51 results about "Matrix projection" patented technology

The invention relates to a random sampling analog circuit compressed sensing measurement and signal reconstruction method, which belongs to the field of electronic system test and fault diagnosis. Aiming at a fault signal having a sparsity distribution characteristic per se or in an orthogonal space in an output response of an analog circuit, a test node is selected according to a circuit topology structure, circuit output responses are randomly sampled under a distributed sensor test network, response signals are expressed in a sparse way on a transform domain by utilizing discrete orthonormal basis, compressed sensing measurement of the sparse signals is completed under observability matrix projection, and when the recovery rate of signal reconstruction by randomly compressed sampling points reaches more than 80 percent, the compressed measurement values of the circuit output responses are effective, can form a characteristic set and can be used for analog circuit fault diagnosis. The method solves the problems that the traditional analog signal sampling occupies a large number of hardware resources, large signal reconstruction calculated amount and the like; and the random sampling compressed sensing measurement method is utilized to improve the efficiency of electronic system testing.

The invention discloses a frequency domain compressive sensing method aiming at sparse SAR (Synthetic Aperture Radar) images in airspace and belonging to the technical field of signal processing. The frequency domain compressive sensing method particularly comprises the following steps of: step 1: determining the directions of original SAR images, with sparsity; step 2: carrying out Fourier transform on the original SAR images along the directions with the sparsity to obtain frequency domain images of the directions; step 3: building frequency domain sparse reconstructed models, solving model parameters, establishing observation vectors, and reconstructing frequency domain signals so as to form reconstructed frequency domain images; and step 4: carrying out the Fourier transform on the reconstructed frequency domain images along the directions to obtain reconstructed images. In the invention, by analyzing the sparsity of the SAR images in the airspace, the frequency domain sparse reconstructed models are built by aiming at the frequency domain signals, the model parameters are estimated, projection is carried out on the basis of an appropriate observation matrix and the frequency domain signals are reconstructed by utilizing a small quantity of observed values.

The invention discloses a phase encoding characteristic and multi-metric learning based vague facial image verification method. The phase encoding characteristic and multi-metric learning based vague facial image verification method comprises (1) a training phase, namely, partitioning sampling images and extracting multi-scale primary characteristics of every image block, performing fisher kernel dictionary learning through the above characteristics to generate into partitioning fisher kernel coding characteristics, performing multi-metric matrix learning on the above coding characteristics to generate a plurality of metric matrixes and obtain the metric distance after training samples are performed on multi-metric matrix projection, calculating the average metric distance and variance of positive samples and negative samples to a set and confirming a final classification threshold through a probability calculation formula of Gaussian distribution and (2) a verification phase, namely, partitioning input facial images and extracting multi-scale primary characteristics, generating partitioning fisher kernel coding characteristics, obtaining the final metric distance through the multi-metric matrix and comparing the distance and the threshold to obtain a facial image verification result. The phase encoding characteristic and multi-metric learning based vague facial image verification method has the advantages that the identification rate is high and the universality is strong.

The invention discloses an inter-class inner-class face change dictionary based single-sample face identification method to solve the problem of limitations of the current single-sample face identification algorithm. The method comprises the steps of step1, obtaining expressions of face images in the compression domain; step2, building a face image training sample matrix containing k classes; step3, building an average face matrix and an inter-class face change matrix of a face database; step4, adding low rank and sparse constraints into the inter-class face change matrix; step5, solving an inter-class similarity matrix and an inter-class difference matrix; step6, projecting the average face matrix, the inter-class similarity matrix and the inter-class difference matrix to low-dimensionality space; step7, performing normalization processing on the dimensionality reduced average face matrix, the inter-class similarity matrix and the inter-class difference matrix through a normalization method, and performing iterative solution on the face image training sample matrix based sparse coefficient vectors through a norm optimization algorithm; step8, selecting column vector face labels in the average face matrix, which are corresponding to the sparse coefficient maximum, to serve as the final face identification result.

The invention discloses an end member extract method of a hyperspectral image, comprising the following steps: projecting a dimensionality reduced hyperspectral image spectrum matrix onto a hyperplane to obtain a monomorphous; selecting the centre direction of an observation spectrum vector set as a projecting direction; projecting the monomorphous alongthe projecting direction; according a projecting result, obtaining a projecting extreme value and a pixel position corresponding thereto, setting up a pixel position index; obtaining a spectrum curve by the projecting extreme value and the position index; getting a next projecting direction with the position index, projecting on a new projecting direction and getting the projecting extreme value, the position index corresponding thereto and the spectrum curve of the end member until that the obtained end member reaches preset end member quantity. The method can reach better extracting precision and better stability.

A method clusters samples using a mean shift procedure. A kernel matrix is determined from the samples in a first dimension. A constraint matrix and a scaling matrix are determined from a constraint set. The kernel matrix is projected to a feature space having a second dimension using the constraint matrix, wherein the second dimension is higher than the first dimension. Then, the samples are clustered according to the kernel matrix.

A method is disclosed for performing multiple classification of an image simultaneously using multiple classifiers, where information between the classifiers is shared explicitly and is achieved with a low-rank decomposition of the classifier weights. The method includes applying an input image to classifiers and, more particularly, multiplying the extracted input image features by |Σ| embedding matrices Ŵc to generate a latent representation of d-dimensions for each of the |Σ| characters. The embedding matrices are uncorrelated with a position of the extracted character. The step of applying the extracted character to the classifiers further includes projecting the latent representation with a decoding matrix shared by all the character embedding matrices to generate scores of every character in an alphabet at every position. At least one of the multiplying the extracted input image features and the projecting the latent representation with the decoding matrix are performed with a processor.

The invention relates to a human face feature extraction and classification method. The method includes the following steps that feature dimension reduction is performed on human face images through a 2D-PCA method, and a high-dimensional image matrix is converted into a low-dimensional image matrix; the low-dimensional image matrix is converted into one-dimensional column vectors; according to the one-dimensional vectors of images of a training set, an intra-class divergence matrix S<W> and an inter-class divergence matrix S in the training set are solved, and eigenvalue decomposition is performed on the S<W> and the S respectively, wherein the eigenvalue decomposition formulae are as follows: and ; D<alpha> is used for estimating , D<beta> is used for estimating , and the formulae and can be obtained; the column space W<1> of and the column space W<2> of are respectively solved, and the optimum projection space W=[W<1>,W<2>] of a feature extraction algorithm with two stages of LDA based on 2D-PCA is obtained; the low-dimensional image matrix in the first step is projected into the optimum projection space W, and then feature vectors of the images are obtained; classifier training is performed on the feature vectors obtained in the sixth step through an SVM+NDA model, and a final human face classifier is obtained.