731results about How to "Meet the needs of practical applications" patented technology

The invention relates to a perovskite / polymer composite luminescent material and a preparation method thereof, which belong to the technical field of composite materials and luminescent materials, wherein the structural formula of the perovskite is R1NH3AB3 or (R2NH3)2AB4, A and B form a ligand octahedral structure, R1NH3 or R2NH3 are filled into gaps of the ligand octahedral structure which is formed through the A and the B, R1 is methyl, R2 is a long-chain organic molecular group, the A is any one of metal Ge, Sn, Pb, Cu, Mn, Sb and Bi, the B is any one of Cl, Br and I. The perovskite / polymer composite luminescent material which is made through the preparation method is high in luminescent intensity, high in luminescent color purity, excellent in light and thermal stability and excellent chemical resistant property and mechanical property, and lays the foundation for theory research and application of the perovskite / polymer composite luminescent material in high-performance photoluminescence devices, flexible display, lasers, nonlinear optical devices. The preparation method can obtain composite materials which can cover a whole visible region through luminescent wavelengths, and has wide application prospect in the fields of wide gamut light emitting diode (LED) and high performance displays and the like.

The invention discloses a human body action recognition method and a mobile intelligent terminal. The human body action recognition method comprises the steps that human body action data are acquired for training so that feature extraction parameters and template data sequences are obtained, and the data requiring performance of human body action recognition are acquired in one time of human body action recognition so that original data sequences are obtained; feature extraction is performed on the original data sequences by utilizing the feature extraction parameters, and the data dimension of the original data sequences is reduced so that test data sequences after dimension reduction are obtained; and the test data sequences and the template data sequences are matched, and generation of human body actions corresponding to the template data sequences to which the test data sequences are correlated is confirmed when the successfully matched test data sequences exist. Dimension reduction is performed on the test data sequences so that the requirements for the human body action attitudes are reduced, and noise is removed. Then the data after dimension reduction are matched with the templates so that calculation complexity is reduced, accurate human body action recognition is realized and user experience is enhanced.

The invention discloses a hierarchical landmark identification method integrating global visual characteristics and local visual characteristics. High-dimensional characteristic vectors of landmark images are obtained and are used as the global visual characteristics of the landmark images; the local visual characteristics of the landmark images are obtained; the global visual characteristics andthe local visual characteristics are stored by adopting a hierarchical tree-shaped structure, and a visual characteristic set is obtained; each image is characterized according to the visual characteristic set; the images are pre-retrieved according to the global visual characteristics xi, and first candidate images are obtained; the first candidate images are further retrieved according to statistical characteristics vi of local outstanding points, and second candidate images are obtained; and the second candidate images are further retrieved according to a characteristic set yi of the localoutstanding points, and final candidate images are obtained and are fed back to a user. By adopting the hierarchical landmark identification method, the images to be identified can be rapidly and accurately retrieved, so the requirement of the user for convenient information acquisition is satisfied; and besides, through removing certain mismatching points, the landmark identification accuracy isimproved and the landmark identification complexity is reduced.

The invention discloses a blurred variable image blind restoration method based on a deep convolution network. The method comprises the following steps: S1, types of blurs that need to be considered for blurred images are modeled, and a range of blurred parameters is determined for each type of blur; S2, according to the range of blurred parameters for each blurred type determined in step S1, a training data set with variable blur is synthesized; each group of training data comprises clear images, blurred images and blur parameters; S3, a deep convolutional network is built and trained, a training data set is input into a neural network, weight of the neural network is optimized, and a trained deep convolutional network is obtained; S4, blurred images to be restored are input into the trained deep convolution network, and reconstructed clear images are output. The method can be free from limitation of a conventional deblurring algorithm, a full convolutional network is adopted for estimating a plurality kinds of variable blur, and precision and speed of the algorithm can be improved.

The present invention discloses a method for identifying a user action and an intelligent mobile terminal. The method comprises: acquiring user action data, and training the user action data to obtain a feature extraction parameter and a template symbol sequence; collecting data that needs to execute user action identification during user action identification to obtain an original data sequence; performing feature extraction on the original data sequence by using the feature extraction parameter to obtain a dimension-reduced test data sequence; converting the test data sequence into a discrete character string to obtain a symbol sequence of the test data sequence; and performing matching on the symbol sequence of the test data sequence and a template symbol sequence, and when the matching is successful, determining that the user action corresponding to the template symbol sequence occurs. According to the method provided by an embodiment of the present invention, dimension reduction is performed on the original data sequence by using the feature extraction parameter, the dimension-reduced data sequence is symbolized, and then the dimension-reduced data sequence is matched with the template symbol sequence, so that calculation complexity is lowered, perceptual identification efficiency is improved, and good user experience is achieved.

The invention discloses a construction waste comprehensive treatment method which comprises the following steps: primary crushing, sorting, primary magnetic separation, secondary crushing, secondary magnetic separation, primary screening, tertiary crushing and tertiary screening. The construction waste comprehensive treatment method can be used for thoroughly sorting foreign matters in the construction waste, thereby enhancing the quality of the regenerative aggregate. The invention has the advantages of reasonable procedure arrangement, low labor consumption and high efficiency, and is simple to operate. When being used for treating construction waste, the method has the social benefits of saving land resources and reducing environmental pollution.

The invention discloses a method and device for distributing network addresses and belongs to the field of internet. The method disclosed by the invention comprises the following steps: receiving the address distribution request of user equipment; obtaining the spare network address from an address pool according to the address distribution request and the address pool mapping relation which is locally stored, wherein the address pool mapping relation comprises the address pool and at least one subaddress pool contained in the address pool; and distributing the spare network address to the user equipment. The device disclosed by the invention comprises a receiving module, an acquisition module and a distribution module. In the invention, the network addresses is distributed for the user equipment according to the address pool mapping relation which is locally stored by virtue of changing the existing address distribution procedure when the address distribution request is received, thereby achieving the purpose of distributing the address for the user equipment with pertinency, improving the distribution efficiency and meeting the requirements in the actual application of a client better.

The invention relates to an enrollment automatic question-answering method and system based on a conversation robot. The method comprises the following steps that 1, characters input by a user are obtained; 2, character processing is conducted on the characters input by the user; 3, a best answer is selected from a knowledge base by means of a fuzzy matching method and an internal reasoning mechanism according to the characters subjected to character processing; 4, the best answer is sent to the user. According to the enrollment automatic question-answering method and system based on the conversation robot, the ALICE open-source conversation robot is improved, a domain ontology base serves as an additional knowledge base of a question-answering system, a hypernym-hyponym relation of a constructed domain ontology is utilized for conducting user intention excavation on a problem put forward by the user, a related content recommendation is given to the user by means of hypernym-hyponym information of the domain ontology on the basis of achieving basic question answering, therefore, an examinee can also obtain some recommended results with associated content when the examinee does not obtain an answer of the relevant question, and the satisfaction degree of the question-answering system is improved.

The invention discloses a circuit and a method for realizing a transport driver interface (TDI) in a complementary metal oxide semiconductor (CMOS) image sensor, and relates to the field of analog integrated circuit designs. The circuit comprises a pixel array, a metal oxide semiconductor (MOS) tube, a sampling capacitor, an operational amplifier, serial buses and an integrator array formed by connecting a plurality of integrators in parallel; each integrator comprises a first switch, a second switch, a third switch, a fourth switch, a fifth switch and an integrating capacitor; the pixel array is respectively connected with one end of the sampling capacitor and the source electrode of the MOS tube through serial buses; the grid electrode of the MOS tube is connected with a bias voltage; the drain electrode of the MOS tube is earthed; the other end of the sampling capacitor is respectively connected with the negative polarity end of the operational amplifier and one end of the first switch; the positive polarity end of the operational amplifier is connected with a reference voltage; the other end of the first switch is respectively connected with one end of the second switch and one end of the fourth switch; the other end of the second switch is respectively connected with one end of the third switch and one end of the fifth switch; the other end of the third switch is connected with the output end of the operational amplifier; the other end of the fourth switch is connected with one end of the integrating capacitor; and the other end of the integrating capacitor is connected with the other end of the fifth switch.

A kind of low-dropout regulator, comprising a differential amplifier circuit, an output amplifier circuit and a voltage divider circuit, said output amplifier circuit comprises a PMOS transistor MPass1 whose source is connected to the input supply voltage Vdd, and connected between the voltage output point and The output capacitor Co between the ground, the two input terminals of the differential amplifier circuit are respectively connected to the reference voltage Ref and the voltage division point of the voltage divider circuit, the output terminal of the differential amplifier circuit is connected to the gate of the PMOS transistor, and the The voltage divider circuit is connected between the voltage output point and the ground, and is characterized in that it also includes a resistor Ra connected between the drain of the PMOS transistor MPass1 and the voltage output point, and a resistor Ra connected between the drain of the PMOS transistor MPass1 and the capacitor Cc between the voltage divider points. The low dropout voltage regulator provided by the invention has the advantages of low quiescent current, high power supply suppression, strong stability, low dropout voltage and low cost, and meets the needs of practical applications.

The invention relates to a testing device and method for the defocused spot and the color bias of an optical system, wherein the device comprises a monochromator, a target wheel, a collimator, a rotary table, a CCD (Charge-Coupled Device) micro-measuring unit and an acquisition and control computer; the monochromator, the target wheel, the collimator and the CCD micro-measuring unit are sequentially positioned on the same optical axis; the rotary table is used for arranging the tested optical system and the CCD micro-measuring unit; and the rotary table, the monochromator and the CCD micro-measuring unit are all connected with the acquisition and control computer. The invention can be used for effectively testing the defocused-spot diameters and the color biases of various optical systemswithin different wavelength ranges and well ensures the imaging quality of the optical systems.

The invention discloses a treated sewage quality prediction method based on combination of support vector classification and a GRU neural network, and belongs to the technical field of sewage treatment. Missing value processing, abnormal value elimination and data standardization are carried out on the collected sewage historical data, a PCA principal component analysis method is adopted to carryout dimension reduction on the data, and the selected auxiliary variable is used as an input variable of a sewage quality prediction model; a sewage effluent key prediction model is established by adopting a GRU neural network suitable for processing time series data, a support vector machine model is firstly introduced to classify sewage quality data, and then the classified data is respectivelymodeled through the GRU neural network algorithm to predict effluent quality. When the SVM model is trained, a grid search method and a cross validation method are used for optimizing model parameters, the prediction precision of the obtained joint prediction model is more accurate, the model effect is better, the network performance can meet the actual application requirements, and accurate prediction of the effluent quality of the sewage treatment system can be realized.

The invention relates to an RTOS-GPOS dual-operation system robot controller based on a ZYNQ dual-core processor. The controller comprises a ZYNQ dual-core processor, a peripheral hardware module withthe ZYNQ dual-core processor as a control core and an RTOS-GPOS dual-operation system tanking the ZYNQ dual-core processor as the hardware basis. The GPOS dual-operation system comprises a master system module, a slave system module, a system starting control module and a dual-core communication module. According to the invention, a dual-operation system parallel operation mechanism based on an asymmetric multi-processing architecture is established on a dual-core chip ZYNQ; a dual-system robot controller is constructed; a general operation system Linux and a real-time operation system FreeRTOS can be operated at the same time; therefore, upper-layer tasks such as man-machine interaction and path planning can be completed, real-time tasks such as motion control and emergency operation canbe executed, a controller which is light, high in communication efficiency and good in reusability is provided for robot development, and the requirements of practical application can be well met.

The invention discloses a rotary speed measuring device and a rotary speed measuring method. The rotary speed measuring device is realized by the following steps: a vibration signal picking sensor is fixedly arranged on a shell of a rotary body to be measured, a supporting seat, a rigid foundation or a rigid object which is in direct contact with the rotary body to be measured and is adjacent to a rotary body; a vibration signal excited by the rotary body to be measured is acquired by the vibration signal picking sensor; the vibration signal is transmitted to a signal analyzing and processing module by a signal transmission lead wire; and the signal analyzing and processing module is used for analyzing and processing the vibration signal so as to extract the rotary speed of the rotary body to be measured. The rotary speed measuring method comprises the steps of: setting a sampling frequency, a data truncation length and a vibration order; carrying out vibration signal picking, filtering, data windowing, data truncation, data conversion, feature data extraction, feature data check, rotary speed calculation, rotary speed check, vibration order check and rotary speed real-time measurement; and according to the vibration order relation, calculating the rotary speed of the rotary body to be measured. The rotary speed measuring device can measure the rotary speed of the rotary body to be measured without directly contacting with the rotary body to be measured; and the measured rotary body does not need an outward extending end, so that the rotary speed measuring device is convenient to use.

The invention relates to an information filling card identifying method based on template matching. A blank information card is used to build the image information of a template and obtain a template information card; the effective filling threshold of the filling point option of the template information card is set; the background mode definition of the filling point option of the template information card to be identified is extracted and different background modes of the template option are respectively transferred to carry out filling point identification; if the filling point result is larger than or equal to the set threshold, then the option is considered to be validly filled; or the filling is invalid. The information filling card identifying method based on template matching can be suitable for identifying the gray scale modes of different types; the identifying results are graded and the flexibility is high; the integral identifying rate can be improved; batch identification is available, the human-computer combination can be realized, better meeting the demands of actual application. The information filling card identifying method defines all the option areas to obtain accurate template information. The difference between the current image and the template image are used for identification, thus solving the problem of grain background interference.

The invention relates to a Chinese word similarity calculation method based on a fusion strategy. The method comprises steps of calculating word similarity based on the combination of four of HowNet,synonym forest, Word2Vec trained Chinese Wikipedia encyclopedia corpus and a Baidu dictionary; for two input words, firstly, judging whether the synonyms exist in a HowNet or synonym forest or not; ifyes, using the HowNet or synonym forest for calculating the similarity, if not, judging whether the HowNet or synonym forest exists in the Wikipedia corpus or the Baidu dictionary or not, and if yes,using the Word2vec or the Baidu dictionary for calculating the similarity of the words. The invention provides a Chinese word similarity calculation method based on a fusion strategy. According to the fusion strategy, the known network, the synonym forest, the word2vec and the Baidu dictionary are comprehensively considered, advantage complementation among strategies is formed, the calculated Spearman correlation coefficient and Pearson correlation coefficient are higher than those of other methods, the accuracy of a word similarity calculation result is improved, and the requirements of practical application can be well met.

The invention relates to a Gd2O3-rich borogermanate scintillation glass, and a preparation method and application thereof. The invention provides a Gd2O3-rich borogermanate scintillation glass which is characterized by comprising the following components in percentage by mol: 20-60 mol% of B2O3, 20-60 mol% of GeO2, 15-40 mol% of Gd2O3 and 0.01-10 mol% of compound containing rare earth or transition metal ions. The rare earth ions comprise Y<3+>, La<3+>, Ce<3+>, Pr<3+>, Nd<3+>, Pm<3+>, Sm<3+>, Eu<3+> (Eu<2+>), Gd<3+>, Tb<3+>, Dy<3+>, Ho<3+>, Er<3+>, Tm<3+>, Yb<3+> and Lu<3+>; and the transition metal ions comprise Mn<2+> (Mn<4+>), Sn<2+>, Zn<2+>, Bi<3+>, Cr<3+>, Ti<4+> and Zr<4+>. The invention also provides a method for preparing the Gd2O3-rich borogermanate scintillation glass, and application of the Gd2O3-rich borogermanate scintillation glass.

The invention discloses a spatial heterodyne spectrometer tester interferometer, improved on the basis of Michelson interferometer, replacing two planar reflectors in the Michelson interferometer with two diffraction gratings; target light source is modulated through collimating light path system into parallel light source, and the interferometer makes spatial interference modulation on light wave to form interference stripes with a certain spatial frequency which is a function of wavelength, and the interference stripes are imaged through imaging object lens in linear array (or planar array) detector and the image is recorded by the detector and stored in computer, which can recover spectrum curve by specific algorithm.

The invention relates to a method for calculating the similarity of patent texts. The method comprises the following steps that patent data is extracted from the two patent texts and preprocessed; word weights are calculated by combining part-of-speech weights, word position weights and a TF-IDF algorithm; the two patent texts are expressed in a vector space model, and two distributed word vectorsare obtained; the similarity of the texts is calculated, when the similarity of the obtained patent texts is larger than a set threshold, it is considered that the two patents are similar, and otherwise, the two patents are not similar. According to the method, patent structure characteristics and inter-word semantic relations are comprehensively considered, and special structures, such as IPC classification numbers, abstracts and claims, of the patent texts are fused into the method for calculating the similarity of the texts; compared with common text similarity calculation methods, the method is more targeted, high accuracy and a high recall rate can be guaranteed, and requirements of actual application can be well met.

A calibration method of a coded structured light three-dimensional vision system comprises the following steps of: 1) camera calibration processes: i.e. a two-dimensional calibration template is adopted, a calibration target consists of standard black and white chess checks, feature points are formed by chess checks points, a camera shoots a plurality of images in different positions, check angular points are selected as the feature points for the plurality of images, and a plane calibration method is adopted for calibrating the camera so as to obtain internal parameter calibration K of the camera as well as a mean value focal distance f; and 2) projector calibration processes: i.e. a light template generated from coding is used for projection onto a coplane calibration target, black and white check angular points are used as positioning points, and image coordinates of the black and white check angular points are simultaneously extracted to generate a world coordinate by utilizing theimages obtained from the camera calibration. The calibration method achieves high camera calibration accuracy and can satisfy the requirements of actual application.

The invention relates to a Chinese patent text similarity calculation method. The method comprises the steps of performing word segmentation on texts; calculating TF-IDF values for word segmentation results, extracting the word segmentation results with the relatively high TF-IDF values to serve as keywords, locating sentences where the keywords are located to serve as key sentences, and taking maximum weight values of the keywords in the key sentences as weight values of the key sentences, thereby obtaining a keyword set of each text; and calculating weights of comparison texts of the key sentences, selecting the key sentences of the to-be-compared texts and the comparison texts in sequence, and based on the sentence similarity of the key sentences, calculating the similarity of the texts. By utilizing existing patent domain ontologies, semantic relationships in the patent texts are analyzed; by utilizing a vector space model and the domain ontologies, patent text similarity is calculated; the correct rate and the recall rate of a calculation result are relatively high; the similarity between patents can be described more accurately; the patent examination speed can be increased;and the need of actual application can be well met.

The invention relates to a pipeline video defect detection method based on a convolutional neural network. The method comprises the steps of extracting frames of a video, training a plurality of CNNsto classify each frame of image; counting the result returned by each CNN, determining the defect type of the frame, segmenting the video into continuous image frames by taking the pipeline closed circuit television video as input, and sending each frame of image into a plurality of trained CNNNs for binary classification, wherein the classification result only comprises a certain specific defectand no defect. According to the invention, the accuracy of pipeline defect detection is obviously improved; a feasible method is provided for video detection; the automatic detection efficiency of thepipeline defects can be improved; the pipeline defect detection method has the advantages that the detection accuracy is high, the detection speed is high, the application value in pipeline video defect detection is high, a satisfactory result is obtained, and the pipeline defect detection method can be used as a technical reference for pipeline defect detection workers and can well meet the requirements of practical application.

The invention provides a preparation method of a porous structural elastic composite material for extracting uranium from seawater. The preparation method comprises preparing an amidoximated polyacrylonitrile solution, mixing the amidoximated polyacrylonitrile solution and a strong alkali aqueous solution, maintaining heating and stirring until dissolved, adding the stirred solution into an aqueous solution of a polymer material capable of cross-linking gelation, mixing uniformly the solution, immersing a porous structure elastic material into the mixed solution, carrying out vacuum defoaming,carrying out drying, immersing the dried porous structure elastic material in a cross-linking agent solution, carrying out cross-linking gelation and carrying out rinsing with water. The aqueous solution of the polymer material which can be cross-linked and gelled is cross-linked and gelled so that the coating strength of amidoximated polyacrylonitrile on the porous structure elastic material skeleton is improved, the material flexibility, mechanical stability and hydrophily are improved and the high adsorptive porous structural elastic composite material for extracting uranium ions from seawater is obtained.

The invention discloses a novel biological pollution-resistant polyamide reverse osmosis composite membrane and a preparation method thereof. The preparation method comprises carrying out graft modification on a polyamide reverse osmosis membrane through atom transfer radical polymerization (ATRP), carrying out sterilization particle loading to obtain a reverse osmosis composite membrane with bioattachment resistance and antibacterial properties. The reverse osmosis composite membrane polymerization process is carried out at the room temperature and has high controllability. The reverse osmosis composite membrane can be conveniently and fast prepared, has stable performances, effectively improves use efficiency and a use life and reduces a seawater desalination cost.

The invention provides a system for comprehensive arrangement and execution management of a production equipment maintenance plan in the process industry. The system can be used for solving the technical problems of the conventional maintenance plan management mode, such as high complexity, low work efficiency and poor information management. The system provided by the invention is characterized by comprising a network platform consisting of a computer, a server computer and a database application computer, wherein the network platform ensures that an annual maintenance plan management module, a monthly maintenance plan management module and a maintenance management module are mutually connected.

A kind of electronic car cards comprises a body, a radio-recognition CMOS chip set in the body, multi- frequency / single frequency resonance antenna connected with the CMOS chips. The body comprises metal structure, which constitutes the whole or local part of the antenna. The body is a metal plate with an antenna pattern. The blank area of the antenna pattern is a hole, a hatch, a notch, or the combination of them. The body is the metalloid or metal plate with antenna pattern and metal-filled film. The metal-filled film is formed using a layer of material that the radio wave can penetrate bonded with a metal-filled layer with antenna pattern. The metal-filled film is pasted onto the metalloid or metal plate. The blank area of the antenna pattern of the metal board is filled with material with low dielectric constant. The blank area of the antenna pattern of the metal board is embedded with the radio-recognition CMOS chip sealed with material with low dielectric constant. Using the car cards to construct the antenna, and the performance of the antenna is fine so that it can satisfy the demand in practical application.

The invention discloses a multi-target collaborative salient-region detection method based on similarity propagation and relates to the field of multimedia information processing and computer vision. The method includes the following steps: with respect to a plurality of image files which are input, segmenting each image into superpixels and performing pairwise measurement on inter-superpixel similarities and establishing a superpixel similarity graph model; realizing inter-image superpixel similarity matrix bipartite matching through the superpixel similarity graph model; performing superpixel graph model similarity diffusion so as to obtain a similarity graph model; and calculating a saliency map through the similarity graph model. The method performs collaborative salience detection on a plurality of targets in a plurality of images which are input so that detection precision is improved; through a parallelization similarity propagation algorithm, the operation speed is reduced and needs in practical application are met; and experiment results show that the method obtains a more accurate detection result during a shorter calculation time.

The invention discloses an earthquake collapse analysis method for a high-rise steel frame structure, which comprises the following steps: obtaining the dynamic equation of the high-rise steel frame structure under the earthquake action; according to a central difference method and the dynamic equation, solving the speed and the acceleration of the displacement differential of the high-rise steel frame structure within time interval delta t; according to the Von Mises yield condition and a corrected K&K model, obtaining unit stress sigma t at the t time; defining a failure criterion corresponding to the K&K model; according to the failure criterion, judging whether a unit fails or not; if so, deleting the failed unit; if not, according to a geometric equation and the displacement u(t+delta t) of the high-rise steel frame structure, obtaining the strain increment d epsilon of the unit; according to a damage evolution equation, obtaining the accumulated damage value D of the unit; according to the accumulated damage value D of the unit, calculating the layer damage value of the high-rise steel frame structure; obtaining the stress sigma t+delta t of the t+delta t time unit; and updating time t to t+delta t. According to the earthquake collapse analysis method, the whole collapse damage process of the high-rise steel frame structure under the earthquake action by considering the damage accumulation effect can be precisely simulated.

The invention provides a snakelike wound software actuator, which is aimed at solving the problem that a software robot is still in a research stage and needs to be further expanded. The snakelike wound software actuator comprises an elastic corrugated pipe, a strain limiting layer, an elastically stretchable layer, a fixed connecting piece, and an air passage pipe for connection with an air source. The novel snakelike wound actuator provided by the invention covers a wound actuator with the strain limiting layer and a wound actuator without the strain limiting layer. The winding performance of the snakelike wound actuator with the strain limiting layer is determined by depending on the strain limiting layer and the elastically stretchable layer; the snakelike wound actuator with the strain limiting layer can wind a long and thin rod rapidly and efficiently; and with the increase of inflation pressure, the snakelike wound actuator winds an object more tightly and can lift a 1.3kg object. The wound actuator without the strain limiting layer is relatively light and soft; the winding manner of the actuator is determined only by depending on the fixed orientation of the elastically stretchable layer; and the wound actuator without the strain limiting layer can also wind and lift the object.

The invention relates to an automatic keyword extracting method. The method includes: extracting general words in a technical standard, extracting candidate keywords, filtering the general words aiming at the candidate keywords, calculating weight scores of the candidate keywords according to position features, word co-occurrence features and context semantic features, calculating a dynamic threshold according to a range of the weight scores of the candidate keywords, and determining a result keyword according to the dynamic threshold. The automatic keyword extracting method has advantages that keyword extraction is realized according to the position features, the word co-occurrence features and the context semantic features, influences of document interior positions and context semantic features on keyword weights are considered comprehensively, high accuracy and high recall rate are realized, 3GPP technical standard retrieval quality is improved, labor cost is reduced, and practicalapplication demands can be well met.