by taking economic efficiency optimum and wind curtailment minimum as two targets, introducing a chance constraint condition; solving active power flow probability distribution of a system on the basis of a probabilistic power flow method for Latin hypercube sampling; establishing a buffer generator set active power optimum allocation model considering the wind power prediction error, and solving a chance constraint planning model by adopting an improved genetic algorithm. According to the electric system active power real-time dispatching method disclosed by the invention, the wind power prediction error is considered during real-time dispatching, power unbalance and flow off-limit brought to the system by wind power output offset can be avoided, and the safe operation of the system is guaranteed; moreover, while the wind power prediction error is assimilated, the wind power accepting ability is increased by the system.

Main points of the invention are as following: using rate distortion theory of message source of possessing statistic characteristics of Laplace distribution to build secondary rate distortion model; using secondary rate distortion model to control video encoding rate. Secondary rate distortion model built in the invention is better accorded with real characteristics of rate distortion of video message source. The invention makes encoding output rate of video sequence closer to given channel bandwidth. The invention raises accuracy for controlling video encoding rate, and raises video encoding quality. Advantages are: better performance of rate distortion, applicable to real time multimedia communication and stream media through network.

The invention discloses a radar system performance index dynamic evaluation method based on Bayesian machine learning. The method comprises the following steps: modeling a likelihood function of radartarget detection data into Gaussian distribution; modeling the prior distribution of the target performance index to be evaluated into Laplace distribution; performing hierarchical modeling on the target to-be-evaluated performance index by utilizing the Bayesian hierarchical probability model; and performing approximate solution on the posterior distribution of the to-be-evaluated performance indexes of each target by using a variational Bayesian expectation maximization method, thereby obtaining a posterior probability density function and the like. The Bayesian machine learning method is adopted to realize the dynamic evaluation of the radar performance indexes, the biggest advantage is that under the model assumption condition, the analyzed dynamic indication result can be obtained only by observing the data once, the test frequency can be obviously reduced, the test cost and period are reduced, the analyzed index dynamic change range is provided, and the availability and the robustness are obviously improved. Therefore, the applicability for the dynamic evaluation of the radar system performance indexes is good.

A method is provided for decoding digital image data in order to improve a picture quality of a reproduced image by performing a dequantizing in consideration of an input DCT coefficient having a laplacian distribution. The digital image data decoding method of a digital image data decoder may dequantize digital image data using a quantizer having characteristics of mapping an input DCT coefficient x_ij to a restoration level y_ij. This may occur by estimating a probability distribution function p(x_ij) of the input DCT coefficient x_ij, calculating a mass center C_m, and setting the mass center as a restoration level.

The invention discloses a method for simulating propagation characteristics of a fading channel in a coupled multi-antenna indoor space. The method comprises the following steps: (1) generating an lthcluster transmitting angle and an lth cluster receiving angle which are randomly distributed; (2) generating a transmitting deviation angle and a receiving deviation angle conforming to the k sub-path in the l cluster of the transmitting end and the receiving end which are in Laplace distribution; Generating an emission angle of an lth cluster and a kth sub-path of the transmittingend; (3) generating an arrival angle of the k subpath in the l cluster of the receiving end; (4) generating the amplitude of each sub-path conforming to Rayleigh distribution; Generatinga phase of each sub-path obeying [0, 2] uniform distribution; (5) generating logarithmic normal shadow fading numerical random variables; (6) generating a guide vector of each antenna by combining the antenna layout; And (7) generating a channel transmission matrix of the whole tight coupling MIMO system. According to the method, the influence of large-scale shadow fading and the array element electromagnetic coupling effect in an indoor propagation environment is fully considered, and the channel characteristics of the indoor MIMO multi-antenna transmission system are accurately simulated and evaluated.

The invention provides a recommendation method based on label and differential privacy protection. The method aims to introduce a label concept, similarity of the tags through a co-occurrence principle of the tags is calculated. The tag similarity is used for replacing the Euclidean distance of fuzzy c-means clustering to carry out fuzzy c-means clustering on the tags, so that the tags belong to different clusters. The problem of low recommendation accuracy caused by the hard clustering problem of conventional clustering is solved, and noise conforming to Laplace distribution is added in the clustering process to achieve the purpose of protecting the privacy of a user. And fuzzy c-means clustering is carried out on the tags, so that the problem of data availability reduction caused by directly adding noise into the tag interest degree vectors is further solved, and the privacy of a user is protected while the recommendation accuracy is improved.

The invention discloses a hybrid Laplace distribution-based wind power fluctuation quantity probability distribution model building method. The method comprises the steps of firstly, calculating a wind power fluctuation quantity sequence according to actually measured wind power data of a wind farm and a default time scale; building a hybrid Laplace distribution model; and determining a parameter of the hybrid Laplace distribution model according to the wind power fluctuation quantity sequence, thereby obtaining a wind power probability distribution model. By the wind power probability distribution model obtained by the method, the wind power fluctuation characteristics can be accurately described; especially, the accuracy of heavy-tailed characteristic description of wind power fluctuation distribution is improved; and aiming at the description problem of wind power fluctuation of different temporal and spatial scale levels, the model can also reach the satisfactory accuracy.

The invention discloses a probability linear discriminant analysis image classification method based on L1 bound norm and solves a problem of an abnormal value existing in an image. Different from a traditional PLDA, Laplace distribution is employed to describe noise, Laplace is a probability density function based on the L1-bound norm, so an error value can be prevented from being amplified; through introducing a hidden variable, parameters of a variational expectation maximization solution model and a dimension reduction matrix are utilized; the dimension reduction matrix is taken as characteristics of a sample, the L1-bound norm is utilized in the model to realize error description, the solved dimension matrix is closer to a main direction, and the image classification effect can be improved.

The invention discloses a nonlinear industrial process robust identification and output estimation method. The invention relates to the field of industrial process modeling and model parameter identification, and aims to solve the problem that in the prior art, when an abnormal value exists in output data, the system identification precision is reduced, and the method comprises the following steps: 1, selecting a system local model, and establishing a robust probability model of a multi-model nonlinear system based on Laplace distribution; 2, according to the variational Bayesian framework, establishing an iterative updating formula of hidden variable posteriori distribution and to-be-estimated parameters; and 3, setting the posteriori distribution of the hidden variables established in the step 2 and a termination condition of a to-be-estimated parameter iteration updating formula, recording a final iteration result as an optimal parameter for estimation when iteration is terminated,and further obtaining a model output value through local model interpolation.

The invention relates to a sparse coding-based human face identification method. In a conventional sparse coding method, residues in coding are often assumed to meet a certain probability distribution form such as Laplace distribution or Gaussian distribution in advance, and based on this, an l1 or l2 norm form is proposed for solving a sparse coding coefficient; and due to the abovementioned processing means, the robustness of human face identification is seriously influenced, especially when shielding, noises or interferences in other forms exist. The method provided by the invention introduces an iterative optimization thought for the deficiencies of the conventional sparse coding method, and mainly solves the following two problems: firstly, the distribution form of residual errors does not need to be assumed in advance, so that the influence of an unreasonable residual error distribution function on the robustness of the human face identification is avoided; and secondly, a part of available pixel points are selectively reserved to perform identification, so that the problems of shielding, pixel damage and the like are better solved while the calculation amount is greatly reduced, and robuster identification performance is obtained.