51 results about "Deep learning" patented technology

The invention discloses an automatic classification method for electrocardiogram signals. The method is achieved according to the following steps of firstly, obtaining electrocardiogram signals of a human body, conducting filtering on the electrocardiogram signals, and detecting R waves of the electrocardiogram signals where filtering is conducted; secondly, establishing a data set after the R waves are detected, wherein the data set is composed of multiple sets of cardiac beat data, and each set of cardiac beat data has a label; thirdly, establishing a sparse automatic coding deep learning network; fourthly, training the sparse automatic coding deep learning network step by step; fifthly, inputting the to-be-measured cardiac beat data into the sparse automatic coding deep learning network according to the network weight, obtained in the fourth step, of the first hidden layer, the network weight, obtained in the fourth step, of the second hidden layer and the network weight, obtained in the fourth step, of the softmax classifier so as to obtain cardiac data which are output in a classified mode. The sparse automatic coding deep learning network is applied to the classification of the cardiac beat data, and by means of the autonomous leaning capacity and the deep characteristic excavation characteristic of the sparse automatic coding deep learning network, deeper characteristics of signals are extracted, and the cardiac beat data are classified.

The invention discloses an emotion recognition method and system based on a deep learning model and a long-short memory network, The method comprises the following steps: data preprocessing and data set partitioning of EEG signals are performed to construct a network model, wherein the network model comprises a picture reconstruction model composed of a variational encoder and an emotion recognition model composed of a long-short memory network; the network model comprises an image reconstruction model composed of a variational encoder and a short-long memory network; The objective function isconstructed according to the network model. The network model is trained by training set, and the objective function is optimized by Adam optimizer in neural network, and the trained network model isobtained. Using the cross-test set to cross-test the trained network model, determining the super-parameters of the network model, and obtaining the final network model; and using the final network model to visualize the seed data and perform emotion recognition. The invention relies on the data artificial intelligence method to learn the collected EEG signal space and time complex structure, reduce the subjective factors in the prediction, and improve the prediction accuracy.

The invention discloses an image processing method and device. The image processing method comprises the following steps: reading a to-be-synthesized image; reading a to-be-cutout image, performing RGB channel splitting on the to-be-cutout image and respectively calculating edge pixel average values of the three channel images; performing binarization processing on the images of the three channelsby using the average values, and combining binarization images of the three channels to obtain a superposed binarized image so as to record target profile information of the binarized image; readingthe target profile information to generate a position of a synthetic image in the to-be-synthesized image; putting the to-be-cutout image on the position of the synthetic image so as to generate the synthetic image, and saving target categories and the generated positions as tag files, wherein numbers of the target object categories and names of the target object categories in the to-be-cutout image are configured in advance, and the numbers of the categories correspond to the names of the categories. According to the image processing method and device disclosed by the invention, by utilizingimage cutout and synthesis, image data required to be learned deeply are acquired and automatic generation of the corresponding labels is realized.

The invention discloses a parking space detection method based on depth learning, comprising the following steps: obtaining a target image of the parking space; setting a training set and a test set of the parking space target image; a tag being arranged for that target region of interest, and the coordinate and categories of the target region being marked; initializing neural network parameters;putting the training set and tag information into the neural network to participate in training; quantitative evaluation of training results. A parking space detection system based on depth learning comprises an image acquisition module, an initial setting module and a training control module. The invention effectively solves the shortcomings of traditional image method that the parking space environment is highly demanded, and has strong adaptability, and the characteristic points of the parking space can be perfectly detected and positioned even in extreme environment, which is convenient for automatic and accurate parking alignment of vehicles, has the advantages of good robustness, and improves the user experience at the same time. The method and system can be widely used in the fieldof parking space recognition.

The invention discloses an automatic electrocardiosignal analyzing and processing method based on deep learning. The method comprises the operation steps of: preprocessing an original electrocardiosignal, segmenting a fixed-length waveform to sort electrocardiosignal data into vectors with a plurality of consistent dimensions, and training the sorted data through a deep learning model of a full connected neural network or a full convolutional neural network; and finally, comprehensively and accurately outputting different electrocardio index categories. By means of the method, the workload ofdoctors during interpretation can be reduced, the doctors are helped to improve the interpretation accuracy, and the method becomes an important auxiliary scheme for doctors to diagnose heart diseases.

The invention discloses a depth learning method based on a block chain platform and an electronic device. The method comprises: a receiving task publisher selects a depth learning model on the block chain platform; a receiving task publisher selects a depth learning model on the block chain platform. Receiving a training set input by the task publisher; generating a depth learning task according to the training set and the depth learning model; distributing the depth learning task to a computing node in the block chain platform; depth learning is performed on the depth learning task by a computing node to which the depth learning task is distributed. The scheme solves the existing technical problems of high cost of deep learning and low training efficiency, and achieves the technical effect of reducing training cost and improving training efficiency.

The invention relates to an interest point check-in prediction method fusing deep learning with a factorization machine and belongs to the field of location check-in prediction. The method comprises the following steps: S1, acquiring check-in data of a user; S2, performing embedding processing on input discrete data; S3, performing sparse elimination processing on the discrete data, and learning implicit second order relations among the data; S4, learning addition of continuous characteristics into a full connection hidden layer, and selecting an appropriate excitation function; S5, inputtinga result obtained by processing discrete characteristics and a result obtained by processing the continuous characteristics and adding the results as an input of a hidden layer h1; S6, enabling an output l1 of the hidden layer h1 to pass a first-order linear and characteristic interaction structure and adding as an input of a hidden layer h2; and S7, receiving an input by a hidden layer h3 from outputs l1 and l2 of the hidden layers h1 and h2, adding a shortcut structure at the same time for guaranteeing gradient stability during parameter learning, determining the best model structure, and finally outputting a prediction result. The method provided by the invention fully excavates and learns check-in rules and predicts interest point check-in problems by analyzing check-in information ofthe user.

The invention discloses a deep learning method for predicting urban traffic jam based on GPS, and relates to the field of intelligent traffic management. Through the deep learning method, a unsupervised large-scale urban network traffic jam prediction method (CRC3D) is innovatively constructed based on GPS real-time data on urban taxies. The method is characterized in that a recursion neural network (RNN), a convolutional neural network (CNN) and a three-dimensional convolutional neural network (C3D) are combined. The recursion neural network and the three-dimensional convolutional neural network can efficiently capture time-varying characteristics of continuous traffic states from different angles. The method is good in effect and high in practicability.

The invention relates to a smoke and fire detection method based on deep learning and image recognition. The method comprises the following steps: step 1, reading a video frame from a to-be-identifiedvideo, and extracting a part of pixel sets in the video frame according to color characteristics of pixels in the video frame and difference characteristics of the pixels between the video frames toform one or more pre-processing areas for identification; Step 2, intercepting the extracted preprocessing area into pictures to serve as input of a smoke and fire detection model; Step 3, extractingfeatures from the picture through the smoke and fire detection model; 4, determining the confidence coefficient of the smoke and fire scene in the picture according to the extracted characteristics; and step 5, if the confidence coefficient exceeds a confidence coefficient threshold, marking an area corresponding to the picture in the video frame.

The invention discloses an elevator monitoring key frame acquisition and detection method based on deep learning. The method comprises a monitoring video acquisition module, a monitoring video preprocessing module, a monitoring video key frame interception module, a monitoring video post-processing application module and a user application module. The elevator monitoring key frame obtaining and detecting method based on deep learning is simple in design and capable of completing monitoring tasks of various scenes.

The invention provides a pipeline magnetic flux leakage detecting data automatic analysis system. The pipeline magnetic flux leakage detecting data automatic analysis system comprises a pipeline magnetic flux leakage detector, a client end and a server, and the pipeline magnetic flux leakage detector collects pipeline magnetic flux leakage detecting data and transfers to the server. A data generator, a trainer, a reasoning machine and a predefined self-learning model are arranged on the server, feature extraction is automatically carried out and mode recognition, defect classification and defect quantization are automatically carried out through deep learning artificial intelligence technology, and labor investment is greatly reduced.

The invention discloses a method for compressing gradients for distributed depth learning, which utilizes the entropy of gradient matrix calculated by each layer to judge the amount of information contained in the gradient of this layer, so as to judge the importance of each layer gradient. Then, entropy is used to calculate the percentage of gradient that each layer needs to send, which can notonly reduce the influence of human experience, but also effectively compress the gradient without degrading the accuracy.

The invention discloses a clutch main cylinder groove surface defect detection method based on deep learning. Deep learning and defect recognition are combined to achieve high-precision and high-efficiency automatic detection. A multi-scale feature fusion pyramid network is adopted to extract defect features more accurately, and autonomous learning is carried out; a single-target FCOS convolutional neural network is adopted as a detection network model, dependence on an anchor box and a feature candidate region is avoided, the network model is simplified, and the network training speed is increased; experimental results show that the effect of the defect detection algorithm based on deep learning is superior to that of a traditional manual detection method and a machine vision detection method, and the time requirement and precision of defect detection of the clutch main cylinder inner groove on an industrial production line are met; defects generated in the production process of the clutch main cylinder inner groove can be rapidly and accurately detected, the production efficiency of a factory is improved, and the practical significance is achieved.

The invention provides an ENSO forecasting method based on deep learning. The ENSO forecasting method comprises the following steps: acquiring a plurality of physical element data observation sets from different ways; processing the physical element data observation sets, so that the data formats of all the physical element data observation sets are the same; pre-extracting a plurality of physicalelement fields in the physical element data observation set to obtain training features; through a deep learning algorithm, importing each element physical quantity in the training features into themodel for training to obtain a forecasting model; and importing the real-time observation data into the forecasting model to obtain a forecasting result. According to the ENSO forecasting method basedon deep learning provided by the invention, effective information hidden in a large amount of data is mined to the greatest extent in a manner of combining deep learning, and the problem of relatively large ENSO forecasting deviation in the prior art is effectively solved, so that more accurate forecasting is realized, and the forecasting precision is improved.

The invention discloses an industrial control safety auditing system and method based on artificial intelligence. The system comprises a safety auditing terminal, a central control terminal and an artificial intelligence learning terminal. According to the invention, the security auditing terminal is arranged to monitor and record the network state, the intrusion behavior and the operation recordrespectively; loopholes and malicious attacks are protected in real time, and when a high-risk security condition occurs, data storage and service interruption are carried out immediately, and an alarm is given; the security of the industrial control network is improved, so that the industrial control network meets the industry compliance auditing requirement; by arranging the central control terminal, operation behaviors in an industrial control network and auditing services of the industrial control network are comprehensively recorded in detail, auditing data are safely reserved, and an information transmission function between the safety auditing terminal and the artificial intelligence learning terminal is achieved; and by arranging the artificial intelligence learning terminal, new security risks are found in time through the industrial control network learning module and the flow behavior learning module of deep learning analysis, and data support is provided for investigation and evidence collection of network security accidents.

The invention provides a text classification data processing method and device, a storage medium and a program product, and relates to the field of data processing, in particular to the fields of natural language processing, big data, intelligent search, deep learning and the like. According to the specific implementation scheme, a plurality of initial label mining functions are constructed in advance, target parameters of the initial label mining functions are tested on a test set, and the label mining functions with the target parameters meeting training requirements of a text classification model are obtained and serve as final target mining functions; the target label mining function is used for generating the classification label of the initial text, a large amount of training data of the classification label with labels in a specific application scene can be obtained, a large amount of training data with accurate classification labels are provided for training of a text classification model, the cost of manual data labeling is greatly saved, and the efficiency of obtaining the training data is improved.

The invention discloses a model training and speech synthesis method and device, equipment and a medium, and relates to the technical field of artificial intelligence, deep learning and speech. The specific implementation scheme is that the model training method comprises the steps: obtaining a sample text in a training data set; determining label information corresponding to the sample text basedon an acoustic model trained by adopting an unsupervised training method in advance, wherein the label information comprises style information and/or role information; training a text classificationmodel based on the sample text and label information corresponding to the sample text, wherein the text classification model is used for outputting corresponding label information according to an input text. According to the embodiment of the invention, the technical effect of automatically determining the label information corresponding to the sample text is realized, the label labeling accuracyand efficiency are improved, and the training speed of the text training model is correspondingly improved.

The invention discloses a face detection and recognition method. The method comprises the following steps: S1, preprocessing an image marked with a face frame to generate a training sample; S2, constructing a face detection and recognition network, wherein the face detection and recognition network adopts a deep learning network, and high-level features and low-level features of the network are fused; S3, inputting a training sample into the constructed face detection and recognition network for training until a training loss value is smaller than a preset threshold value, and completing training to obtain a deep learning network capable of outputting face detection and face recognition results. The face detection and recognition network is designed, face detection is regarded as a face center point problem, face center point detection and face feature vector extraction are combined for learning, a face feature vector corresponding to a face frame can be obtained while the face frame is obtained, the face feature vectors are compared to obtain a face recognition result, and a face detection result and a face recognition result are output through a network.

The invention discloses a data recommendation method and device based on deep learning, and a terminal device. The method comprises the steps of obtaining character portraits of a plurality of applicants; wherein the character portraits of the applicants and the character portraits of the applicants are in one-to-one correspondence; wherein the character portrait of the applicant comprises multi-dimensional information of the applicant; performing feature extraction on the obtained character portraits of the plurality of applicants; inputting the extracted first features into a trained training model to obtain scores of a plurality of applicants respectively; and the score of each applicant in the plurality of applicants being in one-to-one correspondence with each applicant, and determining a target applicant whose score is greater than or equal to a preset threshold value from the plurality of applicants. By the adoption of the method and device, preliminary screening of a large number of applicants can be rapidly and accurately completed, the working efficiency is improved, and the labor cost is saved.