354 results about "Nonnegative matrix" patented technology

Object clustering techniques are disclosed. A nonnegative sparse similarity matrix is constructed for a set of objects. Nonnegative factorization of the nonnegative sparse similarity matrix is performed. Objects of the set of objects are allocated to clusters based on factor matrices generated by the nonnegative factorization of the nonnegative sparse similarity matrix.

We describe a method of blind source separation for use, for example, in a listening or hearing aid. The method processes input data from multiple microphones each receiving a mixed signal from multiple audio sources, performing independent component analysis (ICA) on the data in the time-frequency domain based on an estimation of a spectrogram of each acoustic source. The spectrograms of the sources are determined from non-negative matrix factorization (NMF) models of each source, the NMF model representing time-frequency variations in the output of an acoustic source in the time-frequency domain. The NMF and ICA models are jointly optimized, thus automatically resolving an inter-frequency permutation ambiguity.

A method and system separates components in individual signals, such as time series data streams. A single sensor acquires concurrently multiple individual signals. Each individual signal is generated by a different source. An input non-negative matrix representing the individual signals is constructed. The columns of the input non-negative matrix represent features of the individual signals at different instances in time. The input non-negative matrix is factored into a set of non-negative bases matrices and a non-negative weight matrix. The set of bases matrices and the weight matrix represent the individual signals at the different instances of time.

A method, apparatus and machine-readable medium analyze documents processed by non-negative matrix factorization in accordance with semantic topics. Users construct queries by assigning weights to semantic topics to order documents within a set. The query may be refined in accordance with the user's evaluation of the efficacy of the query. Any document that does not result in data indicative of significant correlation with at least one semantic topic is flagged so that a user may make a manual review. The collection of semantic topics may be continually or periodically updated in response to new documents. Additionally, the collection may also be “downdated” to drop semantic factors no longer appearing in new documents received after an initial set has been analyzed. Different sets of semantic topics may be generated and each document evaluated using each set. Reports may be prepared showing results for a body of documents for each of a plurality of sets of semantic topics.

An implementation of NMF functionality integrated into a relational database management system provides the capability to apply NMF to relational datasets and to sparse datasets. A database management system comprises a multi-dimensional data table operable to store data and a processing unit operable to perform non-negative matrix factorization on data stored in the multi-dimensional data table and to generate a plurality of data tables, each data table being smaller than the multi-dimensional data table and having reduced dimensionality relative to the multi-dimensional data table. The multi-dimensional data table may be a relational data table.

A method and system denoises a mixed signal. A constrained non-negative matrix factorization (NMF) is applied to the mixed signal. The NMF is constrained by a denoising model, in which the denoising model includes training basis matrices of a training acoustic signal and a training noise signal and statistics of weights of the training basis matrices. The applying produces weight of a basis matrix of the acoustic signal, of the mixed signal. A product of the weights of the basis matrix of the acoustic signal and the training basis matrices of the training acoustic signal and the training noise signal is taken to reconstruct the acoustic signal. The mixed signal can be speech and noise.

The invention discloses an abnormal human behavior detecting method based on a video sequence. The abnormal human behavior detecting method based on the video sequence comprises the following steps of; firstly, carrying out movement target detection on a video image sequence acquired by a camera; secondly, extracting movement characteristics of a human movement region, and carrying out characteristic dimension reduction on the movement region by using nonnegative matrix decomposition to obtain local characteristics of the movement target; thirdly, fusing the movement characteristics with the local characteristics to obtain the characteristic vectors of the movement behaviors; and finally, identifying and classifying the movement behaviors by adopting an SVM (Support Vector Machine) multi-class classifier based on a voting algorithm and judging identified behaviors of a movement target. According to the abnormal human behavior detecting method based on the video sequence disclosed by the invention, a plurality of characteristics are fused while the dimensions are reduced, and the respective recognition performances are improved. The abnormal human behavior detecting method based on the video sequence is applied to the intelligent video monitoring and capable of judging the abnormal behaviors by automatically analyzing the behaviors of the moving body in the public place, so that the computer can assist the people to complete the monitoring task and can complete the monitoring task in place of the people.

The invention relates to the technical field of image super-resolution processing, in particular to a non-negative matrix factorization-based face super-resolution processing method. The method comprises the following steps: performing face alignment on high-resolution face images in a sample library, reading the aligned sample image library, utilizing a non-negative matrix factorization algorithm to perform a factorization operation to obtain a basic image W, performing alignment on input low-resolution face images to obtain the non-negative matrix factorization expression coefficient e of a target high-resolution face image, obtaining the target high-resolution image Z1=We in combination with the basic image W and the expression coefficient e and dividing the important areas of the face images in the sample library; performing factorization synthesis on the divided local areas; and weighting and combining the synthesized local area and the image Z1 to obtain a super-resolution image Z2. The method has the advantages of increasing semantic constraint like that the grayscale of the image is non-negative, improving the expression capacity of the characteristic basic image and finally improving the quality of the super-resolution image.

The invention discloses a face recognition method based on weighted diagnostic sparseness constraint nonnegative matrix decomposition, and mainly aims to solve the problem that the method in the prior art is not robust to an obscured face and is of low recognition rate. According to the technical scheme, the method comprises the following steps: (1) constructing a nonnegative weight matrix according to the obscured area of a test image; (2) introducing the weight matrix into a general KL divergence objective function, applying a sparseness constraint to a basis matrix, and applying intra-class and inter-class divergence constraints to a coefficient matrix to get a weighted diagnostic sparseness constraint nonnegative matrix decomposition objective function; (3) solving the objective function, and decomposition-training a data matrix to get a basis matrix and a coefficient matrix; (4) projecting a test data matrix on the basis matrix to get a corresponding low-dimensional representation set, and taking the low-dimensional representation set as final test data; and (5) using a nearest neighbor classifier to classify the test data by taking the coefficient matrix as training data, and outputting the result. By using the method, the effect of obscured face recognition is improved. The method can be used in identity recognition and information security.

A method and system denoises a mixed signal. A constrained non-negative matrix factorization (NMF) is applied to the mixed signal. The NMF is constrained by a denoising model, in which the denoising model includes training basis matrices of a training acoustic signal and a training noise signal, and statistics of weights of the training basis matrices. The applying produces weight of a basis matrix of the acoustic signal of the mixed signal. A product of the weights of the basis matrix of the acoustic signal and the training basis matrices of the training acoustic signal and the training noise signal is taken to reconstruct the acoustic signal. The mixed signal can be speech and noise.

The invention belongs to the technical field of statistical model identification and machine learning, and in particular discloses a nonnegative matrix factorization-based dimensionality reducing method used for clustering. The method comprises the following steps of: adopting a KL distance; adding a data normalization constraint; directly discovering the internal relation among data dimensionalities by minimizing an object error function between data compression and reconstruction; obtaining a mapping matrix; and projecting high-dimensional data to a low-dimensional subspace by using the mapping matrix so as to perform effective data analysis such as clustering and the like. A simpler iterative formula compared with an original factorization method is obtained by the method; and normalization can be maintained naturally in each iterative updating. The normalization makes a final mapping matrix have higher sparsity compared with the original factorization method. In the obtained low-dimensional space, a clustering result shows that a more effective low-dimensional data characteristic can be obtained by the method; and an algorithm is simple and effective.

The invention discloses a self-adaptive hyperspectral image unmixing method based on region segmentation. In consideration of coexistence of linear mixing and bilinear mixing, the method is implemented by adopting the following steps: inputting a hyperspectral image; estimating the number of end elements with a minimum error based hyperspectral signal recognition method; extracting end element matrixes with a vertex component analysis algorithm; clustering hyperspectral data with a K-means clustering method, and segmenting the image into a homogeneous region and a detail region; adopting a linear model for the homogeneous region and performing unmixing with a sparse-constrained non-negative matrix factorization method, and adopting a generalized bilinear model for the detail region and performing unmixing with a sparse-constrained semi-non-negative matrix factorization method. According to the method, characteristics of the hyperspectral data and abundance are combined, the hyperspectral image is represented more accurately, and the unmixing accuracy rate is increased. The sparse constraint condition is added to the abundance, the defect of high probability of local minimum limitation of the semi-non-negative matrix factorization method is overcome, more accurate abundance is obtained, and the method is applied to ground-object recognition for the hyperspectral image.

A system is provided for managing an infrastructure. An extraction engine is in communication with a managed infrastructure that includes physical hardware. A signalizer engine includes one or more of an NMF engine (Non-negative matrix factorization), a k-means clustering engine (a method of vector quantization), and a topology proximity engine. The signalizer engine determines one or more common characteristics of events and produces clusters of events relating to the failure or errors in the infrastructure. The signalizer engine uses graph coordinates and optionally a subset of attributes assigned to each event to generate one or more clusters to bring together events whose characteristics are similar. One or more interactive displays provide a collaborative interface coupled to the extraction and the signalizer engine with a collaborative interface (UI) for decomposing events from the infrastructure. The events are converted into words and subsets to group the events into clusters that relate to security of the managed infrastructure. In response to grouping the events physical changes are made to at least a portion of the physical hardware. In response to production of the clusters security of the managed infrastructure is maintained.

The invention discloses an image retrieval method based on sparse non-negative matrix decomposition. It includes the following steps: 1) Query and extract the image and accompanying text of the retrieval results under two different image data sources; 2) Extract the tags in the accompanying text, and filter the results according to word frequency to form a vocabulary; 3) For each Image set, using the association relationship between labels and images to form an association matrix between labels and images; 4) Using sparse non-negative matrix decomposition to analyze the association matrix obtained in step 3), to obtain the shared subspaces of different sources of data and their corresponding independent subspaces Subspace; 5) The user sends a retrieval request for images on a data source, forms a query vector and maps it to the corresponding subspace of the data source, calculates the similarity with all images and sorts them, and returns the top N most similar images. The present invention makes full use of the associated knowledge of tags and images under multiple data sources, performs migration learning through sparse non-negative matrix decomposition, and improves the accuracy of image retrieval on target data sources.

A method detects components of a non-stationary signal. The non-stationary signal is acquired and a non-negative matrix of the non-stationary signal is constructed. The matrix includes columns representing features of the non-stationary signal at different instances in time. The non-negative matrix is factored into characteristic profiles and temporal profiles.

The invention discloses a voice denoising method and system based on L1 / 2 sparse constraint convolution non-negative matrix decomposition. In single-channel voice enhancement, it is assumed that noised voice signals v(i) are additively relevant to noise signals n(i) and voice signals s(i), i.e., v(i)=n(i)+s(i), and noise-base information is obtained by training specific noise by use of a CNMF method; and then by taking a noise base as prior information, a voice base is obtained by decomposing noised voice by use of a CNMF_L1 / 2 method, and finally, voice after denoising is synthesized. According to the method, correlation of voice between frames can be better described; and strong-sparse constraining is performed on a voice-base coefficient matrix by use of L1 / 2 regular item, and the voice after separation comprises less residual noise. Compared to conventional methods such as a spectral subtraction method, a wiener filtering method and a minimum mean square deviation logarithm domain spectrum estimation method and the like, the voice after enhancement can be understood more easily.

The invention discloses a video motion characteristic extracting method based on local sparse constraint non-negative matrix factorization. The video motion characteristic extracting method mainly solves the problems that static background interference and flash points of a video cannot be filtrated, the convergence rate is low, and the factorization error is over-serious in the prior art. The video motion characteristic extracting method comprises the steps of: firstly converting a video into a video frame group by taking a target frame as the center, and converting the video frame group into a non-negative matrix; next, factorizing the non-negative matrix by a local sparse constraint non-negative matrix factorization method, carrying out sparse constraint on part of base matrix column vectors, and calculating a motion vector of the target frame through weighted summarization of the part of the base matrix column vectors undergoing sparse constraint and the corresponding coefficient matrixes; and finally converting the motion vector of the target frame into the motion characteristic of the target frame. The video motion characteristic extracting method disclosed by the invention is applicable to target tracking and video monitoring, and can be used for extracting the video motion characteristic quickly, accurately and effectively.

The invention discloses a differentially expressed gene identification method based on combined constraint non-negative matrix factorization. The method comprises the following steps of 1, representing a cancer-gene expression data set with a non-negative matrix X, 2, constructing a diagonal matrix Q and an element-full matrix E, 3, introducing manifold learning in the classical non-negative matrix factorization method, conducting orthogonal-constraint sparseness and constraint on a coefficient matrix G, and obtaining a combined constraint non-negative matrix factorization target function, 4, calculating the target function, and obtaining iterative formulas of a basis matrix F and the coefficient matrix G, 5, conducting semi-supervision non-negative matrix factorization on the non-negative data set X, and obtaining the basis matrix F and the coefficient matrix G after iteration convergence, 6, obtaining an evaluation vector (the formula is shown in the description), sorting elements in the evaluation vector (the formula is shown in the description) from large to small according to the basis matrix F, and obtaining differentially expressed genes, 7, testing and analyzing the identified differentially expressed genes through a GO tool. The identification method can effectively extract the differentially expressed genes where cancer data is concentrated, and be applied in discovering differential features in a human disease gene database. The identification method has important clinical significance for early diagnosis and target treatment of diseases.

The invention belongs to the technical field of information processing and particularly relates to a nonnegative matrix factorization method based on low-rank recovery. The method comprises following steps: 1. each image sample in a raw database is converted into a vector to form an m*n original data matrix X, wherein m is the dimension of the image sample, n is the number of image samples; 2. low-rank sparse factorization is performed on the original data matrix X; 2.1 the rank of the low-rank matrix is set as r, and the sparseness of the sparse matrix is set as k; 2.2 a low-rank matrix L with the rank of r and a sparse matrix S with the rank of k of the original data matrix X is solved by means of the bilateral random projection algorithm; 3. nonnegative matrix factorization is performed on the low-rank matrix L obtained in step 2. to obtain a basis matrix W and an encoding matrix H. According to the nonnegative matrix factorization method based on low-rank recovery, data low-rank components and sparse components are obtained through low-rank sparse factorization and nonnegative matrix factorization is performed on the low-rank components removed of sparse noise parts to make the nonnegative matrix factorization results free from noise interference.

The present invention discloses a Non-negative Matrix Factorization (NMF) based on sparse and correlation constraints and the method is applied to processing of decomposition of mixed pixels of a hyper-spectral remote sensing image. According to the method, finally, a given non-negative matrix Vm*n is factorized into a product of a basis matrix Wm*r and a coefficient matrix Hr*n, i.e. Vm*n is approximately equal to Wm*r Hr*n ; firstly, the non-negative matrix V is selected, W and H are randomly initialized, then the minimum correlation constraint is applied to the coefficient matrix H in a target function, the sparse constraint is applied to the basis matrix W and the coefficient matrix H and then iteration is carried out according to an iteration formula until the matrices W and H are converged.

Documents of a set of documents are represented by bag-of-words (BOW) vectors. L labeled topics are provided, each labeled with a word list comprising words of a vocabulary that are representative of the labeled topic and possibly a list of relevant documents. Probabilistic classification of the documents generates for each labeled topic a document vector whose elements store scores of the documents for the labeled topic and a word vector whose elements store scores of the words of the vocabulary for the labeled topic. Non-negative matrix factorization (NMF) is performed to generate a document-topic model that clusters the documents into k topics where k>L. NMF factors representing L topics of the k topics are initialized to the document and word vectors for the L labeled topics. In some embodiments the NMF factors representing the L topics initialized to the document and word vectors are frozen, that is, are not updated by the NMF after the initialization.

The invention discloses a method for fusing a high-spectrum image and a visible light image based on nonnegative matrix decomposition. The method comprises the steps that the high-spectrum image is unmixed in a nonnegative matrix decomposition method based on sparsity constraint to obtain an end member spectrum curve; the mixed coefficient of each pixel in the image is solved by constrained quadratic programming according to the spectrum absorption curve of the visible light image and the end member spectrum curve obtained by unmixing; and the end member spectrum curve obtained by unmixing and the mixed coefficient of each pixel are linearly mixed to obtain the fused image of high spatial resolution and high high-spectrum images. The method disclosed by the invention solves the problem that spectra in some wave bands of the high-spectrum image are irrelevant to the visible light image. The high-spectrum image is unmixed by nonnegative matrix decomposition based on sparsity constraint to obtain a more accurate end member spectrum. At the same time the mixed coefficient of the pixel can be obtained by the constrained quadratic programming. The obtained fused image has high spatial resolution and high spectrum resolution.

Techniques for removing image autoflourescence from fluorescently stained biological images are provided herein. The techniques utilize non-negative matrix factorization that may constrain mixing coefficients to be non-negative. The probability of convergence to local minima is reduced by using smoothness constraints. The non-negative matrix factorization algorithm provides the advantage of removing both dark current and autofluorescence.

The invention discloses a method and a system for intrusion detection based on non-negative matrix factorization under sparse representation. The method includes: acquiring network data and host data, and obtaining a level-one audit privilege program of original network data; preprocessing the network data and the host data, and generating network characteristic data and short-sequence vectors; performing non-negative matrix iterative factorization for a data test matrix, and performing sparse representation for a basis matrix and a weight matrix; sampling weight matrix data subjected to sparse representation by the aid of a projection matrix so that highly characteristic weight coefficient vectors are obtained; and matching the highly characteristic weight coefficient vectors with characteristic vectors in training data by the aid of characteristic vector library data, and judging whether abnormal characteristics are conformed to or not. The method and the system for intrusion detection achieve data dimension reduction by non-negative matrix factorization and uses multi-divergence as a measurement level, an RIP (routing information protocol) condition in sparse representation is added into a combined divergence objective function family to restrain a non-negative matrix factorization iterative process, data detection dimensionality is lowered, and high-dimensional mass data processing of the system for intrusion detection is facilitated.

A color determination method utilizes color matching functions to approximate the imaging system's sensitivity characteristics. The illuminant conditions are modeled according to known illuminant intensity versus wavelength functions. Non-negative Matrix Factorization (NMF) is applied to a set of known reflectance data to decompose the known reflectance data set into a defined number of NMF basis vectors. In general, for an N-color based imaging system, N NMF basis functions are determined. Since basis functions provided by NMF are non-negative, the determined N NMF basis functions are related to actual physical colors. The NMF basis vectors are integrated with the illuminate conditions and color matching function(s) that approximate the imaging system's sensitivity to generate XYZ color values. These are converted to RGB values which are used to determine the optimal N colors for the N-color based imaging system.

The invention belongs to the field of data excavation and discloses a protein functional module excavating method for multi-view data fusion. The method comprises the following steps: firstly performing quantifying description on strong and weak interaction of multiple data sources on protein and forming multi-view data; further performing uniform matrix decomposition on the multi-view data by utilizing a polymerization nonnegative matrix algorithm provided by the invention; determining the functional module of the protein by virtue of obtaining the optimal approximation of the multi-view information. The protein functional module excavating method for multi-view data fusion, provided by the invention, aims at simultaneously analyzing multiple biodata and comprises gene coexpression, GO annotation and PPIN and can be used for extracting the protein functional module with the most consistent polymerization characteristic from the multi-view. The method disclosed by the invention is especially suitable for interaction networks and biodata of the protein and meanwhile can be applied to community excavation problems of social complex networks and communication networks.

The invention discloses an incremental learning and face recognition method based on locality preserving nonnegative matrix factorization (LPNMF), and relates to the technical field of pattern recognition. The incremental learning and face recognition method is essentially a face recognition method based on on-line learning. The method comprises the step (a) of face image preprocessing, the step (b) of initial sample training, the step (c) of incremental learning and the step (d) of face recognition. The incremental learning and face recognition method can be applied to a linear face recognition system and a non-linear face recognition system, keeps local structures of original space of face images, greatly improves recognition rate, and can be actually applied in an on-line mode.

The invention discloses a voice conversion method based on convolutive nonnegative matrix factorization. The method comprises the following steps: (1) training a transformation model through training data: carrying out time calibration and parameter decomposition of training voice data, analyzing STRAIGHT spectrum by using a convolutive nonnegative matrix factorization method, and analyzing pitch frequency of source voice and object voice; (2) converting new input voice based on a training model: carrying out parameter decomposition on source voice data A[c] to be converted by employing a STRAIGHT model, realizing sound channel frequency spectrum parameter conversion based on convolutive nonnegative matrix factorization, realizing conversion of the pitch frequency based on obtained mean value and variance in a training phase, and synthesizing voice after conversion, wherein the voice is voice after synthesis and conversion of the STRAIGHT spectrum S[Bc] which is obtained through conversion, the pitch frequency f[Bc] and original aperiodic component ap[Ac]. According to the invention, training effect of voice conversion is improved, and voice quality of conversion voice is improved.

The embodiment of the invention provides a comic exaggeration method, device and system for a human face digital image supporting posture deflection. The method comprises the following steps of: restoring the approximate depth information of a two-dimensional characteristic point of an input human face image according to a pre-established three-dimensional average face shape model; generating the three-dimensional shape of the input human face image according to the approximate depth information; performing three-dimensional exaggeration on the three-dimensional shape which is rotated to a front view according to a three-dimensional exaggeration rule in a pre-established nonnegative matrix decomposition based human face three-dimensional shape characteristic space to obtain a three-dimensional shape subjected to three-dimensional exaggeration; converting the three-dimensional shape subjected to three-dimensional exaggeration into the initial posture of the input human face image and projecting to generate a two-dimensional characteristic point subjected to three-dimensional exaggeration; and generating a human face image with a comic exaggeration effect according to the two-dimensional characteristic point subjected to three-dimensional exaggeration. By adopting the method, a vivid, humorous and variable comic exaggeration effect can be achieved on the human face image which is deflected by + / -40 degrees in a swinging way and deflected by + / -20 degrees in a pitching way.

A human behavior identification method based on non-negative matrix decomposition and a hidden Markov model comprises an off-line training stage of firstly pre-processing image of each kind of selected behavior sequence training data to obtain a total sample data matrix A of all training data, carrying out non-negative matrix decomposition (NMF) on the A to obtain a basic matrix W and a basic vector number r, and obtaining a characteristic matrix Ei of each kind of training behavior sequence according to the W and the A, and initializing the hidden Markov model (HMM) of each kind of training behavior sequence and respectively estimating an optimal parameter thereof; and an on-line identification stage of firstly pre-processing the image of the input behavior sequence to be identified to obtain an original matrix a of the behavior sequence, obtaining a characteristic matrix e according to the W and the a; and lastly, figuring up a likelihood of the behavior sequence to be identified and each kind of training behavior sequence to determine behavior types. In the invention, the human behavior identification rate is higher, and the automatic analysis ability of the human behavior applied to a real-time intelligent video monitoring system is improved.