281 results about "Canonical correlation" patented technology

This invention discloses a method for media-crossing searches based on content relativity, which applies the typical relativity analysis to analyze the content characters of different mode media data, maps a visual sense character vector of image data and an auditory character vector of audio data in a low dimension isomorphic sub-space simultaneously by a sub-space mapping algorithm, measures the relativities among different mode data based on a general distance function and modifies the topological structure of a multi-mode data set in the sub-space to increase the cross media search efficiency effectively.

The invention relates to a method for identifying facial expressions from a human face image sequence, belonging to the technical field of analyzing and identifying human facial expressions. The method of the invention comprises the following steps of: firstly, adopting a method for tracing feature points, sequentially extracting the displacement amount of the normalized facial key point and the length of the special geometrical characteristic for each frame image of the expression image sequence, and combining the data to form a characteristic column vector; secondly, sequentially arranging all characteristic column vectors of the sequence to form a characteristic matrix, wherein each characteristic matrix represents a facial expression image sequence; finally, comparing the similarities among the characteristic matrixes by using a canonical correlation analysis method, thereby determining the human face images to be identified into one of the basic expressions of happiness, sadness, fear, hate, surprise and anger. In the invention, the canonical correlation analysis method is successfully applied to identifying the human facial expressions, the dynamic information in the expression generation course is utilized effectively and the higher recognition rate and the shorter CPU computation time are acquired.

Provided is a brain activity training apparatus for training to cause a change in correlation of connectivity among brain regions, utilizing measured correlations of connections among brains regions as feedback information. From measured data of resting-state functional connectivity MRI of a healthy group and a patient group (S102), correlation matrix of degree of brain activities among prescribed brain regions is derived for each subject. Feature extraction is executed (S104) by regularized canonical correlation analysis on the correlation matrix and attributes of the subject including a disease / healthy label of the subject. Based on the result of regularized canonical correlation analysis, by discriminant analysis through sparse logistic regression, a discriminator is generated (S108). The brain activity training apparatus feeds back a reward value to the subject based on the result of discriminator on the data of functional connectivity MRI of the subject.

The invention belongs to the cross-media information technology field and particularly is an image optimization clustering method based on the typical correlation analysis. The invention mainly adopts the typical correlation to analyze while considering content characteristics of media data in various modes, maps the characteristics of the media data in various modes to an isomorphism sub-space of a united dimension through the sub-space mapping algorithm and obtains the final clustering result through optimizing clustering algorithm. The invention overcomes single-mode characteristic limitation in the multimedia field where only data is used, effectively solves the isomerism problem of the media data in various modes on the bottom layer characteristics, realizes the united measurement of the media object information between various modes, obtains results which are more accurate, more effective and more comforted to the needs in the large scale image data and has a wide application value in the cross-media processing and the retrieval field.

The invention discloses an automatic image annotation method based on deep learning and canonical correlation analysis. The method includes: using a depth Boltzmann machine to extract the high-level feature vectors of images and annotation words, selecting multiple Bernoulli distribution to fit annotation word samples, and selecting Gaussian distribution to fit image features; performing canonical correlation analysis on the high-level features of the images and the annotation words; calculating the Mahalanobis distance between to-be-annotated images and training set images in canonical variable space, and performing weighted calculation according to the distance to obtain high-level annotation word features; generating image annotation words through mean field estimation. The depth Boltzmann machine comprises I-DBM and T-DBM which are respectively used for extracting the high-level feature vectors of the images and the annotation words. Each of the I-DBM and the T-DBM sequentially comprises a visible layer, a first hidden unit layer and a second hidden unit layer from bottom to top. By the method, the problem of 'semantic gap' during image semantic annotation can be solved effectively, and annotation accuracy is increased.

Provided is a mixing method for a brain-computer interface based on SSVEP and OSP. A subject wears an electrode cap. A SSVEP-OSP mixed paradigm is broadcast in front of the subject by means of a computer screen. The subject stares at any one of simulation units. By a collection system, an electroencephalogram signal generated when the subjects stares at a simulation target is magnified, filtered and subjected to analog-digital conversion by an electroencephalogram acquisition instrument. Digitized electroencephalogram data is inputted into a computer. An electroencephalogram signal feature extraction method based on a typical correlation analysis is adopted for extraction, classification and recognition of features of SSVEP. A support vector machine and naive bayesian algorithm are adopted for extraction and recognition of OSP features. A recognition result is displayed on the screen in order to feed back to the subject. Then neat recognition is carried out. The mixing method for the brain-computer interface based on SSVEP and OSP has following advantages: rate of information transmission of the method for the brain-computer interface is increased based on SSVEP; and the method is easy in operation, few in electrode number and many in target number.

An image data processing system has a learning storage apparatus that stores projection matrixes obtained by canonical correlation analysis so as to derive, based on at least one of an image feature and a word feature, a latent variable as an abstract concept used for associating an image with a word corresponding thereto and that further stores information required for obtaining the latent variable acquired by use of the projection matrixes, a probability of occurrence of an arbitrary image feature from a certain latent variable and a probability of occurrence of an arbitrary word feature from a certain latent variable. In this way, a probability of the image feature and word feature being simultaneously outputted can be easily and quickly determined, thereby executing a high-speed annotation or retrieval with high precision.

Disclosed is a video retrieval method based on multi-core canonical correlation analysis. The method includes grasping text descriptions corresponding to the video on internet, and then operating on the video: firstly dividing the video according to whether a shot is mutated or not, extracting key frames, extracting vision features of the key frames and moving features of the shot to form video feature vectors, and extracting word-frequency features from the text descriptions of each video; then utilizing the method of the multi-core canonical correlation analysis to obtain mapping matrixes and low-dimensional representation of the video features and the word-frequency features, and allowing the mapping matrixes and the low-dimensional representation to have the maximum correlation in low-dimensional space; finally, when a user inputs key words to perform video retrieval, acquiring the low-dimensional representation of the word-frequency features of the key words according to the mapping matrixes of the word-frequency features, and returning video retrieval results sequentially from large to small of the degrees of cosine similarity with the low-dimensional representation of the video features. The method has the advantages that the correlation of video content and the retrieval key words is enhanced, and the accuracy of retrieval by the user is improved.

Provided is a method of analyzing brain activities for realizing a biomarker for neurological / mental disorder, based on brain function imaging. From measured data of resting-state functional connectivity MRI of a healthy group and a patient group, correlation matrix (80) of degree of brain activities among prescribed brain regions is derived for each subject. Feature extraction is executed by regularized canonical correlation analysis (82) on the correlation matrix (80) and attributes of the subject including a disease / healthy label of the subject. Based on the result of regularized canonical correlation analysis, by discriminant analysis (86) through sparse logistic regression, a discriminator (88) is generated.

The invention discloses a brain function network modeling method for resting state synchronization EEG-fMRI, and relates to the research field of neural signal processing. The brain function network modeling method includes the steps that firstly, EEG signal preprocessing is conducted, a tape limit energy signal is extracted and a regression item is established; secondly, fMRI signal preprocessing is conducted, and a BOLD signal of each brain region is extracted; thirdly, the BOLD signal, extracted through fMRI, of each brain region and the regression item obtained from EEG are subjected to principal component analysis; fourthly, canonical correlation analysis is conducted on principal components of the two signals obtained from the last step; fifthly, modeling is conducted on a brain function network of resting state synchronization EEG-fMRI. By means of the brain function network modeling method, a plurality of the brain function networks can be obtained; the brain function networks and existing research results have high coherence so that the effectiveness of the modeling method can be proved, and thus a new thought and a scheme are provided for the resting state brain function research.

The invention discloses a face image super-resolution reconstructing method based on Canonical Correlation Analysis (CCA), which taking principal constituent analysis coefficients of a high resolution face image and a low resolution face image as two different variables with different dimensions and utilizing the CCA to extract relevant subspaces of the high resolution face image and the low resolution face image, thereby enhancing the consistency of a topological structure inside a high and low resolution image data set. In the relevant subspaces, the face image super-resolution reconstructing method can reconstruct a principal constituent analysis coefficient of a corresponding high resolution image to an input low resolution image by utilizing adjacent domain reconstructing idea, thereby reconstructing an entire face image. The invention divides a residual image into seven squares, compensates detail information by using the CCA method and the adjacent domain reconstructing idea and reconstructs the high resolution residual image. The final high resolution face image is the addition of an entire face image and the high resolution residual image.

The invention belongs to the field of power quality managing, and discloses a method for applying the partial least squares regression to power distribution network harmonic source positioning and detecting. According to a model for carrying out equivalent transformation on the electrical power system side and the distortion user side and a harmonic voltage and a harmonic current signal which are obtained through synchronous measurement at the common coupling joint, a regression coefficient is dissolved through the partial least squares regression algorithm, an aggregative variable with the best explanatory capability for a dependent variable is extracted in the manner of decomposing and screening data information in the system, and information and noise in the system are identified. The method integrates the basic functions of multiple linear regression, the canonical correlation analysis and the principal component analysis, the modeling predicting type data analysis method and non-model type data recognition are organically combined, regression modeling, data structure simplifying and variable correlation analyzing can be achieved at the same time, and the method can be widely applied to power quality analyzing, monitoring, evaluating and controlling fields.

The invention relates to a quality evaluation model for natural ageing of redried strips and a building method thereof and belongs to the technical field of tobacco processing and quality evaluation. The building method includes: according to factors influencing quality changes during natural ageing of tobacco, determining a major quality evaluation index, performing screening by means of quantitative characteristic analysis, factor analysis, linear correlation and canonical correlation analysis to obtain quality evaluation indexes highly related to tobacco sensory evaluation quality, subjecting the quality evaluation indexes to factor analysis to obtain factor score coefficient matrixes, subjecting scores of common factors and the sensory evaluation quality to principal component regression to obtain a standard equation, substituting the score coefficient matrixes obtained by factor analysis, to the standard equation so as to obtain the equality evaluation model for the natural ageing of the redried strips. The method and the model are scientific and effective and good in stability, and index detection related is simple and fast and has small errors.

The invention discloses a method for recognizing human face based on typical correlation analysis spatial super-resolution. Aiming at the problem of the low recognition efficiency of a low-resolution face image, the invention provides a method for obtaining the recognition feature of the low-resolution face image in a high-resolution space by using the super-resolution reconstruction of a recognition feature. Based on a manifold learning theory, the recognition features of the high-resolution and low-resolution face images are considered to be generated by a common internal structure, so the method comprises the following steps: enhancing the consistency of neighborhood relationships of the recognition features of the high-resolution and low-resolution face images by using typical correlation analysis so as to better meet the hypothesis on a neighborhood reconstruction concept; and utilizing neighborhood reconstruction to obtain the recognition feature of the high-resolution face image corresponding to the tested low-resolution face image in a related subspace obtained by the transformation of the typical correlation analysis, and finally utilizing the feature to recognize a face. Experiments show that a recognition rate obtained by the method is less influenced by the resolution of the face image and is relatively higher.

The invention discloses a facial expression recognition method based on a rough set and mixed features. The method includes the following steps that (1) facial detection is carried out; (2) local geometric distortion features are extracted by means of a method combining an active appearance model and the rough set; (3) overall features of expressions are extracted with the combination of an improved weighting primary component analysis method and the rough set; (4) feature fusion is carried out on the extracted local geometric distortion features and the overall features by means of kernel canonical correlation analysis under a high-dimensional small sample to eliminate feature redundancy, and fused typical features are obtained; (5) the fused typical features serve as observation vectors of a discrete HMM for classification and recognition, and a classification result is obtained. As is presented by experiments, the improved method can shorten facial expression recognition time and improve the facial expression recognition rate.

The invention discloses a cross-modal retrieval algorithm based on mixed hypergraph learning in subspace, based on the cross-modal public subspace learning of canonical correlation analysis. The similarities inside and among modals are calculated by a mapping of the public subspace. A mixed relation matrix is calculated by the similarities inside and among the modals. A mixed hypergraph model is built by the extraction of the relation matrix. And the cross-modal retrieval and sample sequencing are conducted by adoption of hypergraph learning finally. With aiming at cross-modal heterogeneous variations and high order relation among samples, the algorithm instance applies the hypergraph model combined with the cross-modal public subspace learning to the cross-modal retrieval, so that the model is capable of considering the similarity among the modals and the similarity inside the modals simultaneously and giving consideration to the high order relation among a plurality of the samples meanwhile, improving the final precision ratio and the final recall ratio of the cross-modal retrieval. The algorithm is capable of effectively improving the performance of the cross-modal retrieval and greatly enhancing the precision ratio and the recall ratio of the cross-modal retrieval.

The invention discloses an array antenna based multiple local discharge point positioning method for a local discharge detection device. The array antenna based multiple local discharge point positioning method comprises the steps of establishing a position equation related to local discharge points according to time difference of local discharge signals received by detection antennae and calculating direction angles of the local discharge points; enabling a master control computer to control rotation of cloud platforms of the detection antennae according to the input direction angles till the detection antennae receives the local discharge signals at the same moment; controlling shooting of a video camera, accordingly obtaining images of the local discharge signals and achieving positioning of the local discharge signals. The array antenna based multiple local discharge point positioning method for the local discharge detection device adopts a canonical correlation technology to estimate the number of local discharge sources; adopting a multiple signal classifying algorithm and utilizing covariance matrixes of the received data to separate out local discharge spaces and noise sub-spaces, utilizing direction vectors of the local discharge signals and orthogonality of the noise sub-spaces to form a space angle spectrum, wherein the spectrum peak positions shows the orientations of target signals.

The invention discloses a model training method. The method comprises the steps that training batch data is acquired, the training batch data comprises a plurality of training samples, for each training sample, identity features and age features corresponding to sample images are acquired through a feature extraction module, and typical correlation mapping variables corresponding to the features are acquired through a typical correlation mapping module; based on the typical correlation mapping variables corresponding to the training samples, counting correlation coefficients corresponding to the training batch data through a batch correlation coefficient measurement module; performing joint supervised learning through an identity discriminator and an age discriminator, performing decorrelation training of identity features and age features through correlation coefficients, and performing training to obtain a feature extraction module meeting a training target. And based on the decorrelation-based adversarial learning, the correlation between the identity features identified by the feature extraction module and the age features is extremely low, so that the identification precisionis improved. The invention further discloses a cross-age face recognition method, a corresponding device, equipment and a medium.

The invention discloses a method for fusing a significant structure and a relevant structure of the characteristics of an image. The method for fusing the significant structure and the relevant structure of the characteristics of the image comprises the steps of extracting the HOG characteristic and the LBP characteristic of the image, measuring the significant structure of the image characteristics inside sample sets, measuring the relevant structure of the image characteristics between the sample sets, and conducting fusion and mapping of the significant structure and the relevant structure. According to the method for fusing the significant structure and the relevant structure of the characteristics of the image, the HOG characteristic and the LBP characteristic of the image are extracted firstly, the significant structure of the image characteristics inside the sample sets is measured through x<2> measurement, the relevant structure of the image characteristics between the sample sets is measured through canonical correlation, and finally the structures are fused through a matrix spectrum optimization solution method, so that a fused characteristic set is obtained. By the adoption of the method for fusing the significant structure and the relevant structure of the characteristics of the image, the problem that in the prior art, the significant structure and the relevant structure of the multiple characteristics of the image can not be fused is solved, and structural fusion characteristics high in discrimination capacity are obtained.

The invention relates to a related prediction model-based method for detecting structural deformation in a magnetic resonance image, which is technically characterized in that: a group of normally tested triangular cerebral cortex surfaces which are registered to a standard template is utilized to calculate the structural dependence between the vertex on the group of surfaces and other vertexes; the structural dependence and typical related prediction models are utilized to predict an expected position of the vertex existing in a cerebral atrophy area according to a vertex position of a normal area without structural deformation of the brain; the vertex position obtained by the prediction models is compared with the vertex position before prediction to quantize the deformation on the cerebral cortex surface due to the cerebral atrophy so as to quantize the degree of the structural deformation of the brain resulted from the cerebral atrophy. Compared with other methods, the method has the main advantage that: the method can detect the presence of the structural deformation of the brain and quantize the degree of the structural deformation under the condition that only a single time-point magnetic resonance structure image is available.

The invention relates to an Alzheimer's disease multi-classification diagnosis system based on deep study. The Alzheimer's disease multi-classification diagnosis system comprises an image characteristic extracting module, an index characteristic selecting module, a vector linear merging module and a disease classification and diagnosis module, wherein the image characteristic extracting module is used for extracting characteristic vectors of a cerebral three-orthogonal plane MRI image according to a neural network model; the index characteristic selecting module is used for selecting checking indexes according to medical pertinent literatures to form index characteristic vectors; the vector linear merging module is used for adopting a multivariate data linear merging method based on canonical correlation analysis to merge the characteristic vectors of the image and the index characteristic vectors; and the disease classification and diagnosis module is used for inputting the merged vectors to a multi-classification classifier to distinguish the three stages of the Alzheimer's disease. The Alzheimer's disease multi-classification diagnosis system disclosed by the invention can assist the multi-classification diagnosis of the Alzheimer's disease.

The invention provides a low-resolution human face recognition method based on sparse maintaining canonical correlation analysis. The invention combines with the sparsity and canonical correlation analysis idea, and proposes the low-resolution human face recognition method based on sparse maintaining canonical correlation analysis. The method meets the requirements of maximum correlation of the extracted features through employing the canonical correlation analysis idea, achieves the fusion of the high and low resolution human face feature discrimination information, employs the sparsity idea to maintain the structural information, and improves the robustness of high and low resolution human face recognition. The method achieves the effective fusion of the high and low resolution human face feature discrimination information, improves the feature representation and discrimination capability, and meets the requirements of correlation and structural information maintaining.

The invention relates to an automatic lip language identification system suitable for Chinese language, comprising a wear-type camera, a man-machine interaction module, a lip contour positioning module, a geometric vector acquisition module, a motion vector acquisition module, a characteristic matrix building module, a transformation matrix T acquisition module, a conversion characteristic matrix acquisition module, a memory A, a memory B and a canonical correlation discriminatory analysis module. The wear-type camera is used for recording Chinese character sound image sequences, transmitting the Chinese character sound image sequences to the lip contour positioning module through the man-machine interaction module, and detecting and tracking lip contours by utilizing a convolution virtual electrostatic field Snake module; the geometric vector acquisition module and the motion vector acquisition module respectively extract geometric and motion characteristics from the lip contours and join up the geometric and motion characteristics as an input characteristic matrix of the canonical correlation discriminatory analysis module; and the canonical correlation discriminatory analysis module calculates the similarity among the characteristic matrixes and acquires identification results after processing. Compared with the traditional lip language identification systems, the system has higher identification accuracy.

The invention discloses an improved canonical correlation analysis-based physiological signal fusion identity recognition method, and mainly solves the problem of relatively low identification rate of an existing method. The method is implemented by comprising the steps of 1) obtaining pulse waves and respiratory signals, and preprocessing training data; 2) intercepting waveforms of the training data, and obtaining a pulse wave training set and a respiratory training set; 3) calculating intra-class and inter-class neighborhoods of the two sets; 4) according to the intra-class and inter-class neighborhoods, calculating intra-class and inter-class correlation matrixes of the two sets and constructing a regular canonical correlation analysis objective function; 5) solving regular canonical correlation analysis-based pulse wave and respiratory conversion matrixes; 6) calculating training fusion eigenvectors by utilizing the conversion matrixes; 7) obtaining pulse wave and respiratory signal test data and calculating test fusion vectors; and 8) performing type judgment on the test fusion vectors to obtain an identity recognition result. According to the method, the identity recognition rate is increased; and the method can be applied to the e-commerce and remote medical identity authentication.

The invention discloses an image annotation method and system based on a weak matching probability canonical correlation model, relating to the technical field of processing of network cross-media information. The image annotation method comprises the following steps: obtaining an annotated image and a non-annotated image in an image database, respectively extracting image features and textual features of the annotated image and the non-annotated image, and generating a matched sample set and an unmatched sample set, wherein the matched sample set contains an annotated image feature set and an annotated textual feature set; and the unmatched sample set contains a non-annotated image feature set and a non-annotated textual feature set; training the weak matching probability canonical correlation model according to the matched sample set and the unmatched sample set; and annotating an image to be annotated through the weak matching probability canonical correlation model. According to the invention, correlation between a visual modality and a textual modality is learned by using the annotated image, keywords of the annotated image and the non-annotated image simultaneously; and an unknown image can be accurately annotated.

The invention performs frequency estimation over both the burst preamble, during which known symbols are transmitted, and also during the burst's data packet, which is subsequent to the preamble and extracted by the local detector. During the preamble, an initial frequency estimate is obtained. This estimate is based on a time average of either phase or correlation samples. Atypical phase or correlation samples, attributable to detector symbol errors during the data packet, are detected and filtered, so as to avoid including the atypical samples in a time-averages used to provide the frequency estimate. In a first embodiment correlation samples are time averaged, and atypical correlation samples are suppressed prior to correlation time averaging. In a second embodiment, phase slope values are time averaged, and atypical values of phase slope are suppressed prior to phase slope time averaging.

The invention discloses a steady visual induction paradigm design and identification method for introducing a continuous action of an object. During paradigm design, one continuous action of the object is introduced, the continuous action of the object is disintegrated into a plurality of pictures, continuous black and white flashing is performed according to a time sequence, dynamic picture flashing units are formed, the dynamic picture flashing units are displayed on a computer to stimulate a user to generate a steady visual induction potential, a brain wave signal of the user staring at a certain flashing unit is collected and transmitted to the computer for processing, a canonical correlation analysis algorithm based on personal characteristic correction is adopted to perform target identification, the result is fed back to the user, the object is controlled to complete a corresponding action at the same time, and then target identification is performed for the next time. Through the method, corresponding steady visual induction paradigms can be designed according to different application objects, the control effect of "what you see is what you obtain" is achieved, the difference between user individuals is relieved, the correct rate of identification is increased, and the method can be widely applied to all types of brain-computer interface systems based on steady visual induction paradigms.

The invention discloses a steady motion evoked potential brain-computer interface method based on equal-brightness color strengthening. An equal-brightness color checker stimulation normal form is adopted, the contraction-expansion motion of checkers is achieved in a sine mode; an electroencephalogram generated when a user stares a stimulating pattern is collected through electroencephalogram collecting equipment and subjected to amplifying, filtering and A / D conversion, then the processed electroencephalogram is input into a computer, and feature extraction and classification and recognition of the collected electroencephalogram are achieved through canonical correlation analysis. According to the method, the color, brightness, the shape, motion and other visual information are integrated, the visual fatigue of the user is relieved, the signal-to-noise ratio of the EEG is increased, the identification accuracy of the stare target is higher, the method has the advantages of low flickering and high adaptability, and the interaction performance of a brain-computer interface can be improved.

The present invention discloses a method for eliminating functional magnetic resonance data noise based on independent component space correlation. Said method includes the following steps: firstly, making image segmentation of gray matter region and cerebrospinal fluid region of image; respectively making main component analysis and frequency spectrum analysis; defining and eliminating random noise component in the main component; reconstructing gray matter data and cerebrospinal fluid data after the random noise is eliminated; making independent component decomposition of the reconstructed data; utilizing independent component to respectively construct matrix; then making typical related analysis and sequencing, and zero-setting these components so as to obtain a group of new independent components; reconstructing data after various noise components are eliminated, and repeating the above-mentioned steps until the related maximum various noise components in various layers of gray matter data and cerebrospinal fluid data are eliminated.