441results about How to "Achieve training" patented technology

The invention discloses an execution function evaluating and training system for autism spectrum disorder children, wherein a problem of evaluating and training the execution function of the autism spectrum disorder children is settled by means of advanced man-machine interaction technology and brain information sensing technology. The execution function evaluating and training system comprises the components of a student intelligent terminal, a portable electroencephalogram, a portable set-top-box, a cloud server and a parent handheld smart phone. The functional units are connected through wireless communication technology, wherein the student intelligent terminal provides a user interaction interface and records a user face expression image and spatiotemporal dynamics behaviors of buttons in real time. The portable electroencephalogram records the brain wave rhythm of the user. The data set-top-box is a system data collecting and analyzing center and analyzes the collected data through an algorithm for realizing sensing and estimating of the execution function on the user. Furthermore the man-machine interaction interface is constructed for realizing execution function training based on the analysis. The parent handheld smart phone acquires the execution function evaluating and training result of a child through the cloud server.

The invention provides an image processing model training method and device and a storage medium, an image processing model comprises a backbone network, a regional candidate network and a detection network, and the method comprises the steps: carrying out the feature extraction of a sample image containing a target object through the backbone network, and obtaining a feature map of the sample image; performing region selection on the feature map through a region candidate network, and determining a candidate region; performing target object detection on the candidate area through a detectionnetwork to obtain category parameters and position parameters of a target object, the category parameters including classification results corresponding to the target object, and the position parameters including a bounding box, a segmentation mask and a key point mask corresponding to the target object; determining a value of a target loss function of the image processing model based on the category parameter and the position parameter of the target object; and updating the model parameters based on the value of the target loss function. According to the invention, the target object in the image can be accurately positioned, and the target detection precision is improved.

The invention discloses a dialogue generation model training method, which comprises the following steps: obtaining a dialogue data set, the dialogue data in the dialogue data set comprising questionsand annotation replies corresponding to the questions; based on the questions in the dialogue data set, obtaining the coded representation of the question through an encoder in a constructed dialoguegeneration model; fusing the coded representation of the question and knowledge information of transfer learning of the question from a knowledge base question and answer model through a decoder in the dialogue generation model to obtain a prediction reply corresponding to the question output by the dialogue generation model; and determining a loss function based on the prediction reply and the annotation reply corresponding to the question, and adjusting parameters of the dialogue generation model through the loss function until the loss function of the dialogue generation model is convergent. According to the method, the knowledge information can be better fused by the model, so that the dialogue generation accuracy and reasonability are improved. The invention further discloses a dialogue generation method and device, equipment and a medium.

The invention discloses an auxiliary training tool for comprehensive rehabilitation of cerebral palsy. The auxiliary training tool comprises a chassis, a lower limb exoskeleton device and a suspension device, wherein the chassis comprises universal casters with the locking function, a lower chassis, an upper chassis and an exoskeleton device inclination adjusting mechanism; the lower limb exoskeleton device is fixedly mounted on the chassis, and can realize rehabilitation training such as flexion/extension, adduction/abduction and pronation/supination of hip joints, flexion/extension of the knee joint, and dorsi-flexion/plantar-flexion, strephenopodia/ strephexopodia and pronation ans supination of feet; the suspension device is fixedly mounted on the chassis, and plays a role in slinging, weight reduction and upper body posture correction of paitents. The auxiliary training tool can help patients with cerebral palsy to correct the abnormal posture of lower limbs, carry out the rehabilitation training on all joints of the lower limbs, promotes the development of all the joints of the lower limbs, and gradually establishes the normal standing and walking functions.

The invention discloses a three-dimensional stress cell culture device capable of applying a dynamic load. The device comprises a shell with a liquid outlet and a liquid inlet, a clamping device which is arranged in the shell and used for clamping a biological structure, a driving motor for providing a driving force to the clamping device, a heating device for heating a culture solution in the shell, a temperature sensor for detecting the temperature of the culture solution and outputting temperature signals, a tension sensor for detecting the tension of the biological structure and outputting tension signals, and a controller for receiving the temperature and tension signals and outputting control signals to the driving motor and the heating device according to the received signal information. The device disclosed by the invention is simple in structure, low in manufacturing cost and easy to operate. The device can carry out a force-controllable reciprocating stretching operation on a biological structure according to a set program to simulate an in vivo environment so as to culture cells through operations on a liquid crystal display screen after the biological structure mixed with cells is arranged in the cell culture device.

The invention relates to the technical field of computer processing, and discloses a training method of a text word vector model, electronic equipment and a computer readable storage medium. The training method of the text word vector model comprises the steps: determining a sub-model corresponding to each training statement according to statement tags included in each training statement, the statement tags being used for indicating the sub-models corresponding to the training statements; Respectively training the corresponding semantic word vector quantum model and the text classification sub-model through each training statement to update a first word vector matrix of the text word vector model, so as to train the text word vector model by updating the first word vector matrix. Accordingto the method provided by the embodiment of the invention, through the combination of the semantic word vector quantum model and the text classification sub-model, the close connection and fusion between the word vector training method and the text classification method are realized, and the representation capability of the word vector is enhanced.

The invention discloses a transesophageal echocardiography visual simulation system, comprising an intelligent phantom, a heart ultrasonic image stimulation and acquisition device, a computer connected with the heart ultrasonic image stimulation and acquisition device and a transesophageal ultrasonic probe attitude data acquisition device, wherein, the heart ultrasonic image stimulation and acquisition device is used for stimulating and acquiring heart ultrasonic images, and transmitting the acquired heart ultrasonic images to the computer; when the heat ultrasonic image stimulation and acquisition device acquires the heart ultrasonic images, the transesophageal echocardiography probe attitude data acquisition device and a sensor are used for synchronously acquiring probe attitude data of the heat ultrasonic image stimulation and acquisition device, and transmitting the acquired probe attitude data to the computer; and the computer is used for analyzing, processing and visually displaying the received heart ultrasonic images and the probe attitude data.

The invention provides a classification model training system and an implementation method thereof. The implementation method comprises the steps of S1, establishing a front end management display interface of a SPARK algorithm classification model training system, and defining a front end interactive request interface of the front end management display interface; S2, establishing a rear end service data source system of the SPARK algorithm classification model training system; S3, based on the front end interactive request interface of the front end management display interface, establishinga rear end service control interface, and establishing a corresponding relation of the rear end service control interface and the front end interactive request interface; S4, establishing internal business logic based on a SPARK algorithm training and optimizing classification model in the rear end service control interface. According to the classification model training system and the implementation method thereof, the operation procedure of classification model training can be effectively simplified, accordingly the labour intensity of developers is effectively reduced, and the developing efficiency is improved.

The invention belongs to the field of computer vision and pattern recognition, in particular to a pedestrian re-recognition method based on reinforcement learning adaptive partitioning. The inventionintroduces a reinforcement learning method to train an intelligent network, so that the intelligent network can be driven by the data itself, and adaptively decides a partition strategy (the number ofpartitions and the size of partitions) according to the characteristics of the retrieval pedestrian. Experiments show that the invention not only improves the flexibility and generalization ability of the model, but also has a considerable application prospect.

The invention discloses a face recognition algorithm training method based on multi-objective learning, comprising the following steps: randomly initializing neural network parameters, using a loss function based on face identity and a loss function based on face feature point position constraint to minimize the learning objective to train the deep convolution network; when the prediction accuracyrate of the face identity reaches the threshold, calculating the loss function based on the distance within the face feature classes and the loss function based on the distance between the face feature classes, and performing calculation of the loss function based on the face identity and the loss function of the face feature point position constraint on each sample; based on the artificial weight setting, weighting each loss function to obtain the total loss function, and realizing the back propagation based on the total loss function. Therefore, the update of the network parameters is realized, and the network training is stopped when the accuracy is stable to get the trained face recognition model.

The invention relates to a radar and multi-network fusion based pedestrian pose recognition method and system. The method comprises: pre-processing an echo signal of a radar signal to obtain an outputsignal; suppressing a stationary target in the output signal; searching a pedestrian located distance unit in the suppressed output signal; performing time-frequency analysis on the echo signal corresponding to the pedestrian located distance unit, and obtaining a an echo signal time-frequency map; using a plurality of convolutional neural networks to recognize the echo signal time-frequency map,and obtaining a recognition result of each convolutional neural network; and fusing the recognition results of all the convolutional neural networks, and obtaining the fused pose recognition result.The pedestrian pose recognition method provides a pedestrian pose recognition result in real time by analyzing the radar echo, can ensure high pose recognition accuracy through multi-neural network fusion, is free from interference of factors such as illumination conditions, weather, smoke and the like, and is able to work all day and all weathers.

The invention discloses a bearing fault diagnosis method combining an improved sparse filter and a KELM. The method comprises the following steps: embedding a Min-Max regular term into an original sparse filter to obtain an improved sparse filter. The Min-Max regular term can describe the internal structure information of the original data, and promotes the similar samples to be close to each other and promotes the samples of different classes to be separated from each other, thereby generating discriminative characteristics. Feature discrimination mainly lies in that class label information is used in construction of the Min-Max regular term, and a pseudo label is used for replacing a real label to guide the construction of the Min-Max regular term. Vibration signals of different operation conditions of a rolling bearing are collected to serve as a training set, the training set is used for training an improved sparse filter model and a kernel extreme learning machine model to obtainmodel parameters, and therefore establishment of a fault diagnosis classification model is completed, and the diagnosis model can accurately identify the rolling bearing fault.

The invention discloses a bionic rehabilitation system worn in the whole movement process of a human body. The bionic rehabilitation system mainly comprises a human body movement target joint moment resolving system, a patient muscle force self-supply joint moment resolving system and an external assistant excitation compensation moment calculation system. A mode that limbs of a patient are drivenby an external assistant mechanism of conventional rehabilitation equipment to achieve rehabilitation movement is changed. By adopting a lightened wearable use mode, in the situation that conventional muscle force functions of a patient are sufficiently protected, on the basis of bionics and a healthy body kinematic mechanics model, and by using multi-signal sensation fusion and resolving methodswith inertia signals, sole six-dimensional force signals, surface electromyogram signals, functional electrophotoluminescence signals and the like, space muscle force systems and self-supplied jointmoments of the patient the whole movement process of sitting, standing and walking are analyzed, compensation moments of different joints are calculated in real time, corresponding muscle is stimulated to exert force through the functional electrophotoluminescence signals, and rehabilitation movement assistance and training in the whole movement process of a human body can be achieved.

The invention discloses a 3D culture, passage, cryopreservation, recovery and identification method for in-vitro organoids based on small intestines of different segments of mice. The method comprisesthe following steps: (1) separating recesses of duodenum, jejunum and ileum segments of mice; (2) performing 3D culture on the recesses of the duodenum, jejunum and ileum segments of mice, and forming organoids; (3) performing passage on organoids of small intestines of mice; (4) performing cryopreservation on the organoids of small intestines of mice; (5) performing recovery on the organoids ofsmall intestines of mice; (6) preparing frozen sections of the organoids of small intestines of mice; and (7) performing immunofluorescence labeling and identification on the frozen sections of the organoids of small intestines of mice. Compared with the prior art, the method disclosed by the invention optimizes a recess extraction manner, an in-vitro culture system and a culture medium based on the recesses of the small intestines containing stem cells, and comprises the steps of complete culture, passage, cryopreservation, recovery and identification. The small intestine organoids obtained by the method are capable of performing repeated passage, and the passage organoids have character stability.

The invention provides a target tracking state model training method and device. The method comprises the steps that the known trajectory information of a target in a video sequence in different states and the tracking state model of the target are acquired; the tracking state model is used to track the target in the video sequence, so as to acquire trajectory information outputted by the tracking state model in different states; trajectory information in a state is selected from trajectory information outputted by the tracking state model in different states; according to the trajectory information in the state and the known trajectory information of the corresponding state, whether the tracking state of the target is correct is determined; and if the tracking state of the target is incorrect, the tracking state model is updated. According to the invention, the tracking state model is trained by constantly updating the tracking state model through the known trajectory information of the target in different states, so as to realize automatic conversion of the tracking state of the target in different cases.

The invention discloses a visibility function direct inversion method of an atmospheric temperature and humidity profile. The method comprises the following steps: inputting an actual observed visibility function into a trained BP neural network, so as to obtain an inversed atmospheric temperature and humidity profile physical parameter, wherein the training of the BP neural network comprises: using a numerical weather mode to obtain a sample atmospheric physical parameter, and using an aperture synthesis microwave radiometer model to carry out forward modeling on the sample atmospheric physical parameter to obtain a visibility function corresponding to the sample atmospheric physical parameter; and using the sample atmospheric physical parameter and the visibility function corresponding to the sample atmospheric physical parameter to train the BP neural network, so as to obtain the trained BP neural network. The invention changes a two-step inversion method of visibility-brightness temperature-atmospheric physical parameters required for remote sensing atmospheric physical parameters of a current aperture synthesis microwave radiometer, and proposes a direct inversion method of visibility-atmospheric physical parameters, which simplifies the data processing flow and reduces inversion errors, so as to obtain more accurate atmospheric physical parameters.

Disclosed are an ammonia-oxidizing bacterium screening and testing reagent box, a screening and testing method and a bacterium sample separating method. The ammonia-oxidizing bacterium screening and testing reagent box comprises a bacterium strain enrichment tube, ammonia-oxidizing testing tubes 1, 2 and 3, bacterium strain preservative liquid and color developers A and B. The screening and testing method includes inoculating bacterium strains to be detected to the ammonia-oxidizing testing tubes 1, 2 and 3 to culture after enriching the bacterium strains to be detected via the bacterium strain enrichment tube, and storing enrichment bacterium liquid by the bacterium strain preservative liquid simultaneously; adding the color developer A and B after culturing of the ammonia-oxidizing testing tubes and judging ammonia-oxidizing capability of the bacterium strains according to color development states. The bacterium sample separation method includes culturing bacterium samples, diluting, applying and separating. The ammonia-oxidizing bacterium screening and testing reagent box, the screening and testing method and the bacterium sample separating method has the advantages that effect of quickly screening and testing ammonia-oxidizing bacteria in high throughput is realized, one-stop operations of culturing, testing and preserving of the bacterium strains are realized.

The invention discloses a speech synthesis method and device, a model training method and device, equipment and a storage medium, and relates to the technical field of artificial intelligence such asmachine learning and intelligent speech. The specific implementation scheme is as follows: based on text information, timbre information and rhythm information of to-be-synthesized speech, generatingacoustic feature information of the to-be-synthesized speech by adopting a pre-trained speech synthesis model; based on the acoustic feature information of the to-be-synthesized voice, synthesizing the corresponding voice by adopting a pre-trained vocoder. By adopting the technical scheme of the invention, when the voice is synthesized, any text information, tone information and rhythm informationcan be randomly combined and synthesized into the desired voice, and the voice synthesis mode is very flexible and convenient.

The invention discloses a zero parallel corpus multi-modal neural machine translation method, which comprises the following steps: pre-training a pre-constructed neural machine translation model by utilizing a source language with corresponding picture information and a target language monolingual corpus; translating sentences in the source language into sentences in a target language by using a pre-trained neural machine translation model; calculating a reward value of a sentence level according to an internal relation between the sentence obtained by translation and a corresponding picture, and updating pre-trained neural machine translation model parameters by adopting a strategy gradient reinforcement learning method by taking the maximum expected total reward value as an optimization target so as to obtain a trained neural machine translation model; and translating the given source language sentence by using the trained neural machine translation model. According to the zero parallel corpus multi-modal neural machine translation method, a neural machine translation model of a zero resource language pair can be established by utilizing the internal relation between pictures and characters.

The embodiment of the invention discloses a clustering model training method and device, electronic equipment and a computer storage medium. The method includes the steps that clustering is conductedon new photos through a clustering model and clustered photos to obtain a clustering result of the new photos, wherein the new photos carry class labels; the return function value of the clustering result is calculated based on the clustering result of the new photos and the class labels; the clustering model is trained according to the return function value of the clustering result. According tothe embodiment, the new photos and photos in an initial-state photo album with a classifying result are clustered through the trained clustering model, the obtained clustering result is closer to a manual classifying result, and the clustering accuracy of the trained clustering model is higher.

The embodiment of the invention discloses a voice synthesis model training method. The method comprises the steps of: carrying out voice synthesis on text data based on an initial voice synthesis model to obtain a synthesized voice, carrying out emotion recognition on the synthesized voice based on a speaker classification network to obtain a first feature vector, performing emotion recognition onreal person voice corresponding to the text data based on the speaker classification network to obtain a second feature vector, comparing the first feature vector with the second feature vector, andupdating the network parameters of the initial voice synthesis model based on a comparison result to obtain a target voice synthesis model. Emotion recognition is carried out on the synthesized voicedata and the real person voice data through an emotion recognition network, the network parameters of the initial voice synthesis model are updated according to a feedback result, training of the initial voice synthesis model is completed, a target voice synthesis model is obtained, and training of the voice synthesis model is accurately realized.

The invention belongs to the field of biomedical engineering, and discloses an apparatus and a method for in vitro storage and culture of an organ. According to the present invention, the human body blood circulation system is simulated to provide the in vitro organ environment close to the in vivo organ environment; the in vitro organ environment comprises the constant temperature and the circulation system; the circulation system comprises the oxygenator, the pump, the pressure control device, the organ placement box, the dialysis device, the liquid storage chamber and the agent feeding device; through the pressure control device, the perfusion pressure and the sufficient perfusion amount are maintained when the organ is subjected to the perfusion with the perfusion liquid, such that theorgan can be store in vitro for a long time and can maintain the health, and the culture of the in vitro constructed organ and the collection of the organ secretions are achieved; and the apparatus for reasonably achieving the in vitro organ storing and culturing and the technical scheme thereof are provided.

The invention discloses a pulsation heart model and a pulsation control method thereof. The pulsation heart model comprises a simulated heart model, a pressure sensor, a display device, a control unit, an electric water pump, a man-machine interaction interface and a liquid storage tank. The liquid outlet of the liquid storage tank is communicated with the liquid inlet of the simulated heart model, the liquid outlet of the simulated heart model is communicated with the return port of the liquid storage tank, the electric water pump is installed on an inlet pipe, the electric water pump is connected with the control unit, the pressure sensor is used for detecting the liquid pressure, the pressure sensor is connected to the display device, and the man-machine interaction interface is connected to the control unit and the display device. The pulsation heart model has high simulation degree, can simulate the heart pulsation state under different pathological conditions, and can record anddisplay various physiological signals such as cardiac blood pressure, electrocardio, heart sound, heart rate, and shape variable through the display device in the form of numerical values or charts, which makes it easier to train operators.

The invention provides a method for detecting railway foreign matter invasion in a special scene based on transfer learning. The method comprises the following steps: S1, collecting data; S2, data expansion: converting the image into images of different domains by using an image style converter to realize migration of an image style; S3, data processing; S4, target detection and tracking. Firstly,aiming at the problem that training samples are too few in severe weather environments and night scenes, conventional scene samples are migrated to corresponding scenes, and a railway scene target sample library is established under different weather and time conditions; and secondly, based on a deep convolutional neural network, through combination of a target detection SSD algorithm and a CAFFEframework, detection of an abnormal target in a railway scene is realized, and through combination with the generated complex sample library, model training under complex weather conditions based ontransfer learning is realized, and target detection, tracking and behavior analysis of the model are utilized.

The invention relates to a plant tissue liquid automatic intermittent immersion circulation cultivation device, which is characterized in that at least two groups of cultivation tanks are arranged around the nutrient solution tank, and the nutrient solution tank communicates with the cultivation tank through a high-temperature and high-pressure sterilized diversion tube to cultivate Gas concentration sensors, temperature sensors and humidity sensors are installed on the side wall of the tank, solenoid valve A and solenoid valve B are arranged on the upper and lower sides of the programmable integrated circuit board, and the reinforced polychloride on the left side of the programmable integrated circuit board A pressure regulator is set on the vinyl tube, the right side of the air pump is connected to the pressure regulator through a line, the lower end of the air pump is connected to the programmable integrated circuit board through a line, and an air filter is set on the reinforced polyvinyl chloride tube near the nutrient solution tank The front end of the branch pipe on the right side of the reinforced polyvinyl chloride pipe is provided with an air filter. The invention realizes liquid automatic culture, reduces dependence on agar in plant tissue culture, and reduces labor intensity.

The invention discloses a virtual screen shooting game system. The virtual screen shooting game system comprises a user host computer capable of installing and running software, a display device capable of displaying a game picture, a handheld light emitter capable of emitting light, a motion sensing helmet and a motion sensing garment, wherein the display device, the handheld light emitter, the motion sensing helmet and the motion sensing garment are all in signal connection with the user host computer. A light signal emitted by the light emitter is acquired through a photosensitive sensor array in real time to realize accurate positioning; the body movement information of a user is acquired through a position detection device of the handheld light emitter in real time, so that the reality and physical training capability of a shooting game are improved; the vibratory sensation after shooting can be generated through the motion sensing helmet and the motion sensing garment, so that the immersivity and the fidelity are achieved, and the system is particularly suitable for virtual shooting competitive games, military training simulation of armies and polices and the like.