1982 results about "Recurrent neural network" patented technology

The invention provides a home abnormal state signal detection method and system based on multi-modal sensing, and relates to the technical field of detection, and the method comprises the steps: collecting human body motion, acoustics and environment parameters through a millimeter wave radar, an acoustic sensor and an environment sensor, inputting the parameters into a deep fusion network, a spatio-temporal attention mechanism and a multi-scale convolutional neural network are used to extract a spatio-temporal feature sequence, and joint probability features are formed in combination with an acoustic feature spectrum analysis result. And inputting the joint probability features into a time sequence knowledge graph, and obtaining scene adaptive features after processing by a graph attention network and a double-flow auto-encoder network. And finally, the multi-task learning network is combined with the risk propagation neural network and the recursive neural network to carry out anomaly prediction and risk level evaluation, and is combined with historical early warning information to output an early warning strategy. According to the invention, multi-modal data can be effectively fused, the accuracy and reliability of home abnormal state detection are improved, and a more accurate risk assessment and early warning strategy is provided.

The invention provides a digital twinborn-based park pipe network monitoring and early warning method and system, and relates to the technical field of pipe network monitoring and early warning, and the method comprises the steps: constructing a park pipe network domain knowledge graph, collecting real-time monitoring data and historical operation and maintenance data, and forming a knowledge enhanced pipe network data set; extracting pipe network performance attenuation characteristics by using a bidirectional gating recurrent neural network, and deeply fusing the pipe network performance attenuation characteristics with expert empirical rules in the domain knowledge graph to construct a performance attenuation analysis model; a performance degradation curve and remaining useful life are calculated based on the model, potential fault risks are predicted, and maintenance suggestions are generated. According to the invention, precise evaluation of pipe network performance degradation, early warning of fault risks and intelligent generation of maintenance suggestions are realized, and the intelligent level and reliability of operation and maintenance of the park pipe network are improved.

The invention provides a desilting robot intelligent control method and system based on deep learning, and relates to the technical field of intelligent control, and the method comprises the steps: collecting environment information through a multi-source sensor, extracting environment features through a convolutional neural network and a recurrent neural network, and building a three-dimensional dynamic model of a desilting robot operation environment; based on the model and a hierarchical reward function, training a dual neural network by using a depth deterministic strategy gradient algorithm to obtain an intelligent control strategy model; a motion track output by the model is optimized by an improved ant colony algorithm of a double-target pheromone updating strategy based on path length and energy consumption and then is input into a layered model prediction controller, and cooperative control over a dredging robot propeller and a dredging device is achieved. The working efficiency and the energy utilization rate of the dredging robot are effectively improved, and the dredging cost is reduced.

The invention provides a private domain live broadcast user behavior analysis and drainage recommendation method based on deep learning, and relates to the technical field of deep learning, and the method comprises the steps: collecting user behavior data, constructing a multi-mode deep neural network for processing, extracting user time sequence behavior features, extracting user text intention features through employing a residual connection structure and comparative learning, and carrying out the drainage recommendation of the private domain live broadcast user. User behavior link characteristics are acquired through a time sequence diagram attention network, user layering is performed through a Gaussian hybrid clustering model, a user loss probability is predicted by using a deep cross neural network, future activeness and purchasing power of a user are predicted in combination with a recurrent neural network, and commodities are matched through a diagram attention commodity knowledge network. And performing recommendation by using the improved double-tower deep neural network in combination with difficult sample mining and a variational auto-encoder, and searching an optimized push strategy to obtain an optimal push strategy.

The invention provides a pressure-bearing equipment damage mode identification and risk assessment method and system, and relates to the technical field of safety engineering, and the method comprises the steps: collecting multi-source sensor data and image data, inputting the data into a deep neural network after preprocessing and feature extraction, extracting spatial features through a convolutional layer, and extracting time sequence features through a recurrent neural network. And using the attention mechanism to fuse the features to identify an injury pattern. And then, constructing a multi-level evaluation index system, performing combined weighting by adopting an analytic hierarchy process and an entropy weight method, inputting weights into an improved Bayesian network model based on a D-S evidence theory, dynamically updating a conditional probability table by the model by utilizing a deep neural network and a fuzzy inference rule, and finally obtaining a risk evaluation result. According to the invention, the damage mode of the pressure-bearing equipment can be effectively identified, risk assessment is carried out, and assessment precision and reliability are improved.

The invention relates to a 5G network slice dynamic scheduling method and system based on multi-modal space-time perception and an event knowledge graph, and belongs to the technical field of mobile communication network resource management. According to the method, the change of a physical scene is sensed in real time by constructing a dynamically evolved event knowledge graph and designing a double-flow space-time cross network in combination with visual semantic analysis; dynamically adjusting the resource prediction model by adopting an event-scene dual-drive mechanism, dynamically adjusting parameters of the gated recurrent neural network through an elastic adjustment factor, and optimizing a multi-target resource allocation strategy based on a reinforcement learning algorithm; a two-stage resource scheduling mode is adopted, non-preemptive resource allocation of priority guarantee is implemented in an event triggering stage, and an optimization strategy of continuous adjustment is deployed in a steady-state stage. According to the method, the resource utilization efficiency and the service quality in a high-concurrency scene are remarkably improved, the method is compatible with an O-RAN standard interface, and the method is suitable for high-reliability and low-delay communication scenes such as smart cities and industrial internet.

The embodiment of the invention provides a smart store data processing method and device based on multi-dimensional data analysis, which innovatively realizes alignment and mapping of spatial-temporal feature sequences by collecting sensor, intelligent terminal and social media data. And performing multi-dimensional data fusion based on the feature weight coefficient, and constructing a commodity recommendation model by using a convolutional neural network and a recurrent neural network. The system continuously optimizes the model performance through incremental learning, and realizes personalized commodity recommendation and store operation trend analysis. And meanwhile, a marketing decision optimization model is introduced, intelligent evaluation and dynamic optimization are performed on a marketing strategy, and the model is continuously updated through feedback data. According to the method, the limitation of traditional single data analysis is broken through, and a comprehensive data processing solution is provided for precision marketing of the smart store.

The invention provides a fire risk dynamic assessment and early warning method fused with a deep learning algorithm. According to the method, fire related data is collected in real time from various sources such as a sensor, data cleaning, standardization and feature extraction are carried out, and high-quality input data is generated. And based on the convolution and recurrent neural network, constructing a fire risk prediction model, and performing model training and verification by analyzing historical fire data. The system can monitor building and environment data in real time, assess fire risks and generate corresponding early warning information. And the early warning decision module compares model output with a set threshold value, and once the risk exceeds the threshold value, the system automatically pushes early warning information, so that managers and fire departments can take measures in time. The system also provides a user interface and a feedback mechanism to help a user to monitor a risk assessment result in real time and continuously optimize system performance. The method has the advantages of accurate data analysis and rapid early warning response, and the possibility of fire occurrence is reduced.

The invention provides a multi-task monitoring and scheduling method and system, and relates to the technical field of task scheduling, and the method comprises the steps: receiving a task submission request, and constructing a task execution time prediction model based on a historical execution time sequence and a bidirectional gating recurrent neural network; monitoring a task execution process, and identifying an abnormal task by using a hierarchical Markov decision and a spectral clustering algorithm; for an abnormal task, constructing a task dependency graph based on the resource competition degree, calculating a critical path, determining an optimal calculation node, and performing re-execution; according to the method, the task execution time can be accurately predicted, abnormal tasks can be found in time, resource optimization scheduling is carried out, and the system operation efficiency is improved.

The invention provides a computer room energy consumption and computing power balance optimization method and system based on swarm intelligence, and relates to the technical field of data center management, and the method comprises the steps: collecting server node real-time operation data, constructing a multi-objective optimization function, predicting an energy consumption and computing power change curve through a recurrent neural network, and generating an initial task distribution scheme; and a global optimal allocation scheme is searched by adopting a parallel ant colony algorithm, the running state of the server is monitored, and task migration and energy consumption adjustment are executed, so that collaborative optimization of energy consumption reduction of the machine room and balanced allocation of computing power is realized, and the service quality and the resource utilization efficiency are improved.

The invention relates to the technical field of lithium battery charging and discharging management, and discloses a lithium battery charging and discharging detection and management method, device and equipment and a storage medium, and the method comprises the steps: carrying out the data collection of the charging and discharging process of a lithium battery pack, applying consistent Gaussian noise, and obtaining an electrode-level expansion data set and a battery-level expansion data set; performing feature mapping and feature mixing by minimizing the maximum average difference loss function to obtain a mapping feature vector; estimating the state of charge and the state of health of the battery through a fusion model of a recurrent neural network and a physical information hybrid neural network to obtain a state of charge estimated value and a state of health estimated value; and constructing a battery pack electric heating model of the lithium battery pack, and generating a corresponding charging and discharging control instruction sequence. Self-adaptive adjustment can be carried out according to the real-time state of the battery, potential safety hazards such as over-charging and over-discharging and out-of-control temperature are effectively avoided, and the use safety of the battery pack is remarkably improved.

An AI-driven financial planning system for real-time market adjustment, consisting of: a neural inference coprocessor configured to execute deep financial forecasting models, including recurrent neural networks and attention-based encoders, on the device, and wherein the processor dynamically updates portfolio parameters in response to market signals exhibiting volatility differences above a statistical threshold calculated using an exponentially weighted moving standard deviation; a financial data acquisition module configured to continuously receive and analyze heterogeneous data streams, including market indices, interest rates, stock and bond price fluctuations, economic indicators, regulatory updates, and financial news sentiment feeds; a behavioral analytics engine configured to create a dynamically evolving user-specific financial behavior profile based on real-time analysis of transaction history, income-expenditure cycles, psychometric test results, and temporal lifestyle patterns using supervised and unsupervised machine learning algorithms; A goal optimization module configured to transform high-level, user-defined financial goals into quantitatively tracked multi-level goals. It uses a reinforcement learning framework that predicts optimal asset allocations across multiple time horizons. a real-time strategy simulation engine configured to perform Monte Carlo simulations and deep Q-learning-based assessments to simulate the resilience of proposed financial strategies under different macroeconomic regimes and trigger redistribution events based on predefined confidence thresholds; a compliance-aware execution interface configured to interact with financial institutions through encrypted API channels, ensuring policy enforcement using a smart contract validator and a hardware-enabled secure transaction signing unit; and a recommendation display unit configured to render dynamic dashboards for visualizing investments, reallocation warnings, confidence intervals, and sensitivity sliders, and where user interaction with the unit flows back into the behavioral model for real-time learning.

The invention discloses an electrical equipment life prediction system and method based on deep learning, and belongs to the field of electrical equipment life prediction, and the method comprises the steps: constructing a data set, and carrying out the preprocessing; performing convolution operation on the time sequence data in the data set, extracting local features to obtain a convolutional neural network, capturing a long and short term dependency relationship of the time sequence data in the data set by using LSTM to obtain a recurrent neural network, and fusing output features of the convolutional neural network and the recurrent neural network to obtain comprehensive feature representation; constructing a deep learning model based on the fusion of a convolutional neural network and a recurrent neural network, adding an attention mechanism, adopting a mean square error as a loss function, optimizing model parameters, carrying out model training by utilizing a training data set, evaluating model performance by adopting a verification set, and adjusting hyper-parameters; and inputting real-time data into the trained model, predicting the residual life of the electrical equipment, and designing an online learning mechanism. According to the invention, the accuracy and real-time performance of prediction are improved.

The invention provides a chip production logistics optimization method and system based on a multi-target genetic algorithm, and relates to the technical field of chip production scheduling, and the method comprises the steps: collecting the real-time operation data of a production line, extracting task execution features and equipment layout features through a dual-channel deep neural network, and carrying out the feature fusion through a recurrent neural network, generating a production state vector containing a multi-dimensional scheduling index predicted value; constructing a multi-dimensional optimization target set based on the production state vector, and generating an initial population by adopting a double-layer gene coding structure; executing non-dominated sorting and crowding degree distance calculation based on the fitness value, and selecting high-quality individuals to form a parent population; and performing adaptive crossover, tabu search optimization and adaptive mutation operation on the parent population, iteratively selecting a new generation of population until convergence, and outputting an optimal scheduling scheme.

The invention relates to a proton exchange membrane fuel cell (PEMFC) high current density performance prediction method, system, equipment and medium. The method comprises the following steps: establishing a multi-physical field coupling model which comprehensively considers complex processes such as electrochemical reaction, proton conduction, gas diffusion and heat transfer, and describing the change of each physical quantity by adopting a partial differential equation based on a basic physical law; performing grid division and numerical discretization on the proton exchange membrane fuel cell model; selecting model parameters, and verifying the model through experimental data of different working conditions; inputting actual working condition parameters into a model to predict performance, and analyzing a simulation result; using a convolutional neural network, a recurrent neural network and an auto-encoder to extract features from different types of data and fuse the features to form a comprehensive feature vector; a deep neural network prediction model is constructed, and a cross entropy loss function and an Adam optimizer are adopted for training; dropout, L1 and L2 regularization, k-fold cross validation and transfer learning are utilized to optimize the model, and the generalization ability is improved; the system, the equipment and the medium realize high current density performance prediction of the proton exchange membrane fuel cell (PEMFC) based on the method; the prediction precision is improved, the experiment cost is reduced, the internal mechanism can be deeply understood, and powerful support is provided for design optimization, operation management and fault diagnosis of the fuel cell.

An AI-driven system for personalized advertising in real time, where: ◯ an analytics unit to monitor and analyze user behavior in real time across multiple digital platforms such as websites, mobile applications, social media, and smart devices; the unit collects data on user engagement, browsing patterns, time spent, and content preferences to enable targeted, personalized advertising; ◯ a prediction module that predicts preferences and interests of a user, operatively connected to the user interaction data collection unit, wherein the prediction module uses machine learning models such as deep learning, recurrent neural networks (RNNs) and transformer-based architectures to predict interests of users based on historical interactions and inferred preferences; ◯ an emotion and sentiment analysis unit that assesses the user's mood in real time through computer vision, natural language processing (NLP) and voice analysis, whereby the analysis of facial expressions, voice pitch and linguistic mood is used to determine emotional states and receptivity to advertising content; ◯ an embodiment of a context awareness component in operational communication with the emotion analysis and mood unit, in some cases further augmented by various environmental and situational data such as the device type and its physical location, date and time, and the content processed in the device, processing methods, etc., in establishing adaptability and automatic ad placement to be as relevant as possible to the user and their status as prescribed; ◯ Use reinforcement learning algorithms and generative AI models to drive advertising with personalization engines. Creative elements, messaging, and presentations are dynamically adjusted in real time based on user responses to ensure advertising is personalized and always optimized for best performance; o a privacy-focused AI system with federated learning, differential privacy methods, and on-device AI processing to reduce targeted advertising while complying with international data protection laws such as the General Data Protection Regulation (GDPR) and California consumer privacy laws; o an operationally adapted ad delivery mechanism to engage with real-time bidding (RTB) networks, programmatic advertising exchanges and demand-side platforms (DSPs) and place advertisements through digital advertising networks, which guarantees the delivery of tailored advertising to the most relevant audience in real time; and o a contextual feedback loop in which the machine learning models used in the preference and interest prediction module and in advertising personalization The engine is continuously updated to reflect the latest user engagement data, improving personalization over time and optimizing advertising performance.

The invention relates to the technical field of electric drive transmission system service life prediction, and discloses an AI-based electric drive transmission system comprehensive service life prediction method, which comprises the following steps: acquiring electric drive system operation parameters and mechanical part degradation data; arranging the data into a system operation full-period input tensor and a component degradation full-period input tensor, and respectively extracting multi-order feature tensors through a bidirectional recurrent neural network and a gated time convolutional network; performing cross-modal fusion by using a cross attention fusion network in combination with physical constraints to obtain a system-component degradation fusion feature vector; and after dimension reduction, inputting a multi-layer perceptron network output life prediction value, and generating a residual service life interval with confidence in combination with historical cases. The method integrates multi-source data, realizes cross-domain feature dynamic association and physical constraint fusion, improves the precision and reliability of life prediction, is suitable for intelligent operation and maintenance of the electric drive system, and provides a scientific basis for equipment maintenance.

The invention belongs to the technical field of safety monitoring of subway tunnels, and particularly relates to a subway tunnel water leakage monitoring method and system based on multi-modal data fusion, and the method comprises the steps: obtaining multi-source sensor time sequence data in a subway tunnel, carrying out the feature extraction, and extracting feature parameters related to water leakage; carrying out fusion analysis on the characteristic parameters through a data fusion algorithm, and evaluating the position and severity of water leakage, including extracting time-space correlation characteristics of data by using a dynamic time warping method; after feature extraction is performed on the data set and the space-time correlation features by using a convolutional neural network and a recurrent neural network, feature fusion is performed; and performing anomaly detection and classification on the fused high-level features by using a neural network method to obtain the position and severity of the leakage water. The advanced data fusion technology and the feature extraction algorithm are adopted, various collected data can be analyzed in real time, abnormal conditions can be found in time, and early warning is carried out.

The invention discloses a system for improving the applicability of an intelligent camera, and relates to the technical field of intelligent camera shooting. RGB, infrared and depth sensors are used for cooperative sensing, a dynamic inter-frame interpolation algorithm is used for complementing video data, target trajectory prediction is carried out based on a target motion vector analysis algorithm in combination with a recurrent neural network LSTM + Kalman filtering, and the applicability of the intelligent camera is improved. Meanwhile, three-dimensional scene modeling and shielding recovery are introduced, and environment information complementation of a dead angle area is optimized; through a cross-camera target identity re-identification ReID algorithm, seamless tracking is carried out among multiple cameras by utilizing appearance characteristics, gait characteristics and biological characteristics of a target, and meanwhile, PTZ camera control is optimized in combination with reinforcement learning, and the angles of the cameras are dynamically adjusted; a generative adversarial network GAN based on space-time consistency is adopted, historical video data and environment modeling are combined, missing pictures in visual dead angles are intelligently complemented, and the continuity and authenticity of monitoring pictures are improved.

The invention relates to the technical field of health information, in particular to a dementia risk prediction system based on an associated information processing model, and the system firstly extracts multi-modal hidden variable features through a data processing module, and generates an individualized risk feature matrix through causal reasoning and association rule optimization. And then, the system dynamically generates a personalized intervention path by adopting a reinforcement learning optimization algorithm, and realizes long-term risk prediction through a recurrent neural network. The system outputs personalized prediction reports including dementia risk assessment, trend analysis and intervention suggestions. Experiments show that the accuracy of dementia risk prediction, the dynamic adaptability of an intervention strategy and the efficiency of health management can be remarkably improved.

The invention discloses a voice interaction system and method based on an artificial intelligence technology, and relates to the technical field of voice interaction, and the method comprises the steps: employing the double-layer cooperative recognition of a main channel Transform model and an auxiliary channel CTC-Attention mixed architecture, obtaining a text with a multi-dimensional feature label, carrying out the scene classification and dependency syntactic analysis through a bidirectional recurrent neural network, and obtaining a multi-dimensional feature label text; retrieving associated entities in real time and generating structured intention data in combination with a dynamic knowledge graph, executing three-dimensional analysis through an interpretability engine and generating a multi-dimensional traceability report, and triggering an intelligent decision fusion engine through a confidence threshold to generate structured enhanced interaction data; according to the method, scene classification and dependency syntactic analysis are carried out through the bidirectional recurrent neural network, the associated entities are retrieved in real time in combination with the dynamic knowledge graph, and the structured intention data are generated, so that the intention of the user is accurately understood, and personalized interaction services are provided.

The invention discloses a multivariable time sequence prediction method based on GRU and a computer program product. The method comprises the following steps: firstly, introducing the dynamic characteristics of self-adaptive different time sequences of self-defined time sequence decomposition, and splitting an original sequence into trend, seasonal and residual components so as to reduce the data complexity and improve the interpretability; afterwards, a value embedding module is used for unifying feature representation so as to ensure that the model fully captures the time dependency relationship; in the modeling stage, the model adopts a multi-channel recurrent neural network to independently model the three types of components so as to reduce the interference between modes and improve the learning ability. In the prediction stage, a feature splicing strategy is adopted, information of each component is integrated, and richer time sequence representation is provided. In addition, a segmented prediction strategy is designed for the model, the prediction process is divided into multiple time periods, prediction information is combined, error accumulation is reduced, and the stability and robustness of long-sequence prediction are improved. Experimental results show that the prediction precision can be improved on different data sets.

The invention provides a method for improving the image recognition precision of a new energy station, and relates to the technical field of image recognition, and the method comprises the steps: obtaining monitoring image data and environment parameter data, and carrying out the image preprocessing based on an environment impact factor and an image degradation score; extracting image features by using a deep convolutional neural network, and extracting time sequence correlation features by using a recurrent neural network; and feature optimization is realized through feature stability evaluation and dynamic calibration, and finally the working state and abnormal early warning information of the equipment are output. According to the method, the precision and robustness of new energy station image recognition can be effectively improved, and the interference of environmental factors on a recognition result is reduced.

The invention relates to the technical field of online car-hailing order dispatching, in particular to an online car-hailing intelligent order dispatching method and system based on a multi-dimensional rule, and the method comprises the steps: building a standardized multi-dimensional feature set through collecting passenger orders, driver states, traffic conditions, environmental weather and regional events, and carrying out the combined modeling of regional order demands and driver online conditions based on a recurrent neural network, thereby achieving the intelligent order dispatching of the online car-hailing. Outputting expected passenger and driver thermodynamic distribution, constructing expected difference thermodynamic distribution according to the expected passenger and driver thermodynamic distribution, training a scheduling strategy through reinforcement learning, enabling an unloaded vehicle to actively migrate to a supply and demand gap area before an order is generated, and further combining an order generation condition and driver expected off-duty information to obtain an order generation result; and calculating the accumulated driving time required by the driver to return to the expected off-duty place after the driver completes the service, and selecting the work order for distribution when the time difference between the accumulated driving time and the expected off-duty place is minimum, thereby realizing accurate matching between the order distribution strategy and the driver work cycle. The empty driving rate is effectively reduced, and the quick response capability and the resource configuration efficiency of the system in a dynamic supply and demand environment are improved.

A system for detecting service interruptions in a cable broadband network using telemetry-based anomaly detection, wherein the system comprises the following: a telemetry acquisition unit configured to acquire multi-parameter telemetry data from heterogeneous broadband infrastructure elements, including cable modems, amplifiers, optical nodes and cable modem termination systems (CMTS), wherein the telemetry data includes the signal-to-noise ratio, modulation error ratio, forward error correction counter, power levels and latency statistics; a preprocessing and harmonization module that is operationally coupled with the telemetry acquisition unit, wherein the module is configured to normalize heterogeneous telemetry streams by adjusting sampling rates, synchronizing timestamps, interpolating missing data, and filtering out false outliers; an anomaly detection unit that is communicatively connected to the preprocessing and harmonization module, wherein the unit comprises a hybrid detection framework with statistical prediction models and machine learning models, wherein the statistical prediction models include ARIMA or Holt-Winters models to predict the expected telemetry behavior and the machine learning models include recurrent neural networks and autoencoders trained on historical telemetry; an ensemble evaluation subsystem within the anomaly detection unit, configured to combine the outputs of the statistical prediction models and the machine learning models to generate anomaly probability evaluations with adaptive confidence intervals; an interruption classification module configured to receive anomaly probability values ​​and correlate anomalies across multiple devices, geographic clusters, and time windows, wherein the interruption classification module differentiates between transient anomalies and service-impairing interruptions based on a multidimensional correlation; and an alerting interface configured to transmit outage alerts with severity, root cause metadata, and geolocation to a network management system so that the operator can intervene.

The invention relates to the technical field of robot perception, in particular to an intelligent accompanying human-type robot based on multi-mode emotion interaction and a perception method, and the method comprises the following steps: collecting facial expression and body posture data of a user through a visual perception unit, collecting intonation, speed, volume and keyword data through a voice perception unit, and carrying out the recognition of the speech, speed, volume and keyword data; data preprocessing and feature extraction are carried out; secondly, carrying out time sequence modeling on the extracted visual and voice features based on a recurrent neural network, and generating visual and voice time sequence feature vectors; the method comprises the steps that firstly, time sequence features are extracted, then, a multi-mode fusion model based on an attention mechanism is adopted, the time sequence features are fused into a comprehensive emotion feature vector used for emotion state classification, and finally, personalized interaction content such as voice, expressions and limb actions matched with the emotion state of a user is generated according to an emotion classification result. According to the method, the accuracy of robot emotion perception and the naturalness of man-machine interaction are remarkably improved, and the method is suitable for multi-scene application of education, medical treatment, old-age care and the like.

The invention belongs to the technical field of circuit boards, and particularly relates to a circuit board defect detection system based on machine learning. The method comprises the following steps of: 1, acquiring multi-modal data of a circuit board through various sensors, transmitting the data to a data fusion module, fusing and outputting a unified multi-modal data set; 2, the fused data are transmitted into an adaptive deep convolutional neural network, features in the image are extracted through a deep learning model, and preliminary defect detection is carried out; 3, comparing a result after each detection with an actual defect label, calculating an error and feeding back the error, adjusting a feedback mechanism through a recurrent neural network, and gradually optimizing model parameters after each detection; step 4, after the detection is completed, classifying the defects by using an ensemble learning algorithm, and outputting a final detection result; and 5, continuously collecting new data by the system in the production process and optimizing the model according to real-time feedback.

The invention relates to the technical field of foundation pit detection, and discloses a deep foundation pit deformation prediction method and system based on neural network and rough set classification, and the method comprises the following steps: data collection and preprocessing; carrying out attribute reduction based on a rough set theory; constructing and training a model based on an improved attention mechanism recurrent neural network (A-RNN); carrying out credibility verification and dynamic feedback adjustment; and outputting a result and an early warning response. By integrating multi-parameter monitoring data of displacement, soil pressure, underground water level, support stress and the like, driving factors of deformation of the deep foundation pit are comprehensively captured, misjudgment caused by data deviation of a single sensor is reduced, the combination of bidirectional LSTM and an attention mechanism effectively captures local features of long-term dependence and key time points in time sequence data, and the accuracy of deep foundation pit deformation detection is improved. And the accuracy of deformation prediction is obviously improved.

The invention relates to the technical field of voice processing, in particular to a real-time voice synthesis and conversion method in an intelligent voice call system, which comprises the following steps of: acquiring voice data input by a user through a voice recognition module, preprocessing the voice data, performing semantic analysis through a bidirectional recurrent neural network and an attention mechanism, and obtaining a semantic analysis result; the method comprises the following steps: extracting a key semantic unit, selecting a proper voice template based on user historical interaction data and an adaptive template matching algorithm, performing voice synthesis by using a generative adversarial network, realizing cross-language or dialect real-time voice conversion through a multi-task learning algorithm, and combining phoneme-level optimization processing to ensure the naturalness and continuity of the voice, so as to improve the voice conversion efficiency. And through real-time analysis and a self-adaptive adjustment mechanism fed back by the user, the voice output quality is optimized, and the user experience is improved. According to the invention, efficient and natural speech synthesis and conversion services can be provided in various intelligent devices and application scenarios, and complex requirements in different language environments can be met.

A system for real-time analysis of emotional feedback during motivational speeches, consisting of: a series of multimodal sensors, including at least one visual sensor configured to capture facial expressions of spectators, at least one directional microphone configured to capture the audio responses of the audience, and optionally one or more physiological sensors configured to capture biometric signals from spectators; an edge-based processing unit that is communicatively coupled to the arrangement of multimodal sensors, wherein the edge-based processing unit comprises the following: (a) a feature extraction module configured to extract visual features from captured facial images, acoustic features from voice responses, and physiological features from biometric signals; (b) an emotion inference machine configured to process the features using a deep learning-based emotion recognition model comprising a convolutional neural network (CNN) for classifying facial expressions, a recurrent neural network (RNN) for classifying voice emotions, and a multimodal late fusion layer configured to compute a composite emotion state vector representing the aggregated emotions of the audience; (c) a timestamp and speech alignment module configured to correlate the calculated composite emotion state vector with segmented portions of a live motivational speech based on real-time speech-to-text transcription and semantic analysis; and (d) a session-based storage unit configured to log time-indexed emotional state vectors and corresponding speech segments for post-event analysis; A speaker feedback interface comprising a portable display or a podium-mounted visualization panel, wherein the interface is configured to display visual indicators of emotional feedback in real time, the indicators being derived from the emotional state vector and including at least emotional trend graphs, threshold alerts, or engagement indices.