213results about How to "Accurate features" patented technology

A sound reproducing apparatus and a sound reproducing method. The sound reproducing apparatus includes an actual listening environment feature function database where an actual listening space feature function is stored for correcting the virtual source in response to a feature of an actual listening space provided at the time of listening; and an actual listening space feature correcting unit of reading out the actual listening space feature function stored in the actual listening environment feature function database, and correcting the virtual source based on the reading result. Accordingly, causes of each distortion may be removed to provide sounds having the best quality.

An apparatus for ultrasound irradiation of a body part (208) includes a first ultrasound transducer (216), and a second ultrasound transducer (212) mounted oppositely, and is configured: a) such that at least two ultrasound receiving elements, for determining a relative orientation of the first to the second transducer, are attached to the first transducer; b) for a beam, from the first transducer, causing cavitation, and / or bubble destruction of systemically circulating microbubbles, within the body part; or c) both with the attached elements and for the causing. The apparatus registers, with said first transducer, the second transducer, by using as a reference respectively the features a) and / or b). A detachable subsystem includes either a therapy or imaging transducer, to form a combination imaging and therapy system, the subsystem being configured for removable coupling to correspondingly an imaging or therapy subsystem comprising a head frame (204) to which is mounted an imaging or therapy transducer, the imaging or therapy subsystem for registering the imaging or therapy transducer, one to the other, by the above method.

A keyword extracting device includes high-frequency term extracting means (30) for extracting high-frequency terms which are index terms having a great weight among the index terms in a document group (E) including a plurality of documents (D), the weight including evaluation on the level of an appearance frequency of each index term, clustering means (50) for clustering the high-frequency terms on the basis of a co-occurrence degree C. which is based on the presence / absence of the co-occurrence of each document with the index terms (w) in the document group (E) in each document, score calculating means (70) for calculating a score key(w) of each index term (w) such that a high score is given to the index term among the index terms (w) that co-occurs with the high-frequency term belonging to more clusters (g) and that co-occurs with the high-frequency term in more documents (D), and keyword extracting means (90) for extracting keywords on the basis of the scores. Accordingly, the keywords indicating a feature of a document group including a plurality of documents can be automatically extracted.

A solid imaging apparatus and method employing sub-pixel shifting in multiple exposures of the digitally light projected image of a cross-section of a three-dimensional object on a solidifiable liquid medium. The multiple exposures provide increased resolution, preserving image features in a three-dimensional object and smoothing out rough or uneven edges that would otherwise be occur using digital light projectors that are limited by the number of pixels in an image projected over the size of the image. Algorithms are used to select pixels to be illuminated within the boundary of each image projected in the cross-section being exposed.

A solid imaging apparatus and method employing levels of exposure varied with gray scale or time or both of digitally light projected image of a cross-section of a three-dimensional object on a solidifiable photopolymer build material. The gray scale levels of exposure of projected pixels permits the polymerization boundaries in projected boundary pixels to be controlled to achieve preserved image features in a three-dimensional object and smooth out rough or uneven edges that would otherwise occur using digital light projectors that are limited by the number of pixels in an image projected over the size of the image. Software is used to control intensity parameters applied to pixels to be illuminated in the image projected in the cross-section being exposed in the image plane.

A system and method is provided that employs a frequency domain interpolative CODEC system for low bit rate coding of speech which comprises a linear prediction (LP) front end adapted to process an input signal that provides LP parameters which are quantized and encoded over predetermined intervals and used to compute a LP residual signal. An open loop pitch estimator adapted to process the LP residual signal, a pitch quantizer, and a pitch interpolator and provide a pitch contour within the predetermined intervals is also provided. Also provided is a signal processor responsive to the LP residual signal and the pitch contour and adapted to perform the following: provide a voicing measure, where the voicing measure characterizes a degree of voicing of the input speech signal and is derived from several input parameters that are correlated to degrees of periodicity of the signal over the predetermined intervals; extract a prototype waveform (PW) from the LP residual and the open loop pitch contour for a number of equal sub-intervals within the predetermined intervals; normalize the PW by a gain value of the PW; encode a magnitude of the PW; and directly quantize the PW in a magnitude domain without further decomposition of the PW into complex components, where the direct quantization is performed by a hierarchical quantization method based on a voicing classification using fixed dimension vector quantizers (VQ's).

A method and system for limiting the motion of components such as the optics of a camera are disclosed. The system can comprise a stage and a snubber assembly for controlling motion of the stage in six degrees of freedom. For example, the snubber assembly can permit movement in one translational degree of freedom while substantially limiting motion in the other five degrees of motion so as to facilitate focusing and / or zooming of a camera while inhibiting misalignment of the optics and while providing some protection against shock and vibration. Such motion control can be achieved while mitigating costs associated with precision manufacturing of the snubber assembly.

A sound reproducing apparatus and a sound reproducing method. The sound reproducing apparatus includes an actual listening environment feature function database where an actual listening space feature function is stored for correcting the virtual source in response to a feature of an actual listening space provided at the time of listening; and an actual listening space feature correcting unit of reading out the actual listening space feature function stored in the actual listening environment feature function database, and correcting the virtual source based on the reading result. Accordingly, causes of each distortion may be removed to provide sounds having the best quality.

The invention discloses an image detection method and system based on a convolutional neural network, and the method comprises the following steps of collecting and obtaining a preset number of labeled sample images, and carrying out the preprocessing of the labeled sample images, and obtaining a denoised target area image with a label; performing feature extraction on each labeled target area image to obtain a 3D feature matrix of each labeled target area; training the 3D convolutional neural network weak entity detection model by using all the obtained 3D feature matrixes to obtain a trained3D convolutional neural network weak entity detection model; and performing preprocessing and feature extraction on the to-be-detected picture to obtain a 3D feature matrix of the to-be-detected picture, inputting the 3D feature matrix of the to-be-detected picture into the trained 3D convolutional neural network weak entity detection model, and outputting a detection result of the to-be-detectedpicture, so that the weak entity target detection can be completed more accurately and quickly.

The invention discloses a method for identifying benign and malignant lung nodules based on multi-dimensional information. The method comprises the following steps: 1, representing a three-dimensional nodule in a two-dimensional way; 2, building a feature model; 3, building a fuzzy classifier; 4, evaluating the classifying performance. The method has the beneficial effects that accurate feature modeling is crucial to the identification of benign and malignant lung nodules. More objective bases are laid for the identification of the benign and malignant lung nodules by adopting imaging diagnosis features, common shapes and textual features in image processing, and patient information, feature extraction is performed on a two-dimensional image generated on the basis of a helical scanning technology, and a novel method is adopted for feature modelling, so that the extracted features are more accurate. In the method, a fuzzy C-means (FCM) clustering algorithm is adopted for identifying benign and malignant status of a suspected nodule, and a probability indicating the suspected nodule is benign or malignant is given, so that the method is more accordant with the thinking mode of a doctor.

The invention discloses a remote sensing image processing method combined with the shape self-adaption neighborhood and the texture feature extraction for image preprocessing. The method includes subjecting compressed image to a gray level co-occurrence matrix calculation; subjecting the generated gray level co-occurrence matrix to S coefficient modification of an SAN (Storage Area Networking) irregular object window to obtain a regular matrix; calculating a new co-occurrence matrix according to the modified regular matrix and selecting texture descriptors with obvious feature and low correlation; extracting texture feature map in the SAN irregular images; and calculating to obtain accurate images with combination feature which is overall comprehensive feature of neighborhood. According to the method, the overall classification accuracy based on a shape self-adaption neighborhood method can be improved by 4%. The method can not only extract the texture feature in the SAN irregular images of remote sensing images completely, but also process the extraction of mixed pixel feature of the fuzzy edge of earth surface objects, and is applicable to texture extraction of earth surface objects in natural states.

The invention relates to a micro-expression recognition method based on a space-time appearance movement attention network, and the method comprises the following steps: carrying out the preprocessingof a micro-expression sample, and obtaining an original image sequence and an optical flow sequence with a fixed number of frames; constructing a space-time appearance motion network which comprisesa space-time appearance network STAN and a space-time motion network STMN, designing the STAN and the STMN by adopting a CNN-LSTM structure, learning spatial features of micro-expressions by using a CNN model, and learning time features of the micro-expressions by using an LSTM model; introducing hierarchical convolution attention mechanisms into CNN models of an STAN and an STMN, applying a multi-scale kernel space attention mechanism to a low-level network, applying a global double-pooling channel attention mechanism to a high-level network, and respectively obtaining an STAN network added with the attention mechanism and an STMN network added with the attention mechanism; inputting the original image sequence into the STAN network added with the attention mechanism to be trained, inputting the optical flow sequence into the STMN network added with the attention mechanism to be trained, integrating output results of the original image sequence and the optical flow sequence through the feature cascade SVM to achieve a micro-expression recognition task, and improving the accuracy of micro-expression recognition.

The invention discloses a mosquito-borne infectious disease epidemic situation prediction method and system based on a gradient boosting tree. The mosquito-borne infectious disease epidemic situationprediction method based on a gradient boosting tree includes the steps: widely collecting various factor data influencing the mosquito-borne infectious disease; cleaning the data influencing the mosquito-borne infectious disease so as to perform importance ordering on the factors influencing the mosquito-borne infectious disease, on the basis of the gradient boosting tree; according to the selected important factors influencing the mosquito-borne infectious disease, establishing a mosquito-borne infectious disease epidemic situation prediction model based on Poisson regression; by means of theselected factor and the correlation coefficients of the mosquito-borne infectious disease, initializing the prediction model, and then determining the mosquito-borne infectious disease prediction model parameters by means of S fold cross-validation; and by means of a epidemic situation hot spot graph based on geographical information and an epidemic situation outbreak graph based on a time axis,visually displaying the model prediction result. The mosquito-borne infectious disease epidemic situation prediction method and system based on a gradient boosting tree apply the gradient boosting tree and other machine learning methods to the field of mosquito-borne infectious disease epidemic situation prediction, can improve the mosquito-borne infectious disease epidemic situation prediction accuracy, can assist disease control staff to predict the mosquito-borne infectious disease epidemic situation in advance, and can timely take the corresponding measures to control large scale outbreakof the infectious disease.

The invention discloses a pedestrian re-identification model, method and system for adaptive difficulty mining. The identification method comprises the steps of: randomly dividing sample pictures intoa training set used for each iteration, inputting the training set into a convolutional neural network, obtaining the probability that each sample pair belongs to a positive or negative sample pair by using a softmax function, and then obtaining the loss of each sample pair by using a multinomial logistic function; obtaining a difficult sample pair by using the loss of each sample pair; and training the convolutional neural network by using the difficult sample pair until the current number of iterations reaches the upper limit of the number of iterations, thus obtaining the pedestrian re-identification model. The pedestrian re-identification model is used to extract features of each picture in a picture set to be identified, and then a similarity order of the sample pairs in the pictureset to be identified is obtained. The pedestrian re-identification model, method and system avoid over-fitting and under-fitting, and have high recognition accuracy.

The invention discloses a lithium ion battery health state estimation method based on a codec model, which comprises the following steps of: (1) acquiring battery charging and discharging period dataincluding terminal voltage and current data and the maximum discharging capacity in each charging and discharging period; and (2) constructing a codec model of an attention mechanism according to thecharacteristics of the acquired data, wherein the codec model comprises an encoder and a decoder, voltage and current values are used as encoder input, an SOH estimated value of the battery is obtained and used as decoder output, and the number of nodes of each layer is determined; (3) preprocessing and normalizing the data acquired in the step (1), inputting the data into the codec model with randomly initialized weight, and minimizing the output error of the codec model through Adam algorithm training; and (4) inputting a new test sample into the codec model trained in the step (3), and calculating a prediction error so as to evaluate the accuracy of model prediction.

The invention discloses a feature fusing system and method for low-level visual features and text description information of images in social media. The method includes the steps that word segmentation and other processing are performed on the text description information of the images, so that a text description set, with a word as the unit, of each image is generated, and words which occur in the text description sets in a whole image set and the global occurrence probabilities of the words are counted; the visual features of the images are extracted, wherein the visual features comprise normalized HSV spatial color histogram and edge direction histogram features; according to the low-level features of the images, the visual similarity degrees among the images are calculated, and for each image, k images which are highest in visual similarity degree are taken to generate a neighboring image set; feature fusing is performed on the visual features and the text description information of the images, and the relevancy between a word and a target image is calculated according to the local probabilities of occurring, in image neighbors, of the word in a text description set of the target image and the global probability of occurring, in all the images, of the word. The accuracy of the text description information of the images can be improved.

Ultrafine patterns with dimensions smaller than the chemical and optical limits of lithography are formed by superimposing two photoresist patterns using a double exposure technique. Embodiments include forming a first resist pattern over a target layer to be patterned, forming a protective cover layer over the first resist pattern, forming a second resist pattern on the cover layer superimposed over the first resist pattern while the cover layer protects the first resist pattern, selectively etching the cover layer with high selectivity with respect to the first and second resist patterns leaving an ultrafine target pattern defined by the first and second resist patterns, and etching the underlying target layer using the superimposed first and second resist patterns as a mask.

The invention discloses a method and device for generating user characteristics, equipment and a storage medium and relates to the field of video recommendation. The method for generating the user characteristics comprises the steps of: obtaining a time series corresponding relationship between n groups of target videos and user accounts to obtain a word embedding matrix; training the word embedding matrix by using a loss function; and determining word vectors corresponding to the user accounts in the trained word embedding matrix as the user characteristics of the user accounts. According tothe method, the user characteristics are extracted on the basis of watching history and / or searching history of users; as long as the users normally use a video system, the data are continuously generated and updated without relying on other data sources; and therefore, the problem that the effective user characteristics cannot be generated for the users with empty or incomplete or incorrect attribute information, of a method in related work, can be solved and the accurate user characteristics can be generated for the users who use the video system.

The invention discloses an SAR image segmentation method based on a deconvolution network and sketch direction constraint. The invention mainly solves the problem that existing segmentation technologies are inaccurate in SAR image segmentation. The method comprises the implementation steps of 1, sketching an SAR image, thus acquiring a sketch; 2, dividing pixel sub-spaces of the SAR image according to an area chart of the SAR image; 3, training the deconvolution network; 4, clustering filter directions; 5, segmenting the pixel sub-space with a hybrid aggregation structure by using the SAR image segmentation method based on the deconvolution network and the sketch direction constraint; 6, performing independent target segmentation based on a sketch line gathering characteristic; 7, performing line target segmentation based on a visual semantic rule; 8, segmenting the homogeneous area pixel sub-space by using a polynomial-based logistic regression prior model; and 9, combining segmentation results, thus acquiring an SAR image segmentation result. According to the method provided by the invention, the good segmentation effect of the SAR image is acquired, and thus the method can be used for performing semantic segmentation on the SAR image.

The invention discloses an aircraft system fault diagnosis method based on MSCNN deep learning. The method comprises the following steps of S1, collecting decoded aircraft QAR data; S2, converting anaircraft state parameter in the aircraft QAR data into fixed-size two-dimensional data; S3, establishing the deep learning model MSCNN of a full mission profile; S4, according to an automatic sample identification working condition, adaptively generating a single working condition model, using the deep learning model MSCNN to automatically detect sample data which needs to be detected, and identifying a fault under a single working condition; and S5, comparing the diagnostic results of multiple working conditions, carrying out redundancy and verification, and acquiring a final diagnosis result.

The present invention relates to a method, device and storage medium for locating feature points on a human face. The method for locating feature points on a human face of the present invention comprises: preliminary locating of the position of a human face by combining human face detection with human eye matching, and acquiring preliminary locating information; according to the preliminary locating information, fitting feature points on the human face; and according to the fitting result, completing the locating of the feature points on the human face. The present invention has the beneficial effects that: by using human face detection and human eye matching at the same time, and incorporating at least one feature of an appearance model according to the preliminary locating information, the position information about a human face is more accurately located, and the location of the feature points on the human face is accurately accomplished.

The invention relates to a transformer excitation inrush current recognize method based on wave-shape sine theory, wherein it uses the sine function character of normal operation, short circuit, and external circuit short; the excitation inrush current contains non-periodical component, and the distant angle has no sine character; via comparing the sine degrees, it can judge inrush current and short circuit. The invention defines and calculates out the wave-length non-sine degree sigmaS (t); comparing the wave-length non-sine degree and non-sine preset value sigmaSZd, to judge the inrush current and short circuit current; when the transformer is idle and closes gate, there is inrush current, sigmaS (t) is higher than preset value sigmaSZd; opening judging condition, closing protecting; when the transformer has short circuits, sigmaS (t) is lower than sigmaSZd, it will open protection. The invention has accurate judge and phase-divert brake, with significant effect on CT saturation.

A disk clamp for securing data storage disks to the hub of a spindle motor in a hard disk drive has a circumferential rib protruding from a lower surface of the disk clamp. A complementary recess is formed in the hub of the spindle motor for closely receiving the rib on the disk clamp. The inner diameter of the rib and the outer diameter of the recess are closely toleranced to provide a very precise centering feature for mounting and aligning the disk clamp with respect to the hub and spindle motor. This design reduces the complexity of the tooling and fixtures required to complete the assembly and the overall part cost is reduced.

The invention discloses a color-information-maintaining objectionable image detection method under deformation sensitive organ models. The method comprises the following steps that: a GMM (Gaussian Mixture Model) is adopted for building color models of human body sensitive organs; HoG (Histogram of Oriented Gradients) features and GMM features of a sensitive organ training sample are extracted; for specific postures of each kind of human body sensitive organ, on the basis of features obtained after combining the HoG features and the GMM features of the human body sensitive organ, a deformable part model and a latent support vector machine are used for training detectors of the sensitive organ in the specific postures, and the detectors in various postures are integrated into a mixture deformation model of the sensitive organ; and various human body sensitive organ classifiers respectively detect test images, merge detection results and judge image properties. The method provided by the invention has the advantages that objectionable images are distinguished by using high-level semantic information of the sensitive organs in the objectionable images; the misjudging problem of normal images is effectively solved; and the method can be used for filtering pornographic information in the images.

A solid imaging apparatus and method employing levels of exposure varied with gray scale or time or both of digitally light projected image of a cross-section of a three-dimensional object on a solidifiable photopolymer build material. The gray scale levels of exposure of projected pixels permits the polymerization boundaries in projected boundary pixels to be controlled to achieve preserved image features in a three-dimensional object and smooth out rough or uneven edges that would otherwise occur using digital light projectors that are limited by the number of pixels in an image projected over the size of the image. Software is used to control intensity parameters applied to pixels to be illuminated in the image projected in the cross-section being exposed in the image plane.

The present disclosure discloses a method and apparatus for constructing a product picture. The method includes the following steps that: the characteristic data of the consumption users of a target product are classified according to a plurality of dimensions; the weight of each dimension is determined according to the characteristic data of the consumption users; the score of each of the consumption users is determined according to the characteristic data of the consumption users and the determined weight of each dimension; characteristic users in the consumption users are determined according to the scores; and the product picture of the target product is determined according to the characteristic data of the characteristic users. According to the method and apparatus of the invention, a more accurate product picture can be constructed based on comprehensive consideration on the weights of the dimensions of the characteristic data of the consumption users, and accurate positioning of the consumption users and the improvement of the product can be benefitted, and the benefits of the product can be increased.

A conventional infrared thermometer is equipped with a visible light source having exactly the same size and shape as the infrared detector to illuminate the target so as to visibly identify the area to be thermally measured. The light source is judiciously positioned within the system so as to make the images of the detector and the source formed by achromatic zooming optics coincide on the target surface. As a result, the exact location and dimensions of the target area under test are precisely identified and the system can perform errorless measurements of targets located at variable distances from the thermometer. Alternatively, the system is enabled to errorlessly measure the same target with variable spatial resolution.

The present invention is a LED connector principally comprised of two compatible halves with flat opposing surfaces, which form a complete cylinder when placed together. Its cross sectional area is no larger than that of the LED lamp.The solid-dome shaped LED component has an annular flat bottom that holds an outwardly extended pair of leads. Two parallel L-shaped recesses extend longitudinally over the connector half's flat surface. A large groove aligned with longer recess is included to accommodate the resister's solder. A pair of strategically spaced slots are formed longitudinally in the connector half's central region. The top slot holds a first conductive terminal member, and the bottom slot holds a second terminal member with their bases, respectively.The opposing connector's flat surface bears the same layout of recesses, slots and grooves. When the connector's halves are adjoined, electrical connections between the LED lamp and wires can be firmly secured and insulated. This interaction is conducted within the radius of the LED component, and is easily reversed by simple rewiring via the unique fastening means. Fastening means include three perpendicularly protruding posts with corresponding pinholes within the connector. Each of said posts and pinholes are individually sized and located in corresponding orientations of the first and second connector halves.

The invention discloses a feature extraction method and device as well as a stress detection method and device, relates to the voice detection technology, and aims to solve the problems of low accuracy of stress detection in the prior art. The technical scheme is that the feature extraction method comprises the following steps: according to a preset corresponding relationship between phonemes and acoustic feature pronunciation attributes, a first frame-level eigenvector of the acoustic feature pronunciation attributes is output through a first classifier; a second frame-level eigenvector of vowel and consonant pronunciation attributes is output by a second classifier according to the preset corresponding relationship between preset preset phonemes and the vowel and consonant pronunciation attributes; the first frame-level eigenvector of the acoustic feature pronunciation attributes or the second frame-level eigenvector of the vowel and consonant pronunciation attributes is used for mapping a syllable-level pronunciation eigenvector. The scheme can be applied to a voice detection process.

The invention provides a method and a device for establishing a detecting model and detecting a mobile phone calling / answering behavior. The method for establishing the model comprises the steps of marking first face information and first hand information when a user in a sample image is not calling / answering a mobile phone, and second face information and second hand information in calling / answering the mobile phone, generating a marked training sample, wherein the first face information and the second face information respectively comprise face characteristics and face position information, the first hand information and the second hand information respectively comprise hand characteristics and hand position information; respectively extracting a characteristic graph of the training sample by means of five layers of convolution, connecting pooling characteristic graphs which correspond with a third-layer convolution, a fourth-layer convolution and a fifth-layer convolution; and inputting the characteristic graphs into a convolutional neural network for training, thereby obtaining a face detection model and a face detecting model. The method and the device provided by the invention ensure a global characteristic and a local characteristic of the characteristic graph so that the characteristic graphs represent the characteristics of the training sample in a more comprehensive and accurate manner. Furthermore accuracy of the face detecting model and accuracy of the hand detecting model are improved.