19 results about "Neural network analysis" patented technology

The invention relates to the field of seismology, signal processing and artificial intelligence, namely to automated methods and systems for early detection of earthquakes based on the analysis of seismic data, and can be used to recognize and classify longitudinal waves (P-waves) preceding the main seismic shocks, using deep learning and computer vision methods. The aim of the present invention is to create an automated early earthquake detection system using the classification of seismic signal spectrograms generated by the complex Morlet wave CWT using the YOLO deep neural network architecture. The technical result is an increase in the accuracy and speed of early earthquake detection by analyzing the time-frequency characteristics of seismic signals and automatically localizing P-wave signatures using a neural network model. The device includes a seismic sensor, an analog-to-digital converter, and a microprocessor implementing time-frequency analysis and neural network detection algorithms. The seismic signal is recorded in real time, digitized, segmented into time intervals, and subjected to preliminary digital processing, including noise filtering and amplitude normalization. Each time interval is converted into a time-frequency representation using a continuous wavelet transform, generating a two-dimensional distribution of signal energy over time and frequency. Based on the obtained data, a spectrogram is generated and fed to the YOLO neural network detection model, which is capable of automatically detecting and localizing longitudinal P-wave signatures. Based on the neural network analysis, a determination is made regarding the presence of a P-wave and its arrival time is determined. If a predetermined threshold is exceeded, an early warning signal is generated. The system provides for data accumulation and the possibility of subsequent retraining of the neural network model.

The application discloses a virtual anchor intelligent idle chat system based on a dynamic knowledge graph, relates to the technical field of artificial intelligence and virtual anchors, and comprises data preprocessing, cross-mode fusion, intention distribution, response generation and knowledge updating modules. Current and historical dialogue data of a user is acquired first; dialogue text features and entity relationship features of a knowledge graph are extracted and fused; the correlation degree of user intention and knowledge entities is analyzed by using a graph neural network; idle chat responses fused with knowledge are generated accordingly; and finally, knowledge graph updating strategies are generated according to response feedback to drive the dynamic evolution of the knowledge base. The system realizes the deep dynamic fusion of idle chat dialogue and structured knowledge, and can autonomously optimize knowledge according to interactive feedback, thereby improving the knowledge, accuracy and long-term adaptability of responses of the virtual anchor.

Approaches are disclosed for enhancing frame rate and visual smoothness in real-time video streams through multi-frame interpolation. A classification neural network analyzes two sequential frames, outputting confidence scores that indicate the reliability of motion data for each pixel. These scores determine whether a pixel's motion is accurately described by motion vectors or should be treated as static. The classification results are reused to generate intermediate frames by warping the original frames based on the motion characteristics. Blending weights are calculated by combining warped motion vector confidence values with static values, and a second neural network refines the alignment and blending of candidate frames. This second network predicts intermediate flows and generates new blending weights, which are used to warp and blend the candidate frames, ultimately producing a final interpolated frame that enhances visual smoothness and consistency in the video stream.

The present invention relates to a computer-implemented method for training a neural network for classifying images of objects suspended in a substance, preferably crystals of that substance, and to a computer-implemented method that uses this trained neural network for analyzing objects suspended in said substance, particularly crystals suspended in said substance, especially for real-time monitoring of the progress of a crystallization process of that substance. Figure 7 to be published

This invention relates to the field of battery management system technology, and discloses a power battery system and method based on user interaction upgrades. The system includes: S1, constructing a deeply coupled architecture between an interaction management unit and a battery management unit; S2, constructing a dynamic database; S3, dynamically adjusting the liquid cooling pump power and PTC heating threshold based on driving habits and ambient temperature and humidity data; S4, constructing a two-layer architecture of local edge computing nodes and a public cloud; S5, using an LSTM neural network to analyze historical charge-discharge curves and predict cell lifespan; S6, displaying a degradation degree map through an IMU interface, supporting user scanning to replace degraded cells and automatically configuring topology parameters; and S7, establishing a three-dimensional decision matrix of user settings, environmental perception, and battery status to dynamically optimize charging strategies. This application solves the problems of traditional systems such as lag in interaction, insufficient intelligence, maintenance difficulties, and data synchronization bottlenecks, reducing maintenance costs, extending battery pack lifespan, and achieving collaborative full-cycle management.

The present invention relates to an apparatus and method for assessing work risk that can provide work risk assessment information and safety measure information using process work data, worker history data, and environmental biological data. To this end, the work risk assessment apparatus according to the present invention comprises: a first preprocessing unit that removes stop words and unnecessary symbols and performs padding or truncation on process work data input as text; a language model that outputs a vector in which the preprocessed text is converted into an integer through an encoder based on the meaning and context of the preprocessed text; a second preprocessing unit that normalizes process work data and worker history data input as numerical values; a third preprocessing unit that preprocesses environmental biological data representing the work environment and the biological state of a worker; a first model that outputs risk assessment information by analyzing the vector output from the language model and the values ​​output from the second and third preprocessing units through a first neural network; and a second model that outputs safety measure information by analyzing the values ​​output from the language model through a second artificial neural network.

This invention discloses an intelligent prediction and intervention system for enterprise safety risks based on graph neural networks, belonging to the field of enterprise safety governance technology. The intelligent safety management system includes a risk data module and an intelligent analysis module, with the risk data module connected to a detection and early warning unit. By constructing a risk correlation graph, a graph neural network analysis unit, a causal inference unit, and a risk prediction unit, this invention enables correlation analysis and dynamic prediction of multi-source enterprise risks, facilitating real-time monitoring of the enterprise's safety situation and improving the initiative and accuracy of risk management. This invention constructs a visualized risk graph from entities such as equipment, chemicals, work types, and hazard records, and uses the graph structure for message transmission and causal reasoning. Even in complex nonlinear risk scenarios, it can intuitively display the risk propagation chain and generate targeted intervention suggestions, thereby reducing blind spots and decision-making delays in safety management.

This disclosure relates to multi-frame interpolation for real-time video processing using deep neural networks. A method is disclosed to improve frame rate and visual smoothness in real-time video streams through multi-frame interpolation. A classification neural network analyzes two consecutive frames and outputs confidence scores indicating the reliability of motion data for each pixel. These scores determine whether the motion of a pixel is accurately described by a motion vector or should be considered static. The classification results are reused to generate intermediate frames by warping the original frames based on motion features. Blending weights are calculated by combining the warped motion vector confidence values ​​with static values, and a second neural network refines the alignment and blending of candidate frames. This second network predicts the intermediate stream and generates new blending weights, which are used to warp and blend candidate frames, ultimately producing final interpolated frames that enhance the visual smoothness and consistency of the video stream.

A method for analyzing a cross-section of a joint includes: capturing, using a camera, an image of the cross-section of the joint; analyzing, using one or more neural networks, the image of the cross-section of the joint to classify two or more body segments that comprise the joint; isolating, using the one or more neural networks, the body segments of the joint; reassembling the body segments to form an assembled image of the joint; identifying key points in the assembled image of the joint; and measuring, using the key points, a value for a characteristic of the joint. A system for automated sectional analysis of a joint is also provided.

The present invention relates to a wearable device-based early prediction system for collecting movement data of a user in real time by using an inertial measurement unit and analyzing the movement data through a neural network-based analysis module, thereby detecting an early sign of a degenerative brain disease. When the present invention is used, excellent gait data suitable for prevention of degenerative brain disease can be collected using a low-cost inertial measurement unit. Alternatively, the present invention relates to a system and a method for collecting movement path and gait data from a user to which both a GNSS-based positioning device and an IMU-based inertial measurement unit are attached, and analyzing the movement path and the gait data in an integrated manner through neural network analysis to determine whether wandering occurs in real time.

An apparatus for performing image analysis to identify human actions represented in an image, comprising: a joint-determination module configured to analyse an image depicting one or more people using a first computational neural network to determine a set of joint candidates for the one or more people depicted in the image; a pose estimation module configured to derive pose estimates from the set of joint candidates that estimate a body configuration for the one or more people depicted in the image; and an action-identification module configured to analyse a region of interest within the image identified from the derived pose estimates using a second computational neural network to identify an action performed by a person depicted in the image.

This invention provides an AI-based method and system for airborne fault prediction and remote support, relating to the field of artificial intelligence technology. It includes extracting envelope transition features from time-series state data of airborne equipment and performing dual-envelope confidence analysis using an airborne deep neural network to achieve transition fault prediction. When communication is limited, key data is selected based on the marginal information gain value of data segments and sent to the ground station. The ground station combines the received features and data, using physical equation deduction and deep neural network analysis to generate rapid and complete prediction results, which are then cross-validated. Finally, a support plan is generated when the verification is valid. This invention achieves efficient and accurate fault prediction and remote decision support under limited communication bandwidth.

A computer-implemented method (CIM), according to one embodiment, includes training a first neural network on email behavioral characteristics drawn from a threat intelligence data lake, and training a second neural network on email technical characteristics drawn from the threat intelligence data lake. The method further includes inputting data from an Endpoint Detection and Response (EDR) system, network traffic, and email to the neural networks for causing the neural networks to analyze the data. An alert related to Business Email Compromise (BEC) is generated based on output(s) from the neural networks. A computer program product (CPP), according to another embodiment, includes a set of one or more computer-readable storage media, and program instructions, collectively stored in the set of one or more storage media, for causing a processor set to perform the foregoing method.

This invention discloses a data processing method and system for an electrothermal acupuncture therapy device, relating to the field of data processing technology. The method collects in real-time electrical parameters of the electrothermal acupuncture therapy device, thermal parameters of the needle and body surface, and physiological and subjective feedback signals from the patient. It calculates composite features and inputs them into a pre-trained operational status recognition model. Through interpretable rules and deep neural networks, it analyzes discriminable local subsequences within the composite features to accurately identify the operational status. Based on the recognition results, it invokes associated candidate parameter adjustment strategies, inputting them into an impedance dynamic model, a heat conduction model, and a patient response model to simulate the treatment process. It outputs multi-index prediction results and quantitatively evaluates the multi-index prediction results of each candidate strategy, selecting the strategy with the highest comprehensive score to adjust the relevant parameters of the therapy device. This solves the problem of existing methods being unable to accurately identify key operational statuses from complex multi-source data, achieving dynamic, safe, and personalized adaptation of therapy device parameters.

The application provides a license plate intelligent recognition method and system based on image enhancement, and belongs to the technical field of image processing. The method performs Fourier transform on license plate image data to determine the spectral distribution characteristics; then, a gray level histogram is constructed based on the characteristics, and when the gray level variance exceeds the preset variance threshold, it is determined that the license plate is affected by light changes and the dynamic characteristic index of the gray level distribution is determined; then, the index is analyzed using a convolutional neural network to generate modulation parameters of the contrast enhancement factor and the sharpening intensity; when the deviation between the spectral peak position and the preset peak value exceeds the preset deviation threshold, it is determined that the license plate is affected by motion blur, and an enhanced spectral representation is constructed; finally, the enhanced license plate image data is generated through inverse Fourier transform. The quality of the license plate image in a complex environment is significantly improved, the characters are clear and identifiable, and the license plate recognition accuracy and system robustness are effectively improved.

The application discloses a real-time positioning method of a tumor target area and a gold mark implant, and belongs to the technical field of tumor positioning, and specifically comprises the following steps: implanting a gold mark in a tumor area, constructing a gold mark-tumor dynamic model through CT and acquiring a respiratory phase feature vector; a two-dimensional projection is matched with a three-dimensional model in space by using an improved hybrid registration algorithm, and a tumor composite displacement vector is acquired; a respiratory-displacement correlation model is established through time sequence alignment, a tumor motion atlas containing a real-time position and a predicted trajectory is generated, and a reference coordinate system is updated; a rigid safety boundary is established based on the gold mark, an elastic treatment boundary is generated by combining convolutional neural network analysis of soft tissue texture and the predicted trajectory, and a non-marked area dose weight is dynamically adjusted; and finally, sub-millimeter dynamic tracking is executed through a multi-leaf collimator; the problems of tumor motion uncertainty and individual deformation difference are solved, and the target area coverage precision and dose distribution rationality are improved.

A system for complexity assessments for agent performance analysis using a memory neural network according to an embodiment includes receiving a plurality of call recordings of calls between a contact center agent and one or more users, generating, for each call recording, a respective conversation summary of the call recording, analyzing each respective conversation summary using the memory neural network to determine a respective agent performance category for performance of the agent during the respective call, each respective agent performance category being selected from a plurality of predefined agent performance categories and associated with a respective call score, comparing each respective call score to a predefined threshold value and updating a topics array based on the respective conversation summary in response to determining that the respective call score is below the predefined threshold value, and determining a total performance score for the agent based on each respective call score.