82 results about "Dimensional reduction" patented technology

A method of identifying and localizing objects belonging to one of three or more classes, includes deriving vectors, each being mapped to one of the objects, where each of the vectors is an element of an N-dimensional space. The method includes training an ensemble of binary classifiers with a CISS technique, using an ECOC technique. For each object corresponding to a class, the method includes calculating a probability that the associated vector belongs to a particular class, using an ECOC probability estimation technique. In another embodiment, increased detection accuracy is achieved by using images obtained with different contrast methods. A nonlinear dimensional reduction technique, Kernel PCA, was employed to extract features from the multi-contrast composite image. The Kernel PCA preprocessing shows improvements over traditional linear PCA preprocessing possibly due to its ability to capture high-order, nonlinear correlations in the high dimensional image space.

The present invention discloses an object image detection method, which uses a coarse-to-fine strategy to detect objects. The method of the present invention comprises steps: acquiring an image and pre-processing the image to achieve dimensional reduction and information fusion; using a trained filter to screen features; and sequentially using a coarse-level MLP verifier and a fine-level MLP verifier to perform a neural network image detection to determine whether the features of the image match the features of the image of a target object. The present invention simultaneously uses three mainstream image detection methods, including the statistic method, neural network method and adaboost method, to perform image detection. Therefore, the present invention has the advantages of the rapidity of the adaboost method and the accuracy of the neural network method at the same time.

The present invention is suitable for the unmanned aerial vehicle field, and provides a system of matching somatosensory operation to realize virtual flight. The system comprises a wearable sensor, a remote controller, an unmanned aerial vehicle and a visual device, wherein an unmanned aerial vehicle controller is arranged inside the unmanned aerial vehicle, the unmanned aerial vehicle is equipped with a panorama camera, and an instruction converter is arranged inside the remote controller or the unmanned aerial vehicle. According to the present invention, the wearable sensor captures the human body action and converts the human body action into a corresponding flight instruction, after receiving the flight instruction, the unmanned aerial vehicle controller controls the motion track of the unmanned aerial vehicle, the panorama camera real-timely obtains images and finally transmits back to a virtual reality (VR) visual device, and the VR visual device carries out the three dimensional reduction and reconstruction on the images to realize the VR flight effects for players. According to the present invention, the wearable sensor can capture the fine action change of the human bodies, can determine the actions accurately, and converts the actions into the instructions capable of being identified by the unmanned aerial vehicle correspondingly in a protocol conversion manner, so that the system is very strong in flight interactivity, and brings the brand new experience for the unmanned aerial vehicle flight.

The invention discloses an airborne radar clutter suppression method, belonging to the technical field of radar signal processing and aiming at solving the problems of the current clutter suppression technology such as high sample requirement, large calculation amount and poor real-time performance. The implementation steps are as follows: firstly, carrying out clutter pre-processing on the received data according to the prior information of the clutter spectrum distribution to realize dimensional reduction of the space-time bidimensions and reduce the clutter freedom degree; then carrying out cascade connection low-rank adaptive processing and performing multiple filtering and multiple dimensional reduction of the space-time bidimensions on the echoed signal to realize the clutter suppression and target detection of the signals. Indicated by the results of computer simulation analysis and actually measured data processing, the invention has remarkably reduced sample requirement and calculation amount under the precondition of good performance and the filtering performance is greatly improved under the circumstance of equivalent sample requirement, thus the invention can be used in dynamic target detection.

The invention discloses a radar radiation source identification method based on phase noise unintentional modulation characteristic, relates to an identification method of a radar radiation source, and aims to solve the problem that the identification rate of an existing radiation source identification method based on phase noise is not high. The method disclosed by the invention comprises the following steps of analyzing the structure of a phase-lock frequency synthesizer in a radar transmitter system; building a model of phase noise generated by the phase-lock frequency synthesizer; calculating a bispectrum diagonal slice characteristic and a bispectrum non-diagonal slice characteristic; forming a characteristic matrix Y by using a bispectrum diagonal slice characteristic matrix A1 and a bispectrum non-diagonal slice characteristic matrix B1; performing PCA (Principal Component Analysis) dimensional reduction and building a type-known transmitter vector machine model; identifying a transmission signal of a type-unknown transmitter by utilizing the built vector machine model so as to realize the identification of a radar radiation source. The method disclosed by the invention is applicable to the identification of the radar radiation source.

The invention discloses a method for shooting and three-dimensional reduction and reconstruction. The method comprises the following steps of: 1, interior calibration: measuring an image distance and a focal distance of a digital camera (2) and parallelism of an optical shaft and a mechanical shaft; measuring an image distance and a focal distance of a light interference source and the parallelism of the optical shaft and the mechanical shaft; 2, exterior calibration: measuring parameters, including included angles X, Y and Z of axles X, Y and Z, of relative poses and absolute poses of a system and a distance between the digital camera (2) and the light interference source and focusing parameters; 3, shooting: shooting a shot object by the digital camera (2) under the irradiation of the light interference source; 4, image acquisition and image processing: extracting characteristic information of the surface of the shot object under grating mapping to realize output of multiple image formats; and 5, three-dimensional reconstruction: reading parameters in the interior calibration 1 and parameters in the exterior calibration 2, and calling the acquired image and the processed image in the step 4 to finish the stereoscopic matching of pixel points, so that the three-dimensional reconstruction and reduction on the pixel points is realized, and the precision and the speed of three-dimensional reduction and reconstruction are improved to a large extent.

The invention discloses a hypergraph and random forest (HG-RF)-based intrusion detection method, and belongs to the technical field of network intrusion detection. The method comprises the following steps: (1) carrying out data preprocessing; (2) carrying out feature screening to obtain a new feature subset; (3) calculating a Fisher score of each feature in each category, and carrying out descending-order arrangement; and (4) inputting a test sample set to a weighted random forest classifier to obtain final intrusion detection results of test samples. The method is based on a method of featurepreference, firstly carries out dimensional reduction processing on data, then carries out classification, improves intrusion detection speed and an accuracy rate of the classifier, and reduces a detection false-reporting rate.

In the present work, quantum clustering is extended to provide a dynamical approach for data clustering using a time-dependent Schrödinger equation. To expedite computations, we can approximate the time-dependent Hamiltonian formalism by a truncated calculation within a set of Gaussian wave-functions (coherent states) centered around the original points. This allows for analytic evaluation of the time evolution of all such states, opening up the possibility of exploration of relationships among data points through observation of varying dynamical-distances among points and convergence of points into clusters. This formalism may be further supplemented by preprocessing, such as dimensional reduction through singular value decomposition and / or feature filtering. Additionally, the parameters of the analysis can be modified in order to improve the efficiency of the dynamic quantum clustering processes.

The invention discloses a Chinese text categorization method based on a multi-hidden-layer extreme learning machine. A regularization extreme learning machine model is applied to a Chinese text categorization problem, and text is categorized by means of a model of the multi-hidden-layer extreme learning machine. The Chinese corpus of Fudan University is used as a training set and testing set of text categorization; operation such as pre-processing is conducted on text data, including encoding mode unification, word segmentation, removal of stop words, symbols and figures and the like; the text is represented by means of a spatial vector model, and a data set is transformed into a text matrix; the text is categorized by means of the multi-hidden-layer extreme learning machine, wherein the process includes text dimensional reduction, characteristic mapping and text categorization. Text dimensional reduction is to transform high-dimensional text data into low-dimensional text data which can be calculated. The characteristics of the text are mapped by a multi-hidden-layer result of the multi-hidden-layer extreme learning machine, and high-level characteristic representation is conducted. The text is categorized by the regularization extreme learning machine of the multi-hidden-layer extreme learning machine.

The invention relates to a high-speed camera calibration system and method. The high-speed camera calibration system comprises a calibration board, a laser ranging radar, a high-speed camera, an exposure time and dropped frame measurement device and an imaging distortion calibration device. According to the system and method, a high-precision picosecond laser ranging technology is adopted to be combined with a camera parameter calibration algorithm, the camera focus and the distortion coefficient are determined, a distortion matrix is optimized, a precise camera image-to-spatial position three-dimensional reduction algorithm is formed, and spatial calibration on camera imaging distortion is achieved; the adopted measurement reference clock frequency is 4-5 orders of magnitude higher than the measured high-speed camera frequency; acquisition and analysis equipment with the sampling rate of 25 GHz is adopted to acquire the exposure time of the high-speed camera, the high-precision exposure time and dropped frame rate are obtained through measurement, and a data basis is provided for test parameters acquired for calibrating the high-speed camera.

A medical ventilator capable of early detecting and recognizing types of pneumonia, a gas recognition chip, and a method for recognizing gas thereof are disclosed. The gas recognition chip of the medical ventilator comprises a sensor array, a sensor interface circuit, a stochastic neural network chip, a memory and a microcontroller. The sensor array receives a plurality of multiple types of gases to produce odor signals corresponding to each type of gas. The sensor interface circuit analyzes the odor signals to produce gas pattern signals corresponding to each type of gas. The stochastic neural network chip amplifies the differences between the gas pattern signals and performs dimensional reduction on the gas pattern signals to aid the analysis. The memory stores training data. The microcontroller performs a mixed gas recognizing algorithm to early detect and recognize the type of the pneumonia according to the gas training data.

The invention discloses a method for the dimensional-reduction four-channel sun-difference beam angle measurement of a phased array radar. The method includes the following steps that: the phased array radar is determined; sub-array division is performed on N array elements contained by the phased array radar, so that M sub-arrays and the number of array elements contained by each sub-array can be obtained; the sub-array-level sum weight, sub-array-level pitch difference weight, sub-array-level azimuth difference weight, sub-array-level double-difference weight and sub-array element-level weight of the optimized phased array radar are calculated, and the optimal dimension reduction matrix of the phased array radar is calculated; the directivity function of the sum beams of the phased array radar, the directivity function of the azimuth difference beams of the phased array radar, the directivity function of the pitch difference beams of the phased array radar and the directivity function of the double-difference beams of the phased array radar are calculated; and the final pitch directional angle of the target of the phased array radar and the final azimuth directional angle of the target of the phased array radar are estimated.

The invention discloses a cross-media searching method based on discrimination correlation analysis. The method comprises the following steps of establishing a cross-media training database, carrying out feature extraction, mean-value pretreatment and linear projection transformation sequentially for different modal samples, and setting a target function according to a projection space; solving the target function to acquire a linear projection vector; establishing a cross-media test database; sequentially performing the feature extraction and mean-value pretreatment for a target to be searched; utilizing the linear projection vector to perform the linear projection transformation for the feature data after the mean-value pretreatment; and calculating an Euclidean distance between two modal data projection vectors, listing the Euclidean distance in ascending order, and acquiring a cross-media searching result. Due to the adoption of the method, dimensional reduction can be effectively performed for the feature data, so that the feature data can be widely applied to other multi-modal work, for example the multi-modal biological feature recognition.

A method and system for simulating and analyzing the behavior of a structural component of a computerized model in response to a simulated event to determine an optimized shape for the component is disclosed. The shape is optimized using an implicit dimensional reduction rather than an explicit geometric replacement by discarding data of a 3D discretization that has little or no bearing on the performance of the component to a simulated event. The reduced dataset is then collapsed onto a lower dimension projection that is applied over a force vector that is representative of the simulated event to determine the behavior of the component to the simulated event. Optimization tools may then be used to modify the physical attributes of the component and performance of the component once again simulated until an optimized component is determined.

The invention discloses an abnormal event detection method based on low-rank adaptive sparse reconstruction, and the method comprises three processes: feature extraction, training, and detection, and specifically comprises the steps: 1), extracting multi-scale three-dimensional gradient features of a video sequence; 2), carrying out the dimensional reduction of the multi-scale three-dimensional gradient features, and forming a training feature set and a test feature set; 3), initializing the remaining training features and related parameters; 4), carrying out the iteration group sparse dictionary learning of the remaining training features, and obtaining a normal mode dictionary set and the low-rank information thereof; 5), carrying out the weighted sparse reconstruction of the test features through a group sparse dictionary set and the low-rank information, which are obtained in a training process; 6), judging whether the test features are abnormal features or not according to a reconstruction error. The method irons out defects of low-rank characteristics of video data which is not fully exploited and poor detection efficiency in the technology of abnormality detection.

The invention discloses a broadband signal source positioning method based on subspace weight sparse recovery. The method comprises steps that multiple broadband signals having unknown correlation in the space are irradiated to a linear sensor array formed by multiple isotropy sensors, and a complex observation matrix is acquired according to the linear sensor array; a target signal arrival angle is estimated under a compression sensing framework according to the complex observation matrix to carry out reconstruction to acquire a first signal model; truncated singular value decomposition of the complex observation matrix is carried out to construct a weight matrix; a weight vector is calculated according to the weight matrix; dimensional reduction is carried out according to the first signal model to acquire a second signal model; the second signal model is optimized to acquire a third presentation form of the target signal under the compression sensing framework; a spectral function is calculated according to the third presentation form of the target signal under the compression sensing framework to acquire an estimate of the target signal arrival angle. The method is advantaged in that decorrelation capability of a compression sensing algorithm and estimation precision of an MUSIC algorithm are realized.

The invention provides a three-dimensional joint estimation method based on Wi-Fi channel state information (CSI). Firstly, in order to overcome the limitation of the number of antennas and the channel bandwidth to a two-dimensional joint estimation model, the obtained channel state information forms a three-dimensional matrix from three dimensions of subcarriers, the antennas and data packets; secondly, dimensional reduction processing is performed on the three-dimensional matrix, and smooth processing between the subcarriers, the antennas and the data packets is performed on the basis; and finally, signal subspace decomposition and noise subspace decomposition are performed on the smoothed matrix to construct a spectral function. Based on the spectral function, three-dimensional joint search of an angle of arrival (AoA), time of flight (ToF) and doppler frequency shift (DFS) is performed. A designed three-dimensional joint estimation algorithm can still achieve high estimation precision under the conditions of a small number of antennas and a narrow channel bandwidth, and provides a theoretical basis for application such as accurate indoor tracking and positioning.

The invention discloses a nonnegative local coordinate factorization-based clustering method, which comprises the following steps that: (1) a sample characteristic matrix is built; (2) a low-dimensional sparse matrix is iteratively outputted; (3) and the low-dimensional sparse matrix is clustered and analyzed. A sparse coding concept is introduced into the nonnegative matrix factorization (NMF) process, nonnegative local coordinate factorization is undertaken on a high-dimensional sample characteristic matrix, so the factorized coefficient matrix is used as a low-dimensional expression of the high-dimensional sample characteristic matrix, and the low-dimensional matrix is clustered to analyze, so the clustering analysis is simple and valid; and at the same time, data after the dimensional reduction has good explanatory property. Compared with the dimensional reduction method of the prior art, the judgment capacity of the clustering analysis can be further improved.

The invention discloses a personal credit risk evaluation method and evaluation system for large-scale unbalanced credit data. The method comprises steps that firstly, according to the acquired high-dimensional large-scale unbalanced historical credit data, a training matrix X constructed by the acquired historical data set is constructed, and the adaptive elastic network characteristic selectionalgorithm is utilized to implement dimensional reduction for the given historical credit data set; and secondly, a training sample set after dimension reduction is divided into a few of sample classesand a plenty of sample classes, the membership degree of each sample is calculated according to the cluster-like centroid distance exponential attenuation function, a weight matrix W is constructed,relevant parameters are set, and an IWELM model is utilized to implement personal credit risk evaluation. The method is advantaged in that a high-degree unbalanced problem shown by the large-scale credit data is overcome, the speed and efficiency of personal credit risk evaluation are further improved, and reliability and credibility of the evaluation result are improved.

The present invention relates to a device and a method for extracting information from remotely detected characteristic signals. A data stream (24) derivable from electromagnetic radiation (14) emitted or reflected by an object (12) is received. The data stream (24) comprises a sequence of frames (66, 70, 92, 94; 108, 110, 112, 122, 124; 164, 66, 168), at least some of the frames comprise a frame section (68, 72; 174, 180, 186) 5 representative of a region of interest (56) attributable to the object (12). The region of interest (56) exhibits a continuous or discrete characteristic signal (136; 192) including physiological information (200) indicative of at least one at least partially periodic vital signal (20; 208). The sequence of frames (66, 70, 92, 94; 108, 110, 112, 122, 124; 164, 166, 168) further comprises a disturbing signal portion at least partially indicative of undesired object motion. 10 The characteristic signal (136; 192) can be stabilized by deriving a derivative signal form (78, 88, 98, 100, 102; 172, 176) from at least some frames of said sequence of frames (66, 70, 92, 94; 108, 110, 112, 122, 124; 164, 166, 168) through a dimensional reduction. A positional shift (74, 178) of a present derivative signal form (88, 98, 100, 102; 176) relative to a previous derivative signal form (78, 172) can be estimated. A present frame section (72; 180, 186) can be determined under consideration of the estimated positional shift (74, 178). Hence, the region of interest (56) can be tracked for at least partially compensating undesired object motion. Consequently, the disturbing signal portion can be at least partially compensated. The characteristic signal (136; 192) can be extracted from the sequence of frames (66, 70, 92, 94; 108, 110, 112, 122, 124; 164, 166, 168) under consideration of a sequence of determined frame sections (68, 72; 174, 180, 186).

The invention relates to a preparation method of a transition metal oxide / graphene composite material, a lithium ion battery negative pole and a lithium ion battery. The preparation method comprises the following steps: soaking three-dimensional reduction-oxidation graphene in a soaking solution containing transition metal salt, and carrying out predecomposition and roasting after freeze drying. The transition metal oxide / graphene composite material prepared in the invention is applied to lithium ion batteries, and has the excellent performances such as high capacity, long cycle life, low cost and mass production.

A sensor data processing system and method is described. Contemplated systems and methods derive a first recognition trait of an object from a first data set that represents the object in a first environmental state. A second recognition trait of the object is then derived from a second data set that represents the object in a second environmental state. The sensor data processing systems and methods then identifies a mapping of elements of the first and second recognition traits in a new representation space. The mapping of elements satisfies a variance criterion for corresponding elements, which allows the mapping to be used for object recognition. The sensor data processing systems and methods described herein provide new object recognition techniques that are computationally efficient and can be performed in real-time by the mobile phone technology that is currently available.

The invention provides a preparation method of a rGO / In2O3 composite room temperature gas-sensitive material with mesoporous structure cubic In2O3 grown on three dimensional reduction graphene oxide (rGO). The preparation method includes the following steps: using graphene oxide GO, tetrahydrate indium trichloride and urea as raw materials, and using trisodium citrate as a surfactant and a reductant for hydrothermal method and calcinations to obtain the omposite room temperature gas-sensitive material with the mesoporous structure cubic In2O3 uniformly distributed on a three dimensional rGO carrier overlapped into a pore structure. The production process is simple. The prepared mesoporous structure cubic In2O3 is uniformly distributed on the three dimensional rGO carrier overlapped into the pore structure, and the gas-sensitive detection material with sensitivity to trimethylamine gas at room temperature is obtained.

The invention relates to a preparation method of a nano-Pd hydrogenation catalyst loaded with three-dimensional reduction graphene oxide. The preparation method comprises the steps that a certain amount of PdCl2 and ascorbic acid are added into an aqueous solution of graphene oxide by taking the graphene oxide as a carrier precursor, Pd is uniformly deposited on the generated three-dimensional reduction graphene oxide through a one-step hydrothermal method, and then high-dispersity loaded nano-Pd particles are prepared. By applying the catalyst to a hydrogenation reaction of olefin, the very high catalytic activity and stability are achieved, and the high recycling rate is achieved; the catalyst is mild in preparation condition, simple in process and low in cost, and industrialized production is easy to achieve.

The invention relates to a method for mining data of an electronic nose based on supervised explicit manifold learning algorithm. The method for mining the data of the electronic nose through the explicit manifold learning algorithm comprises the following steps of collection of gas samples, characteristic extraction of the gas samples, determination of near neighbor of each point in a characteristic value matrix, relation calculation of any two characteristic value points and data dimension reduction of the explicit manifold algorithm. The data mining method of the electronic nose with the supervised explicit manifold learning algorithm comprises all above steps and is additionally provided with one step after the characteristic extraction of the gas sample: considering the type information, and determining the near neighbor of each point in the characteristic value matrix. The method has beneficial effects that the explicit manifold learning algorithm is used for reducing the dimension of the electronic nose data, and an explicit dimensional-reduction expression is provided; and the supervised manifold learning algorithm is used for reducing the dimension of the electronic nose data, the relation difference of each point of difference sources in the characteristic value matrix is considered, and the reservation of the detail information guarantees high resolution of an electronic nose system.

The invention discloses a VoIP voice quality objective evaluating method based on network parameters, and belongs to the voice quality evaluation research field. The method includes the following steps: obtaining network parameters during VoIP conversation; pre-processing the acquired network parameters, pre-processing including obtaining basic main statistics such as mean value, variance, extreme value, mode and median, wherein the network parameters mainly refer to a packet loss ratio, a jitter value and a time-delay value; performing PCA dimensional reduction on the preprocessed data; and finally inputting the data undergoing dimensional reduction to a multivariable linear regression model, and obtaining a voice quality objective evaluation score through calculation. In this way, evaluation of the VoIP voice quality based on the network parameters is conducted. The invention brings forward the new VoIP voice quality objective evaluating method, which can accurately evaluate the voice quality only by monitoring network parameter values with no need of obtaining voice signals.