96results about How to "Able to learn" patented technology

The present invention discloses a method for extracting keywords based on the K-MEANS and the WORD2VEC. The method is characterized in that: the global semantics and the theme of each branch are integrated; a space vector is constructed by using the WORD2VEC algorithm; the K-means algorithm is used to remove the fuzzy words; the centroids distance is calculated; high quality keywords are obtained after the clustering evaluation; and by lifting the weight value, the dynamic optimization of the lexicon is realized, so that the keyword extraction has the learning evolution ability. According to the method disclosed by the present invention, the extracted keywords can reflect the internal classification theme of the document, and each keyword can well reflect the classification; and the method has the characteristics of high quality of the final keywords, wider adaptability, a more accurate result, and the like.

The invention discloses a semantics-based automatic service combination system for web service relation network, which inputs a user request and a field body description, and outputs a solution for meeting the functional requirements of a user. The system is characterized by comprising various data sets and modules, and executing automatic service combination according to a first stage and a second stage, wherein the first stage mainly comprises a mixed service discovery module, a service transition and relational matrix establishment module, an operation instantiation module, an action dependency graph establishment module and an action filtering module, while the second stage mainly comprises a case acquisition module, a case modification module, a case correction module and an abstractservice instantiation module. Compared with the prior art, a mixed service discovery algorithm also can accurately and highly efficiently find out proper web services from a service network to conclude and sort out the discrete service information on the web service relation network, and the combination of the mixed service discovery algorithm and the web service relation network also greatly improves the service discovery efficiency.

The invention discloses a sparse matrix and convolutional neural network-based picture compression method. The method comprises the following steps of converting an image into two-dimensional matrix data; assessing, optimizing and constructing a sparse matrix image by utilizing a convolutional neural network; optimizing a compression scheme through the convolutional neural network; performing compression processing on the sparse matrix image by using the selected optimal compression scheme; and finally obtaining a high-compression-ratio and low-distortion compressed image. According to the method, the problem of difficult use and identification of the image due to low compression ratio and excessively reduced quality of the compressed image in an existing compression technology is solved; and the optimization can be carried out according to picture contents, and the proper image construction and compression method can be self-selected, so that the picture feature extraction and compression method has self-optimization and learning capabilities.

The invention provides an equipment monitoring and diagnosing method, comprising the following steps of obtaining a vibration signal of the equipment under a fixed rotation speed; calculating the characteristic value complying with Gaussian distribution within the corresponding frequency width range according to the vibration signal so as to establish at least one benchmark line; subsequently judging the normality or abnormality of the equipment according to the benchmark line, the corresponding frequency width range and the characteristic value; therefore, if the vibration signal exceeds the benchmark line of safe range, the abnormality and failure type can be diagnosed and recognized; and furthermore, the aim of owning the automatic monitoring and diagnosing and learning capabilities can be realized. Furthermore, the equipment monitoring and diagnosing method does not need to consider all failure types that may occur in the whole service life of the equipment so as to greatly reduce the development scale and difficulty of the failure monitoring and diagnosing system, thus improving the visualization of the equipment condition and increasing the completion rate of the equipment.

The invention discloses a geologic environment disaster monitor and a geologic environment disaster monitoring method. The monitor is connected to a central control system, which is formed by environment parameter sensing modules, a video data acquisition and processing module as well as a module communication channel and a central control system communication channel. The environment parameter sensing modules are used for acquiring monitoring parameters, the monitoring parameters are transmitted to the video data acquisition and processing module via the module communication channel, and a microprocessor is used for transmitting instructions to the environment parameter sensing modules via a communication interface and for receiving data transmitted from the environment parameter sensing modules; and meanwhile, the central control system communication interface receives the instructions of the central control system to carry out control, algorithm is identified via different video data modes according to environment parameters, data analysis and processing are carried out, and an analysis and processing result, mode identification and calculation result and video data are transmitted to the central control system for manual intervention judgment. According to the invention, reasonable compressed video data is selected and transmitted, machine visual algorithm can be corrected, and demands for transmission bandwidth are greatly reduced.

A communication method of word information includes matching language type of receiving terminal to language used in received word information by service server and retransmitting word information to translation platform to let it be translated then sending it to receiving terminal if language used in received word information is not matched to language type of receiving terminal. The translation platform is worked as querying databank to know qualified translator then sending word information to said translator to let word information be translated at his terminal.

The invention provides a mobile phone user behavior similarity analysis method based on mobile big data, and the method comprises the steps: collecting the movement track information of a user group in a time period through the signaling data of a telecom operator; counting the occurrence frequency or residence time length PF of each user in each sector in the time period and the mean value GF ofthe occurrence frequency or residence time length of all the users in each sector, and calculating PF / GF to obtain a PF-IGF value of each user; calculating an included angle cosine value of the PF-IGFvalues of every two mobile phone users, and judging the behavior similarity of the two users according to the obtained included angle cosine value: the larger the included angle cosine value theta is, the closer of the activity behavior tracks of the two mobile phone users are; wherein the smaller the value of the included angle cosine value is, the more irrelevant of the activity behavior tracksof the two mobile phone users are. The method can be used for researching the similarity degree between every two persons and finding the similar crowd of each person in a specified range.

The invention provides a hearing aid system based on an intelligent speech recognition technology, and relates to the field of medical equipment. The hearing aid system is characterized in that the system comprises a recognition terminal, a mobile terminal, and a cloud terminal. The mobile terminal comprises a voice collection device, a signal conversion device, a decomposition filter, an enhancing system, a preprocessing computation device, a judgment system, a synthetic filter, a data sending device, a local database, and a data receiving device. The mobile terminal comprises a controller, a processor, a display, a speech recognition device, and a mobile data transmission device. The cloud terminal comprises a cloud data transmission device, a cloud database, and a cloud analytical processor. The system has the advantages of accurate Chinese speech recognition, self-learning ability, leading recognition algorithm, good later maintenance, and the like.

The invention provides an intelligent community system which comprises a transmission network, an online intelligent community service module, an intelligent home furnishing module, an intelligent family security module, a data processing module and a data routing module, wherein the transmission network is used for transmitting data among the modules; the online intelligent community service module comprises recreation and sports reservation, medical treatment reservation, health examination, electric car rental, community affair processing, community distribution, comprehensive traffic query, online property, online shopping, health record query, service agency, express collection, remote family doctor, emergency agency and warm greetings; the intelligent home furnishing module comprises the control of an air purifier, a television set, a projector, a water heater, an air conditioner, an intelligent mattress, a sound box, a fan, a curtain and an electric lamp; the intelligent family security module comprises an IP (Internet Protocol) video camera, an emergency call button, an infrared intrusion alarm and a fire alarm; the data processing module comprises a data storage module, an audio module, a network processing module, a WIFI (Wireless Fidelity) module, a power supply module, a display module and a control module; the dada routing module comprises a network router and a WIFI hotspot.

The invention discloses a railway wind disaster prevention prediction method in the technical fields of railway disaster prevention alarming and monitoring. The method comprises the following steps of: acquiring wind speed and wind direction data of a current moment measuring point; calculating the first-order difference, the second-order difference and the third-order difference of the wind speed and wind direction data of the current moment measuring point; retrieving the wind speed and wind direction data, the first-order difference, the second-order difference and the third-order difference of three continuous moments before the current moment and establishing an identification equation of a wind speed and wind direction model; establishing a quaternion first-order equation set according to the identification equation of the wind speed and wind direction model and acquiring the model parameters of the wind speed and the wind direction through a Gaussian elimination method; updating the model parameters according to an average sliding coefficient; and calculating the wind speed and wind direction data of the next moment according to the updated model parameters. The invention improves the prediction accuracy and the accuracy of a prediction result.

The present invention provides an automatic questions selecting method of an examination system on a network based on a layered genetic algorithm. A key point of the invention is introducing characteristics of the layered genetic algorithm into the automatic questions selection of the examination system on the network. The method can select questions which are best matched with a question requirement to generate examination paper rapidly and satisfy the rapid and high-efficiency question requirement by the high capacity item bank of the examination system according to the crucial attributes of investigation scope and difficulty of paper etc. with respect to users. The automatic questions selecting method based on the layered genetic algorithm can integrate situations such as questions difficulty of various levels, total difficulty of the paper and questions that have already been used in examines, so that a question selection efficiency of the examination system can be greatly enhanced, a matching degree between the paper and the question selection requirement can be improved, and a checking efficiency can be increased.

The invention discloses a monitoring image processing method and system, and a monitoring terminal. The method comprises the following steps that: S1: obtaining an image, and extracting the image intoan RoI (Region of Interest) and a non-RoI through a preset RoI extraction model; S2: comparing the RoI by a preset first compression rate, and compressing the non-RoI by a preset second compression rate, wherein the first compression rate is smaller than the second compression rate; and S3: sending the compressed RoI and non-ROI to external equipment to be subjected to image identification. The method has the advantages that a data transmission amount can be greatly lowered, network requirements are low, network cost is greatly lowered, image detection efficiency is improved, response speed is high, detection accuracy is high and the like.

The invention relates to an artificial intelligent calligraphy robot. The artificial intelligent calligraphy robot comprises a base, a stepping motor A, a rotary disk, a stepping motor B, a stepping motor C, a big cantilever, a small cantilever, a stepping motor D, a paper rolling motor, Chinese art paper, a paper winding tube, a connection rod, a rotating shaft, a stepping motor E, a clamping plate, a stepping motor F, a self-learning camera, a synchronous belt, a pen holder, a fastening screw, an automatic ink outlet brush pen, an ink conveying hose, an ink conveying pump and an ink box. According to the artificial intelligent calligraphy robot, a folding neutral network and a sequence neutral network are combined to obtain output of a sequence, and the complex feedback from images to movement data is achieved; and after basic training is carried out, by means of self-repeated writing, self-learning and training are carried out, and the calligraphy level of the calligraphy robot canbe improved without human intervention; and the calligraphy robot has a learning ability, the calligraphy robot can be applied to the calligraphic art study, and after certain training, the calligraphy robot can try to imitate famous fonts and can also help an ordinary person to learn and understand calligraphy.

The embodiment of the invention discloses a self-adapting fuzzy pulse neurolemma system which has the following structure: TT=(A, Np, Nr, syn, I, O), wherein A-{a} is a single letter set, NP={sigma p1, sigma p2,..., sigma pm} is a proposition neuron set, sigma pi is a proposition neuron, i is larger than or equal to 1 and smaller than or equal to m, Nr={sigma r1, sigma r2,..., sigma rn} is a regular neuron set, sigma ri is the i-th regular neuron set, and i is larger than or equal to 1 and smaller than or equal to n; the system represents the relation of connection between the proposition neuron and the regular neuron, but no relation of connection exists between two proposition neurons or between two regular neurons; I which is an element of a set Np is an input neuron set; O which is an element of the set Np is an output neuron set. According to the self-adapting fuzzy pulse neurolemma system provided by the invention, the fuzzy pulse neurolemma system can be reasoned by setting two types of neuron sets, and the system has learning capacity and can be used for handling with dynamic problems in application.

The invention relates to a method for customizing skin care products. The method comprises the steps of inputting user skin information into a recipe management system; generating solution codes through the recipe management system; inputting the solution codes into a single-piece production line; and executing single-piece production through the single-piece production line according to the input solution codes. Through the method, skin care products applicable to a user are customized for the user. According to the method, a formulator does not need to execute customization for each user, customization can be executed fully automatically in the whole course, and the method is applicable to respectively executing customization of a small amount of skin care products for each user rather than large-scale manufacture for various people with one recipe.

The invention relates to a garbage can for intelligently sorting medical waste and a sorting method thereof, and belongs to the technical field of auxiliary sorting and recognition. The garbage can comprises a box body assembly, a box cover assembly, an image recognition system and a power supply module, wherein the box body assembly comprises a box body housing and a plurality of built-in garbagecans, the box cover assembly comprises a connecting cover, a central motor and an inner rotating bucket, the image recognition system comprises a camera, an information storage module, an informationprocessor and a human-computer interaction module, the box cover assembly is located above the box body housing, the connecting cover is used for fixing the inner rotating bucket, the central motor and the image recognition system, and the power supply module is connected with power supplies of all devices to which power needs to be supplied in the garbage can for intelligently sorting medical waste. The garbage can for intelligently sorting medical waste has the function of automatically identifying and sorting the medical waste by a machine, the use is convenient, and the sorting accuracy of the medical waste is increased; in addition, according to the garbage can for intelligently sorting medical waste, the number of garbage inlets is reduced, so that the condition of environmental pollution caused by the exposure of the medical waste is avoided, and the effects of cleanness and sanitation are achieved.

The invention discloses an intelligent exit and entry system based on faces and vehicle-related characteristics. The intelligent exit and entry system comprises a front-end recognition module, an information positioning module, a data analysis module, an intelligent decision module and an output control module. The front-end recognition module comprises a biometric identification intelligent camera and a vehicle identification intelligent camera. The information positioning module comprises the functions of personnel information positioning and vehicle information positioning, and based on artificial intelligence analysis of big data, the data analysis module analyzes intelligent exit and entry relationship information through vehicle information and personnel information acquired by the positioning module and relevant entry and exit record information. The intelligent decision module makes intelligent probabilistic decisions on the intelligent exit and entry relationship information to obtain probability assessment of various release decisions. According to the intelligent exit and entry system based on the faces and the vehicle-related characteristics, face information of a driver and passengers on a vehicle are also identified while the vehicle is identified, whether the vehicle can enter and exit the field or not is automatically determined by an artificial intelligence analysis system mining relevant information of vehicles and people, the entry and exit experience and the traffic efficiency are greatly improved, and manual management work is reduced.

The invention provides a speed sensorless control method, comprising the following steps: a deviation between the given speed of the motor and the feedback value of the rotor angular speed is adjustedby the CMAC-PID intelligent controller and the torque current is output, and the deviation between torque current and torque current feedback is regulated by PI regulator, and the output voltage is uq. The deviation between the excitation current and the feedback value of the excitation current is regulated by the PI regulator and the output voltage ud is obtained. The voltage uq and the voltageud are inversely transformed to obtain two-phase control voltages. The injected high-frequency carrier signal is superposed on the two-phase control voltage, and the superposed voltage signal is usedas the input signal of SVPWM modulation. The output of SVPWM modulates six PWM pulse signals needed to control the on-off of the trinomial inverter switching devices, and the inverter outputs three-phase voltage to control the permanent magnet synchronous motor. Thus, the purpose of adjusting PID parameters on line is realized, and the motor has good dynamic and static performance in a wide rangeof speed.

The invention provides a convolutional neural network-based image geographic positioning system and method, and relates to the field of satellite positioning. The system is characterized by comprising a video uploading module, a picture uploading module, an image frame extraction module, a geographic mark detection module, a decomposition module, a feature extraction module, an activation function, a pooling module, a full connection module, a classifier analysis and judgement module, a learning module, a regional database, a pixel database and a result display module. The system and method provided by the invention have the advantages of rapid in image analysis and correct in image recognition, positioning video information and having learning ability.

The invention relates a device and method having infrared transponding and infrared detection functions. The device comprises a processor, a power module, a wireless transceiver module, a low-power-consumption control module, an infrared receiving module, an infrared transmission module and an infrared detection module, wherein the processor is connected with the infrared receiving module, the wireless transceiver module, the low-power-consumption control module, the infrared transmission module and the infrared detection module, respectively; the power module is connected with the low-power-consumption control module, the infrared transmission module, the infrared detection module and the processor, respectively; the wireless transceiver module is connected with an external gateway; the low-power-consumption control module is connected with the infrared receiving module; the infrared transmission module is connected with an external household appliance. Compared with the prior art, the device and the method are capable of realizing infrared transponding and infrared detection simultaneously; the device is capable of sensing a human body by use of a pyroelectric sensor, has the capability of learning, and further is low in power consumption.

The invention discloses an artificial neuron circuit with a plastic threshold, which comprises an accumulation module, a threshold issuing module and a threshold adjusting module, and is characterizedin that the accumulation module is connected with the threshold issuing module and is used for accumulating signals generated by synapses and sending the signals to the threshold issuing module; thethreshold value issuing module is used for receiving the signals output by the accumulation module and comparing the received signals with a preset threshold value, if the received signals are largerthan the preset threshold value, neurons are activated, and if the received signals are smaller than the preset threshold value, the neurons are not activated; the threshold value adjusting module isconnected with the threshold value issuing module and used for adjusting the size of a preset threshold value according to a learning algorithm. In the training process of the circuit, synapses and neurons are adjustable at the same time, so that the artificial neural network has stronger learning ability, and is richer and more flexible in function and more intelligent.

The invention relates to the field of medical equipment and provides an intelligent medical ultrasonic image processing equipment, a system and a method. The equipment is characterized by comprising an image acquisition device, an image receiving device, an image classifying device, a general image processing device, a first algorithm database, an analytical judgment device, a display device, an ultrasonic image processing device and a second database. The equipment, the system and the method have advantages of low cost, high intelligent level, objectivity and accuracy of results, clearness in image processing, learning ability and the like.

The invention discloses a semi-supervised community discovery method based on maximum clique, comprising the following steps: selecting an initial node based on node degree, looking for each independent clique structure, and forming a clique set in a network; selecting k biggest cliques as initial core communities, and marking the nodes in the initial core communities, wherein the nodes in the same community are marked in the same way, and the nodes in different communities are marked differently; respectively adding MustLink and CannotLink relations according to whether the nodes are marked in the same way or differently, and calculating the membership degree between neighbor nodes; and finally, carrying out reasoning based on the properties of the MustLink and CannotLink relations to get more MustLink and CannotLink relations, and determining the membership of the nodes based on the maximum membership degree between adjacent nodes, thus completing community division. The method can help analyze the structure of a network and find features hidden in the network, and is of important practical significance.

The invention provides a radio frequency positioning system with a learning capability and a positioning method thereof, and relates to the technical field of positioning. The radio frequency positioning system is characterized in that the system comprises a label node, a coordinate reader-writer, a gateway node, a mobile end and a cloud end; the label node comprises a label controller and a label radio frequency transceiver module; the reader-writer comprises a reading-writing storage and calculation module, a reading-writing controller and a reading-writing radio frequency transceiver module; the gateway node comprises a communication interface, a micro-processor and a gateway radio frequency transceiver module; the cloud end comprises a cloud end analysis processor, a cloud end database and a cloud end data transmission module; and the mobile end comprises a mobile end data transmission module, a processor and an operation interface. The radio frequency positioning system has the advantages that the cost is low, the positioning is accurate, the algorithm trap is avoided, the learning capability is realized, and the practicability is high.

The invention discloses an intelligent type barrier avoiding grabbing system based on deep learning. The intelligent type barrier avoiding grabbing system based on deep learning comprises a wireless base station unit, a binocular camera, a three-axis pan-tilt unit, an Mecanum system unit, a manipulator unit, a power source and an industrial computer; the power source is connected with the binocular camera, the three-axis pan-tilt unit, the Mecanum system unit, the manipulator unit and the industrial computer; the industrial computer is connected with an external terminal through the wireless base station unit; the binocular camera, the three-axis pan-tilt unit, the Mecanum system unit and the manipulator unit are all connected with the industrial computer; and the three-axis pan-tilt unitis connected with the binocular camera. The system can automatically recognize and lock a target object, barrier avoiding tracing is combined with a manipulator to grab a target object under a complicated background, the learning capacity is achieved, work can be independently completed, human intervention is not needed, it is achieved that the target object is automatically picked through the manipulator, the applying range of the intelligent type barrier avoiding grabbing system based on deep learning is widened, and the intelligent degree of the intelligent type barrier avoiding grabbing system based on deep learning is improved.

The invention discloses an image retrieval system and method, and relates to the field of image identification. The image retrieval system comprises an image reading module, an image noise elimination module, an image segmentation module, a characteristic quantity calculation module, a screening module, a retrieval module, and a feedback module. The image reading module is used for reading image information; the image noise elimination module is used for eliminating the image noise; the image segmentation module is used for segmenting the image; the characteristic quantity calculation module is used for calculating the characteristic quantity of an unit image; the screening module is used for screening the characteristic quantity; the retrieval module is used for retrieving similar images based on the characteristic quantity; and the feedback module is used for feeding back a retrieval result. Aiming at the problems such as poor accuracy, poor retrieval speed and poor learnability of an existing image retrieval system and method, an improved system is provided; the improved system provided by the invention not only enables the retrieval result to be more accurate and the efficiency to be higher, but also has a learning capability and can improve the retrieval quality continuously.

The invention relates to a universal protocol analysis framework implementation method and device, and the method comprises the steps: S1, building an association relation among a protocol name, a port number and an analysis function, and storing the association relation in a global hash table; s2, acquiring a protocol data message needing to be analyzed; s3, decomposing the protocol data messageneeding to be analyzed, and extracting one or more port numbers from the protocol data message; s4, carrying out searching in a global hash table according to the extracted one or more port numbers, and if an analytical function associated with the one or more port numbers is found, executing the step S5; and S5, analyzing the protocol data message needing to be analyzed by utilizing the searchedanalysis function to obtain an analysis result. The method has the advantages of cross-platform, wide protocol support, simple message analysis, convenient deployment and the like.

The invention provides a trajectory tracking control algorithm for a pneumatic muscle driving system. The algorithm comprises the steps that an expected trajectory of the pneumatic muscle driving system is input; a tracking error is determined according to the feedback actual trajectory of the pneumatic muscle driving system; a tracking error derivative is obtained through the tracking error; according to a hierarchical fusion deep fuzzy neural network, a control signal is obtained through combination with the tracking error and the tracking error derivative; the control signal is transmittedto a pneumatic driving joint; and the pneumatic driving joint conducts trajectory tracking according to the control signal. According to the trajectory tracking control algorithm, through a hierarchical fusion deep fuzzy neural network controller without pre-training, the problems in the existing technical scheme are effectively solved, and the trajectory tracking control of the pneumatic muscle driving system is achieved.

The invention discloses a working method for a safe man-robot cooperation assembly robot and relates to the technical field of intelligent manufacturing. Firstly, a worker set working parameters and instructions through a controller and a touch screen, and a first joint arm and a second joint arm conduct rotational adjustment in the horizontal direction under driving of an alternating current torque electromotor; under driving of a hollow shaft motor inside a connecting sleeve, a screw rod conducts vertical adjustment movement, meanwhile, a vacuum suction nozzle is started to suck air, and a screw is sucked into a locking claw head by a material supplying device through a feeding pipeline and positioned by the vacuum suction nozzle; when moving to a corresponding screw hole position, the screw rod is driven by an electric screwdriver to press rotationally, and the screw is driven by the screw rod to be continuously screwed in the hole position till totally locked in a screw hole; and after the screw is locked, the screw rod moves up to leave the screw hole position. According to the working method, safety and stability are good, highly difficult and fine actions are achieved, programming is easy, and the working efficiency is high.

The invention discloses an operating method of an assembly robot with a monitoring function and relates to the technical field of intelligent manufacturing. The operating method comprises that steps that firstly, operating parameters and commands are set through a controller and a touch screen, and a first joint arm and a second joint arm are driven by an alternating-current torque motor to conduct rotary adjustment in the horizontal direction; driven by a hollow shaft motor inside a connection sleeve, a screw rod starts to make vertical adjustment motion, meanwhile, a vacuum suction nozzle is started to suck air, a negative pressure vacuum loop is formed under the action of a sucker assembly, and a screw is sucked into a locking claw by a supply unit and positioned by the vacuum suction nozzle; when moving to a corresponding screw hole site, the screw rod is driven by an electric screwdriver to be pressed rotationally, and the screw is driven by the screw rod to be continuously screwed into the hole site till being completely locked into the screw hole; and after the screw is locked, the screw rod moves upwards and leaves the screw hole site. The operating method has good safety and stability, is high in operating efficiency, and ensures long-term efficient and stable operation of assembly.