88 results about "Residual matrix" patented technology

The present invention relates to a quick scanning method for a residual error matrix in video coding process, which belongs to the field of video coding in signal processing. The method comprises: storing elements in a two dimension residual error matrix to a one-dimension serial according to a preset mapping way; extracting the lots of residual error elements gradually and orderly from the one-dimension serial, computing a level value, a map value and a run value of each residual error element, and storing them respectively. In the process of scanning the residual error matrix, the invention can parallelizably and effectively read the residual error element, accomplish parallelizably data computing and information extracting, and can quickly obtain the required scanning results, thereby, the invention greatly reduces the time cost in whole scanning process, enhances the coding rate of a coder, and facilitates the realization of real-time encoding under high distinguishability.

The invention provides an optimized video data coding method based on boundary adaptive transform. The method mainly comprises steps: according to the existing DCT (Discrete Cosine Transform) matrix for video coding, a new transform matrix is generated, the new transform matrix and the existing DST (Discrete Sine Transform)/DCT matrix are combined into a new transform matrix set; according to a prediction mode, the size of a transform unit, and the relationship between the boundary information of a transform unit and the boundary information of a prediction unit, a group of transform matrixes is selected from the new transform matrix set; and by using the group of new transform matrixes, inputted video data are transformed, and a coding stream is generated. The method of the invention makes full use of distribution characteristics of residual matrixes after inter-frame prediction, and based on the boundary information of the prediction unit and the transform unit, a different transform matrix is adopted for a different boundary characteristic to carry out transform. The optimized method can select a more efficient transform matrix for transform in a targeted mode according to the characteristics of residuals after inter-frame prediction, and the coding efficiency is thus improved.

The invention provides an abnormity detection and fault diagnosis algorithm of a gas turbine based on sensor associated network and relates to the technical field of gas turbine fault diagnosis. The diagnosis algorithm comprises following steps: step 1. signal processing; step 2: evaluating the correlation between any two sensor test points in a pretreated test point set; step 3. hierarchically clustering an obtained gas turbine sensor associated network model; step 4. evaluating the structure distribution of the residual matrix corresponding to the smallest granularity sub-class by using the information entropy index; step 5. calculating information entropy of all sub-classes in all layers and calculating super-class information entropy index of sub-classes whose information entropy indexes are transfinite. The abnormity detection and fault diagnosis algorithm of gas turbine based on sensor associated network is used achieving abnormity detection on the complete operation status of large gas turbines, especially suitable for the case where sensor test point information is very rich.

The invention discloses an anti-interference realization method of a self-adaptive array antenna, and aims at providing an anti-interference method with small calculated amount, strong real-time and better anti-interference effect. The method is realized through the following technical schemes: a data pre-processing module performs N-order time domain tapping on M path digital signals to obtain continuous space-time two-dimension snapshot data in a uniform circular array antenna with antenna array element number of M, and a space-time two-dimension guiding vector is obtained according to a desired signal direction provided by an upper computer; a guiding vector residual matrix is computed by using the space-time two-dimensional guiding vector through a minimum variance distortionless response MVDR algorithm simplification realization module, and the space-time two-dimension beam synthesis and space-time two-dimension weight iteration are synchronously performed according to a built-inbeam synthesis module and a weight iteration module by combining the space-time two-dimension snapshot data; and an array output module converts the digital signal after the beam synthesis into an analog signal through a digital-to-analog DA converter, and the analog signal is sent to a back-end processor after being output in a up-conversion way through an up-conversion module.

The invention provides a method for compressing the space of a finite automaton, comprising the steps: executing each status in the finite automaton in a clustering way to obtain a plurality of clusters for expressing a status set; classifying transfer edges of the each status in the finite automaton according to the clusters so as to obtain a plurality of cluster matrixes, bitmaps corresponding to the cluster matrixes and remaining matrixes, wherein the cluster matrixes comprise transfer edges for pointing to the same cluster, the bitmaps are used for describing the effectiveness of correlative elements in the cluster matrixes corresponding thereto, and the remaining matrixes comprise remaining transfer edges which are not contained in the cluster matrixes in the finite automaton; and extracting a basic value for each row in the cluster matrixes, converting the cluster matrixes into an offset matrix, merging the each row in the offset matrix, and adding with a index data set for marking a status which can be merged, to obtain compression matrixes of the cluster matrixes. The method improves not only compression effect but also matching speed when realizing regular expression match.

The invention relates to a method of estimating a noise and interference covariance matrix and to a receiver. The method comprises: estimating an initial noise and interference covariance matrix on the basis of a received signal; reducing the impact of the background noise of the receiver from the initial noise and interference covariance matrix for obtaining a residual matrix; accepting the obtained residual matrix when the residual matrix is at least approximately positive semidefinite; modifying the obtained residual matrix such that the positive semidefinity of the residual matrix is achieved when the residual matrix is not at least approximately positive semidefinite; and adding the impact of background noise back to the residual matrix for estimating a final noise and interference covariance matrix.

A robust width learning system collects training data and performs linear conversion processing on the training data; solving an extended input matrix by using the input data matrix and the enhanced node matrix, solving an iterative initial connection weight matrix by using a ridge regression algorithm, and solving a residual matrix by using a residual formula; obtaining a residual probability density function by using a kernel density estimation algorithm, and calculating a weight matrix formed by all training data; and solving the connection weight matrix of the kth iteration, and if the maximum value of the absolute value of the difference between the output weights of two adjacent steps is not greater than a set threshold value or the number of iterations reaches a preset maximum number of iterations, ending the iteration, stopping the training of the model by the robust width learning system, and establishing a robust width learning system model. According to the system, the robustness of a width learning system can be improved, the problem of adverse effects on modeling precision caused by outliers can be effectively suppressed, and a robust width system model can be conveniently established so as to be suitable for prediction of related indexes in a complex industrial process.

The invention discloses a vectorization realization method for a deconvolution matrix of a GPDSP. A CPU core of the GPDSP distributes corresponding scalar storage space and vector storage space for a weight matrix generated at a forward propagation stage and a residual matrix at a reverse calculation stage in a convolutional neural network. The method comprises the steps that S1, a residual matrix A(m,m), a weight matrix B(n,n) and a deconvolution result matrix C(m+n-1,m+n-1) are set, wherein m is greater than n; S2, by controlling the number of cycles, elements in the first (n-1) rows of the deconvolution result matrix C are calculated first; S3, the number of cycles is fixed, and elements from the n(th) row to the m(th) row of the deconvolution result matrix C are calculated; and S4, by controlling the number of cycles, elements from the reciprocal (n-1)(th) row to the reciprocal first row of the deconvolution result matrix C are calculated. The method has the advantages that the principle is simple, operation is convenient, a vector processor can be fully utilized to complete special data calculation, total algorithm running time is shortened, and algorithm execution efficiency is improved.

The invention discloses a multiple-input-multiple-output radar near-field positioning algorithm which includes the steps: firstly, performing estimation of a first weighted least square method by the aid of thinking of a chan algorithm; secondly, defining a new residual matrix, and performing estimation of a second weighted least square method; finally, substituting a result initial value of second estimation into a Taylor algorithm, performing iterative processing until an outputted positioning result is smaller than a minimum set threshold value. According to the positioning algorithm, near-field targets can be accurately positioned, and the positioning algorithm is applicable to multiple-input-multiple-output radar near-field positioning systems.

The invention relates to a coal mill fault early warning method based on DPC-MND and multivariate state estimation, and the method comprises the following steps: 1), building a training matrix K through operation data, corresponding to observation parameters related to a coal mill fault, in historical operation data of a coal mill, enabling the training matrix K to cover all dynamic change conditions of each observation parameter during normal operation of the coal mill; 2) adopting a density peak clustering algorithm based on mutual proximity to extract and construct a memory matrix D in thetraining matrix K, and constructing an MSET early warning model based on the memory matrix D; 3) obtaining an observation matrix Xobs in a to-be-estimated time period, inputting the observation matrixXobs into the MSET early warning model, and adopting multivariate state estimation to obtain an estimation matrix Xest; and 4) obtaining a residual matrix RL = Xobs-Xest between the estimation matrixXest and the observation matrix Xobs, setting a fault early warning threshold, and performing fault early warning through a sliding window method. Compared with the prior art, the method has the advantages of accurate early warning, real-time monitoring, accordance with actual operation, high fault early warning sensitivity and the like.

The invention relates to an efficient video coding zero-coefficient early detection method. The method includes the steps that firstly, a quantization parameter Qp, a two-dimensional DCT transformation matrix C, a residual matrix Re and the size N of a TU block are set; secondly, a zero block detection threshold value is worked out according to the Qp, the C, the Re and the N; thirdly, a detection parameter SAD corresponding to the TU block with the size of N is worked out; fourthly, for the TU block with the size of N, whether the SAD meets the inequation SAD<Th0 or not is judged, if yes, it is determined that the TU block is an all-zero block and the third step is executed again, and otherwise the fifth step is executed; fifthly, whether the SAD can meet the inequation SAD<Thj, j=1,2,...,N-1 or not is judged, if yes, a final quantization block from Z(0,0) to Z(j-1,j-1) with the size of j*j is subjected to DCT, quantization, inverse quantization and inverse DCT operation and the third step is executed again, and otherwise the total TU block is subjected to DCT, quantization, inverse quantization and inverse DCT operation and the third step is executed again. Compared with the prior art, the efficient video coding zero-coefficient early detection method has the advantages of being capable of improving video coding speed and reducing coding calculation complexity, high in detection efficiency and the like.

The invention discloses a water supply pipe network leakage monitoring and early warning method based on time sequence analysis. According to the method, relevance among data coming from pressure sensors distributed in different spaces and received at different times is analyzed in real time, and leakage warning is conducted after monitoring indexes exceed threshold values. The water supply pipe network leakage monitoring and early warning method includes the following steps that firstly, a leakage pressure residual matrix and real-time pressure residual vectors are established; secondly, a space-time relevance matrix is established; and thirdly, leakage monitoring and early warning are conducted. By means of the method, under the given scanning time window length (STWL), influences of model uncertainty and measurement uncertainty can be reduced, and therefore the real-time leakage recognition rate is increased, the false alarm rate (misinformation rate) is decreased, and the water supply pipe network leakage monitoring and early warning method has important significance in improving the comprehensive performance of a leakage monitoring system.

The invention relates to a label query and change method based on active learning, belonging to the active learning field. The method comprises the following steps: firstly, training a plurality of classification models by using a 10-fold cross-validation method and a strong-weak classification algorithm; Then calculating the residual matrix and confidence matrix; Secondly, splicing the two new matrices into one matrix horizontally as the eigenmatrix. using the K-Means algorithm to cluster the generated feature matrices, and obtaining four types of clusters A, B, C and D according to the threshold; Finally, relabeling some samples by active learning method for the suspected noise clusters (type A and type C), and obtaining the final noise clusters according to the labeling results. The sample labels of noise clusters are modified to correct labels, and the modified A, C samples and B, D samples are merged into the final training sample set. The invention has excellent effect of screening out label noise, can reduce the cost of manual labeling, and can obtain higher classification accuracy rate with very small active learning cost.

The invention discloses a priori knowledge-based method for monitoring a fault of a magnesite electric melting furnace by kernel flexible manifold embedding, and belongs to the technical field of fault detection and diagnosis. The method comprises the steps of performing fault type marking on partial data in a historical data set of a running process of the magnesite electric melting furnace; mapping historical data to a high-dimensional space; building an optimization model capable of obtaining a fault prediction matrix, and calculating the fault prediction matrix by utilizing the historical data set comprising marking data; and obtaining monitoring data of the magnesite electric melting furnace in real time, calculating a fault monitoring matrix by utilizing the historical data and the fault prediction matrix, and obtaining a fault type of the monitoring data. According to the method, priori knowledge is converted into knowledge identifiable for a machine, so that the magnesite electric melting furnace can safely and robustly run even if old workers are not on site; a regression residual matrix is added to the fault prediction matrix, so that a prediction matrix model is more flexible and better in generalization ability; and the kernel flexible manifold embedding method is used, so that data disasters caused by excessively high data dimension are avoided, and processing and fault diagnosis of a nonlinear system are realized.

The invention discloses an on-line gas sensor drifting correction method based on an internet of things and a mobile robot, and belongs to the field of the drifting correction of a sensor. The method comprises the following steps: step I, distributing and networking sensor nodes; step II, determining a route for the mobile robot to traverse all the nodes, and implementing the route; step III, setting a plurality of time windows, so as to obtain a residual matrix and a drifting component matrix; step IV, correcting two matrixes of a previous time window to data of the current time window according to a given weight; and V, completing the online correction for the drifting of the sensor. By adopting the correction method, the on-line drifting compensation of the gas sensor can be realized, the reliability of a wireless sensor network can be improved, and the service life of a system can be prolonged.