532 results about "Graph recognition" patented technology

An apparatus for recognizing an indoor location using an RSSI (Received Signal Strength Intensity) map, includes an environment information acquirer configured to acquire indoor environment information and store the acquired indoor environment information on a node basis; a numerical map creator configured to create a numerical map; a field intensity creator configured to create the RSSI map; a dead-reckoning sensor configured to locate an area less than several meters within which the environment information acquire is actually located to locate a more accurate location within the area; a path information producer configured to produce path information including movable potential trajectories; and an initializer configured to initialize an initialization location of one or more candidate entities and fingerprint information.

An attribute collector may collect an attribute set for each test entity of a plurality of test entities associated with a software test executed in a software environment. An attribute analysis signal handler may receive an attribute analysis signal associated with a change in the software environment, and a view generator may provide an attribute-based view associated with an affected attribute set associated with the change, the attribute-based view identifying an affected test entity that is affected by the change.

The invention relates to the technical field of image processing, in particular to a 3D skeleton modeling and hand detecting method. The method includes the steps that a figure video is shot through a depth video camera; face detection is carried out in the video; based on face depth information, a body silhouette is extracted; the body silhouette is verified, whether the silhouette is a body silhouette is judged, if yes, the next step is executed, and if not, the face detection step is returned to be executed; image filtering and smoothing are conducted; a skeleton line is extracted by using a detailing algorithm; the distance of body parts is calculated, and a probability distribution diagram is established; all the parts of a human body are recognized; precise joint points are obtained, and coordinates of hands in the 3D space are also obtained; all the joint points are connected to form a complete 3D human skeleton; the 3D human skeleton and the coordinates of the hands are output; the body silhouette is tracked, and information is provided for the next frame. According to the 3D skeleton modeling and hand detecting method, people can be detected rapidly under an existing technical condition, a human skeleton model is obtained, each joint point of people can be accurately positioned, the 3D skeleton is established, the arithmetic speed is high, the computation complexity is low, the 3D skeleton modeling and hand detecting method is suitable for various complicated backgrounds, and only 5ms is needed for each image frame.

The invention discloses a method and system for recognizing and judging a pump indicator diagram. The method comprises the following steps of: pre-processing the pump indicator diagram, and extracting pump indicator diagram information; normalizing the pump indicator diagram according to the pump indicator diagram information; performing polygonal approximation processing on the normalized pump indicator diagram; and recognizing a fault in the pump indicator diagram, which is subjected to the polygonal approximation processing, by using a vector characteristic method. The system comprises a pump indicator diagram pre-processing module, a normalization processing module, a polygonal approximation processing module and a fault recognizing module. By the method and system for recognizing andjudging the pump indicator diagram, provided by the invention, the pump indicator diagram recognition accuracy rate of a deep well can be improved, and the working condition of an oil well is accurately judged.

The invention relates to the field of artificial intelligence, discloses an attention mechanism-based intention recognition method, device and equipment and a storage medium, and is used for improvingthe accuracy of multi-modal intention recognition of information needing to be reasoned. The method comprises the steps of obtaining text intention features of text information and image intention features of image information; respectively calculating a text attention value and an image attention value; according to the text attention value, the text intention feature, the image attention valueand the image intention feature, respectively obtaining a text weighting feature matrix and an image weighting feature matrix; generating attention fusion intention features and gating mechanism fusion intention features according to the text intention features, the image intention features, the text bias feature matrix, the image bias feature matrix and a preset gating mechanism; splicing the attention fusion intention feature and the gating mechanism fusion intention feature to obtain a target intention feature; and performing intention classification on the target intention features to obtain a corresponding target intention.

The embodiment of the invention discloses a graph recognition method, device and system and a terminal device. The method includes the steps of firstly, scanning a target image, and analyzing the target image on the basis of a preset graphic code processing library; secondly, if matched graphic code information is obtained through analysis, determining the target image to be a graphic code and outputting the graphic code information obtained through analysis and matched with the graphic code; if analysis fails, shooting the target image to obtain images to be recognized including the target image, sending the obtained images to be recognized to a server, and receiving content information which is returned after the server recognizes image features of the target image and matched with the recognized image features. By means of the graph recognition method, device and system and the terminal device, software and hardware resource waste caused by switching initiated in the prior art can be avoided, cost is saved, and user time is saved.

The invention provides an intention recognition model training method, an intention recognition method and a related device. The intention recognition model training method comprises the following steps: performing preliminary training on a convolutional neural network by utilizing labeled training sample data; performing data enhancement on the labeled training sample data by adopting a conditional variation auto-encoder to obtain labeled extended sample data; and retraining the preliminarily trained convolutional neural network by using the labeled extended sample data to obtain an intentionrecognition model. Through carrying out data enhancement on training sample data with labels by adopting a conditional variation auto-encoder to obtain expanded sample data with labels; and then, theconvolutional neural network subjected to preliminary training is retrained by utilizing the extension sample data with the labels, so that the model precision can be improved and the labor cost canbe reduced on the basis of utilizing a small amount of accurately labeled training corpora.

The invention provides a method for identifying the complex intentions in the task type multi-round dialogue, and belongs to the field of natural language processing. The method is characterized by defining a task of multi-intention tracking and recognition, introducing a whole set of intention transfer mode set, designing a door structure controller to better utilize the information in the dialogue and identify the intention of the current round of the dialogue in the dialogue process. In addition, the method can predict the next possible intention of the user when the current dialogue intention is finished, and provide the useful information in advance. The mechanism of the prepositive prediction can borrow the information from other related intentions, so that a long dialogue round is avoided to a certain extent. After the dialogue intention of the current round and the potential next round is obtained, the method generates a reply by combining the intention and the information slotaccording to a template library pre-defined manually, so that the more natural dialogue reply result is obtained.

The invention discloses a three-dimensional micro-Doppler gesture identification method based on a convolutional neural network and relates to the fields of man-machine interaction, wireless sensing and image processing, in particular to a method for using the convolutional neural network for identifying three-dimensional micro-Doppler gesture temporal-frequency graphs detected by a three-way radar framework. According to the method, firstly, a three-way placement system framework capable of fully collecting gesture speed information is put forward; an energy window statistic technology is used for continuously extracting effective gesture time-domain signals; by means of temporal-frequency graph synthesis, fusion and processing of three-way temporal-frequency graph information can be achieved at the same time. The convolutional neural network which is subjected to cropping and additionally provided with an SVM layer is designed, image information can be fully extracted, and the identification accuracy is high.

The invention relates to an information pushing method and device based on human-computer interaction and computer equipment. The method comprises the following steps: receiving session information sent by a user terminal, and carrying out preprocessing and word segmentation processing on the session information to obtain a plurality of session texts; inputting the plurality of session texts intoan intention recognition model, and outputting an intention type corresponding to the session information; inputting the plurality of session texts into a trained information extraction model, and calculating the matching degree of the plurality of session texts and a plurality of structured texts in a structured corpus to obtain target field information corresponding to the plurality of session texts; generating corresponding target consultation information according to the intention type and the target field information; and matching corresponding target push information according to the target consultation information, and sending the target push information to the corresponding user terminal. By adopting the method, the accuracy of identifying the structured text in the specific fieldin the session information can be effectively improved, so that the matched information can be accurately and effectively pushed to the user.

The invention provides an intelligent automatic door as well as a graph recognition unlocking method and an automatic control method of the intelligent automatic door, and relates to the field of safety doors. With the adoption of the intelligent automatic door as well as the graph recognition unlocking method and the automatic control method of the intelligent automatic door, the problem that an intelligent door is unsafe to be locked and unlocked with a chip card is solved. A microprocessor is adopted to intelligently control the electrically-control automatic door, a graphic code is drawn through a graph recognition code recognizer, and accordingly, the problem that the chip card is prone to be stolen or the safety is low if a digital code is adopted is solved. A graph recognition code is adopted, a user can set the graph recognition code according to requirements of the user and can draw the graph recognition code during unlocking, the drawn code can be sent to the microprocessor, the microprocessor compares the received graph code with the set graph recognition code, and when the similarity is up to 90% or higher, the microprocessor sends an unlocking drive signal to the electrically-control automatic door. The user can also set a new graph recognition code before going out or set several graph recognition codes simultaneously, so that the graph recognition code is not prone to be revealed stolen, and the safety is high. The intelligent automatic door is applied to a safety door.

A stackable filesystem that transparently tracks process file writes for forensic analysis. The filesystem comprises a base filesystem, and an overlay filesystem. Processes see the union of the upper and lower filesystems, but process writes are only reflected in the overlay. By providing per-process views of the filesystem using this stackable approach, a forensic analyzer can record a process's file-based activity—i.e., file creation, deletion, modification. These activities are then analyzed to identify indicators of compromise (IoCs). These indicators are then fed into a forensics analysis engine, which then quickly decides whether a subject (e.g., process, user) is malicious. If so, the system takes some proactive action to alert a proper authority, to quarantine the potential attack, or to provide other remediation. The approach enables forensic analysis without requiring file access mediation, or conducting system event-level collection and analysis, making it a lightweight, and non-intrusive solution.

The embodiment of the invention provides a defect detection and recognition method and device, a computer readable medium and electronic equipment. The defect detection and identification method comprises the following steps: acquiring a target product image; matching the target product image with a template image corresponding to the target product image to obtain a first defect image of the target product image; performing defect detection processing on the target product image through a neural network model to obtain a second defect graph corresponding to the target product image; and identifying defects contained in the target product image by combining the first defect image and the second defect image. According to the technical scheme, the accuracy of the quality inspection result can be improved, manual quality inspection can be replaced, and the quality inspection efficiency is effectively improved.

The invention provides an intention recognition model training method and device and electronic equipment, and relates to the technical field of artificial intelligence. The method comprises the following steps: determining slot characteristics corresponding to each intention in a training sample set; determining a first intention vector of each sample according to the matching degree of each sample and the slot feature corresponding to each intention; utilizing a first preset encoder to encode the word segmentation vector, the part-of-speech vector and the entity vector corresponding to eachsample, and determining a second intention vector corresponding to each sample; utilizing a preset decoder to decode the first intention vector and the second intention vector corresponding to each sample, and determining a prediction intention corresponding to each sample; and updating a first preset encoder and a preset decoder according to the difference between the prediction intention and thelabeling intention corresponding to each sample. Therefore, through the intention recognition model training method, the accuracy of intention recognition of the deep neural network model under a small-scale training sample is improved.

The invention discloses an intelligent robot integrating chat, knowledge and task questions and answers, and the robot comprises a system interaction module which is used for providing a visual interaction interface and receiving the input information of a user as question input; a dialogue management module which is used for processing dialogue logic, and the dialogue management module comprises a problem rewriting sub-module, an intention recognition sub-module and a problem response sub-module; a dialogue engine module which is used for realizing internal logic of each subsystem, including chat questions and answers, knowledge questions and answers and task questions and answers, and providing algorithm support for the questions and answers; a dialogue data module which is used for storing and managing models and corpora related to each sub-engine and providing data support for a question-answering system; and a system support module which is used for managing configuration files and logs and supporting modular deployment and testing. By judging user intention and processing dialogue logic, questions and answers are obtained through processing of the dialogue engine module and the dialogue data module.

The embodiment of the invention discloses a machine model training method, an intention recognition method and related devices in the field of artificial intelligence. The method comprises the following steps: training a capsule network model according to a training sample, wherein the training process comprises the steps of iteratively adjusting a current weight coefficient corresponding to a first prediction vector according to the similarity between a first activation vector and the first prediction vector, wherein the first activation vector is the weighted sum of a plurality of predictionvectors, and represents the probability that the intention of the training text is predicted as a first real intention; and the first prediction vector represents a contribution of the first semanticfeature to the first real intent. Furthermore, the weight coefficient corresponding to the prediction vector with the large similarity with the first activation vector is increased; therefore, the semantic features corresponding to the prediction vectors with the large similarity with the first activation vectors are screened out, the semantic features corresponding to the prediction vectors withthe small similarity with the first activation vectors are filtered out, the semantic features with the high correlation degree are screened out to form the intention, and the accuracy of intention recognition of the model is improved.

The invention provides an intention recognition method and device, equipment and a medium. The intention recognition method comprises the steps of obtaining a dialogue based on a natural language input by a user; preprocessing the corpus information in the dialogue; and identifying the corpus information by using a pre-trained intention identification model to obtain a corresponding intention. Compared with a traditional intention recognition method, the intention recognition method has the advantages that the intention recognition module is trained in advance, and the obtained dialogue basedon the natural language is input into the intention recognition module to recognize the intention in the corpus information after being preprocessed; the intention recognition module can be adjusted according to scene changes and can be quickly migrated to a new scene; and the application range of the intention recognition module is widened.

The invention discloses a house type image recognition method based on deep learning and image processing, and relates to the technical field of artificial intelligence recognition. The basic image type house type image recognition method is the same as a black and white image type house type image recognition method, and is carried out in two paths, and the method comprises the following steps: S1, obtaining a wall line segment set L1 through a first path; s2, obtaining a window line segment set L2 and a wall line segment set L3 of the outermost contour in a second path; s3, fusing results ofthe wall line segment set L1, the window line segment set L2 and the wall line segment set, and removing repeated line segments; s4, performing linear correction on the image according to the correction method; and S5, extracting a contour of the binary image, and outputting a json file. According to the method, the spatial positions of the house type images are identified for different categories, and the image processing and deep learning methods are combined, so that the house type image identification accuracy is high, the identification speed is high, manpower and material resources aregreatly liberated, and the efficiency is improved.

The invention provides an intention recognition model training method and device and an intention recognition method and device. The training method comprises the steps of obtaining a question pair, aquestion answer pair and an intention category in a dialogue corpus; constructing a question similarity annotation data set according to the question pairs, constructing a question answer similarityannotation data set according to the question answer pairs, and constructing an intention recognition data set according to each round of sentences in the dialogue corpus and category labels corresponding to each round of sentences; constructing a similarity task loss function of sentences according to the question similarity annotation data set and the question answer similarity annotation data set, and constructing an intention recognition task loss function according to the intention recognition data set; and according to the question similarity annotation data set, the question answer similarity annotation data set and the intention recognition data set, carrying out optimization training on the similarity task loss function and the intention recognition task loss function of the sentence to obtain an intention recognition model.

The invention discloses an intelligent question answering system for the automobile field. The system comprises a knowledge base module, a visual interaction module, an intention recognition module, a graph matching module, a template matching module, a retrieval module and an end-to-end module. The knowledge base module stores a knowledge graph and a corpus of the automobile field; after a user inputs a question, the input content of the user is judged, corresponding modules of the system are called for processing according to different judged user purposes, and an answer to the question is obtained. User purposes are divided into automobile field questions and chats, and for the automobile field questions, question answers are obtained by using a question answering method based on an automobile field knowledge graph; and for the chat, an end-to-end module based on deep learning is used to generate an answer. The method can improve the classification precision, and can accurately recognize the user intention.

The invention discloses a man-machine interaction method based on a semantic net and intention recognition, relates to the field of natural language processing, and aims to achieve accurate recognition of question intentions and improve the question and answer quality of man-machine interaction. The method comprises the following steps: obtaining common question answers in the industry as an interactive data source; performing semantic annotation on standard questions in the common question solutions, and constructing an industry semantic network; obtaining a training corpus; training an intention recognition classification model through the training corpus; and receiving a user question, performing intention recognition on the user question through the intention recognition classificationmodel to obtain an intention candidate set, performing multiple rounds of man-machine interaction through the industry semantic network based on the intention candidate set, determining a standard question matched with the intention of the user question, and outputting an answer. The invention further discloses electronic equipment and a computer storage medium.

The invention discloses an individuation mobile phone news system which comprises a motion sensing capturing module, a voice collecting module, a graph collecting module, a character content collecting module, an interactive display screen, an intelligent analysis and recognition module, a content filtering and classifying module, a hobby data base, an action data base, a content knowledge base, an offline content source, an editing content and an online content. According to the individuation mobile phone news system, by means of a voice recognition technology, a graph recognition technology and a character content understanding technology, action (body aspect) and thought (spirit aspect) data of readers are captured, and contents which meet requirements of the readers are interacted with the readers by means of an intelligent analysis technology to achieve accurate content transmission, so that the problems of homogenization and non-individuation of mobile phone news are thoroughly solved, the individuation of mobile media is really achieved, valuable time and download web traffic of the readers are greatly saved, and the waste of bandwidths is prevented.

The invention relates to the technical field of intelligent questions and answers, in particular to a Bert model-based intention recognition and slot value filling joint prediction method, which comprises the following steps of: inputting a target text to obtain a word vector, a segment vector and a position vector of the target text, splicing the word vector, the segment vector and the position vector as an input vector of a Bert model, and performing prediction on the input vector of the Bert model; inputting a trained Bert model, outputting an intention representation vector and a slot value sequence representation vector by the trained Bert model, performing weight calculation on the intention representation vector and the slot value sequence representation vector in a Gate layer to calculate a joint action factor, acting the joint action factor on the slot value sequence representation vector, and finally outputting predicted intention classification and a slot value sequence. According to the method, a Gate mechanism is used on a Bert layer, the internal relation between intention recognition and slot value filling is fully utilized, and the task error prediction rate is reduced.

The invention relates to the technical field of artificial intelligence. The invention provides an intention recognition method, device and equipment and a computer readable storage medium, and the method comprises the steps: carrying out the word vector feature extraction of an unmarked text through a first language model, obtaining an unmarked feature, carrying out the marking of the unmarked text according to the unmarked feature, and obtaining a marked training text; constructing an attention neural network model based on a second language model and the annotation training text; obtaininga to-be-recognized text, and performing feature extraction on the to-be-recognized text through the attention neural network model to obtain a candidate feature set; and calculating the similarity ofthe to-be-recognized text according to the candidate feature set, and judging whether the to-be-recognized text corresponds to the same expression intention or not according to the similarity to obtain an intention recognition result. According to the method, feature extraction is carried out in a neural network mode, text features can be considered more comprehensively, and the accuracy of intention recognition is improved.

The invention discloses a knowledge graph-based question and answer intention recognition method. The method comprises the steps of constructing a domain topic dictionary; constructing a template; performing part-of-speech analysis and syntactic dependency analysis on the retrieval statement; calculating the similarity between the search text and the template sample by utilizing the word vector and an LDA algorithm; on the basis of a known word vector, carrying out wider intention recognition on the basis of TextCNN to serve as a result of open information; extracting keywords from the domainmap. According to the method, intention recognition is achieved by integrating multiple methods, comprehensive retrieval of accurate answers and related information is achieved in combination with theknowledge graph of the field, and various requirements of users are met.

The invention discloses a social network model construction module of a company image improvement system. The social network model construction module comprises the following five sub-modules: construction of a complex social network user model, construction of an inter-user relationship module, construction of a multi-source heterogeneous complex social network topological graph, identification of key nodes, discovering and dividing of communities, wherein the construction of the complex social network user model comprises user data extraction and user attribute feature definition; the construction of the inter-user relationship module comprises user relationship extraction and potential relationship prediction; and the identification of the key nodes comprises user node importance indexes and event propagation node importance indexes. According to the invention, related data on social media is collected efficiently; the complex social network user model is constructed on the basis ofacquired data; meanwhile, a specific relationship among users is modeled; a one-way edge model among the users is constructed; a complex social network topological structure model is comprehensivelyobtained; and the complex social network topological structure model is taken as an object.

The invention discloses an intention recognition method, device and equipment, and a storage medium. The intention recognition method comprises the steps of receiving to-be-recognized text information; segmenting each statement in the text into statement sub-fragments with different semantics respectively; extracting a keyword of each statement sub-segment; inputting the statement sub-segments andthe keywords extracted from the statement sub-segments into a text classification model for classification to obtain a classification result; recombining the statement sub-fragments of which the classification result is the specified category in sequence to obtain a recombined text; and performing intention recognition on the recombined text. According to the intention recognition method providedby the invention, the accuracy and efficiency of text intention recognition are improved.