25 results about "Frequency map" patented technology

This invention relates to the field of target recognition technology, and in particular to a radar target recognition method and system based on micro-Doppler sensing attention. First, a micro-Doppler time-frequency map is generated based on the radar echo signal. Second, the micro-Doppler time-frequency map is input into a lightweight backbone network for feature extraction to obtain an initial spatiotemporal feature map. Third, a dual attention mechanism is used for the first residual compression feature extraction to obtain a first-stage feature map. Then, the first-stage feature map is input into a max-pooling layer for feature downsampling to obtain a downsampled feature map. Next, a dual attention mechanism is used for the second residual compression feature extraction to obtain a second-stage feature map. Finally, the second-stage feature map is input into a classification output module for classification calculation to obtain the radar target recognition result. This invention achieves high-precision target classification under strong background noise by introducing a time-axis integral pooling mechanism and a channel attention mechanism that conform to the physical laws of incoherent accumulation of radar signals.

This invention proposes a soft measurement method for the internal state of a ball mill based on DEM simulation and acoustic signals. The method includes: constructing a DEM simulation model based on the physical parameters of a laboratory ball mill and performing iterative calibration; under different operating parameters, acquiring acoustic signals from the laboratory ball mill experiment and obtaining corresponding internal state data from the calibrated simulation model to collaboratively construct a soft measurement training dataset; generating paired noise reduction module training data through continuous wavelet transform, time-domain superposition, and pairing processing; and training the noise reduction module and the soft measurement module based on this data to achieve noise reduction processing of noisy time-frequency maps and accurate prediction of the internal state variables of the ball mill. This invention uses internal state variables such as filling rate, collision energy distribution, and material particle size distribution extracted from DEM simulation as training labels, fundamentally solving the problem of limited soft measurement accuracy caused by insufficient label accuracy in existing methods.

This invention discloses a method for identifying unipolar and bipolar depression based on deep learning-based multi-state EEG fusion, comprising: Step 1, performing continuous wavelet transform on EEG signals in open and closed states respectively to obtain open-eye time-frequency maps and closed-eye time-frequency maps; Step 2, using a deep learning model to extract features from the open-eye and closed-eye time-frequency maps obtained in Step 1 to obtain feature vectors for open-eye and closed-eye states; then, fusing the feature vectors in open-eye and closed-eye states to obtain a multi-state fused feature vector; finally, using a deep learning classification network to classify and identify the multi-state fused feature vector to obtain the identification result, thus completing the identification. This invention uses a deep learning model based on EEG signals to perform three-class classification identification of unipolar depression, bipolar disorder, and healthy individuals, improving classification performance.

This invention relates to a method and system for identifying individual open-set radiation sources based on space-frequency domain fusion and Hybrid-EVT, belonging to the field of radiation source identification technology. It addresses the problems of frequent occurrences of unknown emission sources and the failure of the closed-set hypothesis in radiation source identification under complex electromagnetic environments. The method includes: acquiring raw IQ signals, preprocessing them to obtain a time-frequency map; extracting features from the time-frequency map based on the SFFNet model to obtain a discriminant activation vector; training the SFFNet model and a classification head based on samples of known radiation source categories, and calculating the mean activation vector for each category; constructing a Hybrid-EVT hybrid extremum model based on the training set activation vectors; inputting test samples into the SFFNet model with frozen parameters to obtain the activation vector of the test samples; evaluating the tail probability of the distance between the test samples and candidate categories, introducing unknown channel probabilities; if the unknown confidence level meets preset conditions, it is determined to be an unknown radiation source; otherwise, the corresponding known category is output. This invention is applicable to scenarios involving enhanced security for wireless device authentication.

An example method includes storing reconstructed electrophysiological data sets for at least two baseline time intervals and at least one modulation time interval. The electrophysiological data sets are representative of electrophysiological signals reconstructed at locations across a respective surface of interest within a patient's body based on non-invasively measured electrophysiological signals for respective time intervals. The modulation time interval can overlap with modulating an arrhythmia or arrhythmogenic activity. The method can also include computing dominant frequency maps from the electrophysiological signals reconstructed for at least two baseline data sets and the at least one modulation data set. The method can also include analyzing the dominant frequency maps to determine an indication of dominant frequency stability across at least a portion of the surface of interest.

This invention provides a method, apparatus, and device for suppressing SAR composite interference. The method includes: acquiring SAR echo data containing interference; performing short-time Fourier transform processing on the interfering SAR echo data to obtain echo time-frequency map data; inputting the echo time-frequency map data into a pre-trained anti-interference model for interference suppression processing to obtain interference-suppressed echo data; processing the interference-suppressed echo data using inverse short-time Fourier transform to obtain a de-interference reconstructed signal; and performing imaging processing on the de-interference reconstructed signal to obtain an interference-suppressed SAR imaging result. The pre-trained anti-interference model is a U-shaped network framework based on a combination of a Swing Transformer module and a multi-scale feature guidance module; the pre-trained anti-interference model is trained using a composite loss function. This significantly reduces target signal loss and enhances the model's robustness in complex coupled interference scenarios.

This invention discloses an automatic extraction method for subway train passing data based on time-frequency features, relating to the field of rail transit monitoring. The method includes: collecting subway track vibration signals and performing wavelet transform to obtain a two-dimensional time-frequency map; constructing a wheelset recognition plugin based on a YOLO model to recognize the two-dimensional time-frequency map and obtain initial wheelset recognition results; obtaining a timestamp interval sequence, calculating the coefficient of variation of the timestamp interval sequence, filtering initial wheelset recognition results with a coefficient of variation greater than or equal to a threshold to obtain a negative sample set, and using the remaining initial wheelset recognition results as a multi-train sample set; performing feature extraction based on a random sample consensus algorithm to obtain a positive sample set; performing transfer learning on the wheelset recognition plugin to obtain an optimized wheelset recognition plugin, performing wheelset recognition, and calculating and obtaining subway train passing data. This method solves the problems of inconsistent data acquisition standards, non-standard processing procedures, low overall efficiency, and insufficient reliability in existing technologies.

This application provides a method, device, electronic device, and storage medium for detecting sleep apnea, relating to the field of medical testing technology. The method includes: acquiring multi-source physiological signals from a patient in a conscious state; filtering the multi-source physiological signals to obtain time-frequency maps corresponding to electroencephalograms (EEGs) and electrocardiomyograms (ECMs) corresponding to respiratory physiological signals for each channel; using frequency feature dimensionality reduction models and respiratory feature dimensionality reduction models to perform feature dimensionality reduction processing on the time-frequency maps and respiratory physiological signals, respectively, to obtain frequency features and respiratory features; inputting the frequency features and respiratory features into an apnea detection model for feature fusion and time-series dependency capture to obtain sleep apnea detection results. By employing the above-mentioned method, device, electronic device, and storage medium for detecting sleep apnea, the problems of high cost, low detection efficiency, and poor accessibility in sleep apnea detection are solved.

This invention discloses a method for identifying lightning overvoltage signals based on wavelet time-frequency maps and an improved CNN, relating to the field of lightning protection technology for power transmission lines. The method includes: acquiring and preprocessing lightning overvoltage waveform data to obtain three-phase time-domain signals; extracting time-frequency features from the three-phase time-domain signals based on continuous wavelet transform and generating a three-phase time-frequency map; performing deep feature mining on the three-phase time-frequency map using a parallel feature extraction network with three pre-trained ResNet branches; weighted fusion of the output features of each branch through compression and attention activation modules; and classifying and outputting the lightning overvoltage type. This invention displays the time-frequency joint distribution characteristics of the retained signal through wavelet time-frequency maps and adaptively extracts deep nonlinear features using an improved multi-channel CNN model, significantly improving the accuracy and robustness of lightning overvoltage type identification.

This invention discloses a method for semantic detection and multi-frame joint signal reconstruction of radar composite interference, belonging to the field of radar signal processing technology. For signal reconstruction, it includes performing short-time Fourier transform and power-law nonlinear transform on mixed and clean interference signals to obtain enhanced time-frequency maps and binary masks; constructing a U-shaped network guided by physical features, embedding two-dimensional constant false alarm rate (CFAR) detection in skip connections, generating a matrix of interference signal-to-noise ratio (SNR), confidence level, and neighborhood interference proportion as physical features, and outputting a semantic mask; using the CLEAN algorithm to suppress interference, and achieving signal reconstruction based on a corrected dictionary; and solving for sparse vectors in the Doppler domain through sparse Bayesian learning to obtain a high-resolution two-dimensional distribution of the target in the range and Doppler domains. This invention improves the accuracy and anti-interference robustness of high-resolution two-dimensional reconstruction of radar targets in the range and Doppler domains through U-shaped network semantic segmentation guided by physical features and sparse Bayesian learning.

This invention discloses a bearing fault diagnosis method based on deep transfer learning under varying operating conditions. The steps include: Step 1, obtaining the source domain and target domain of the bearing; Step 2, obtaining time-frequency maps of the source domain and target domain; Step 3, pre-training a CNN model using the source domain time-frequency map to obtain a pre-trained CNN model and a source domain deep feature set; Step 4, transferring the pre-trained CNN model to obtain a transfer CNN model; Step 5, training the transfer CNN model using a small number of target domain time-frequency maps, and obtaining a target domain deep feature set from the trained transfer CNN model; Step 6, processing the source domain and target domain deep feature sets using a balanced distribution domain adaptation method based on the fusion margin criterion to obtain new source domain and target domain deep feature sets; Step 7, training a machine learning classifier using the new source domain deep feature set, and using the trained machine learning classifier to achieve fault diagnosis. This invention achieves ideal fault diagnosis performance.

This application relates to the field of industrial production, and particularly to a method, apparatus, equipment, and storage medium for detecting anomalies in continuous casting. It involves acquiring a sequence of continuous casting state parameters, including multiple sets of continuous casting state parameters collected at a preset sampling frequency. A corresponding time-frequency map is determined based on the changes in each type of continuous casting state parameter in the sequence. The time-frequency maps corresponding to each type of continuous casting state parameter are used as a color channel to construct a multi-scale fused time-frequency map. This multi-scale fused time-frequency map is input into a trained representation learning model to obtain the corresponding anomaly detection result. This application improves the accuracy of anomaly detection results by mapping the time-frequency maps of multiple key parameters to multiple color channels of a unified image, thus explicitly presenting the coupling relationship between different parameters in the image space.

This invention discloses a method, system, and medium for shortwave time-frequency map target detection based on multi-map patch fusion, belonging to the field of shortwave communication technology. The method includes the following steps: time-frequency map construction; dividing the time-frequency map and constructing a multi-channel joint detection input containing a central patch and at least one adjacent patch; performing feature extraction and context fusion processing to obtain a fused feature map; performing target detection processing on the fused feature map and outputting candidate detection results for each patch; performing cross-map weighted non-maximum suppression processing, fusing overlapping candidate boxes across patches, and outputting a weighted fused candidate box set; mapping the weighted fused candidate box set to the original time-frequency map coordinate system to obtain the final target detection result. The method in this invention effectively avoids the loss of contextual information caused by processing the time-frequency map by dividing it into multiple patches, ensuring the integrity of targets in the edge regions of each patch, and avoiding missed detection of weak signals and repeated detection across patches.

This invention discloses a radar active interference weighted joint optimization threshold localization method, comprising: obtaining a discrete domain received signal sequence based on the radar received baseband echo signal; obtaining a final time-frequency map sequentially based on a first time-frequency map; obtaining a probability distribution result based on the final time-frequency map; obtaining the inter-class variance based on the foreground probability and background probability; obtaining the optimal Otsu segmentation threshold based on the inter-class variance; obtaining the optimal Tsallis segmentation threshold based on the non-additive combined entropy; determining adaptive joint weights based on normalized inter-class variance index and normalized Tsallis gain index; fusing the optimal Otsu segmentation threshold and the optimal Tsallis segmentation threshold through a weighted method to obtain a joint segmentation threshold; and segmenting a standard grayscale image based on the joint segmentation threshold to obtain the interference localization result. This invention improves the accuracy of radar active interference localization under low interference-to-signal ratio conditions and enhances the robustness of interference localization in complex electromagnetic environments.

This invention provides a method and apparatus for optimizing a rainstorm frequency map, which can accurately identify outliers in the rainstorm frequency map. It delineates a neighborhood range centered on the outlier, constructs a continuous spatial surface function, establishes an objective functional including a fitting error term and a spatial curvature penalty term, and uses the penalized least squares method to solve for the optimal spatial surface function. This method can correct outliers while maintaining fitting accuracy and spatial continuity, smoothing errors while preserving the true design rainstorm extreme values ​​to the maximum extent. It overcomes the oversmoothing and undersmoothing problems of existing technologies, providing technical support for improving the reliability and practicality of rainstorm frequency maps in flood control, disaster reduction, and hydrological engineering design.

This invention discloses an artificial intelligence-based electromagnetic environment detection system and method, relating to the field of electromagnetic spectrum monitoring technology. The method includes the following steps: initiating SDR to perform full scan and convert the signal, extracting the topological skeleton, and generating a key sampling mask matrix; reading the mask matrix to perform non-uniform sampling, retaining key time-frequency point data, removing redundancy, forming a sparse point cloud dataset and transmitting it to the AI ​​processor; the AI ​​processor performs Thiessen polygon partitioning and interpolation completion on the sparse point cloud data, and generates a two-dimensional grayscale time-frequency map after grayscale processing; the AI ​​processor extracts features of the grayscale time-frequency map through a pre-trained model to determine the signal modulation type and protocol type.

This invention provides a method and system for autonomous iteration and real-time control of forging process based on machine vision for extracting time-frequency map features. The method transforms the thermal and deformation history data during forging into a time-frequency map through time-frequency transformation, and establishes a mapping relationship between the time-frequency map and the microstructure using a convolutional neural network (CNN). Then, a generative model is used to construct a bidirectional mapping model between process parameters and the time-frequency map. Based on this, combined with a model predictive control (MPC) framework, the process parameters are optimized in real time with microstructure targets as constraints. After the process is completed, the model parameters are iteratively updated autonomously, thereby achieving iterative optimization of the microstructure quality of the forging. This invention can effectively extract deep features of the thermal deformation history, establish a reliable mapping relationship between "process parameters - thermal deformation history - microstructure," and achieve online optimization of process parameters during forging, providing an intelligent production process optimization solution for the high-end equipment manufacturing field.

This invention discloses a radar human behavior recognition method based on the fusion of multiple attention mechanisms. It constructs an adaptive convolutional-hybrid attention module, leveraging the advantages of convolution and self-attention mechanisms in the shallow layers of the network. Through a dynamic weight learning mechanism, it adaptively balances local feature extraction and global context modeling, significantly improving the ability to discriminate complex human behaviors. A convolutional block attention mechanism is designed, adaptively recalibrating channel feature responses through a channel attention submodule, while simultaneously utilizing a spatial attention submodule to accurately locate key regions of human motion, effectively enhancing the motion feature representation in the micro-Doppler time-frequency map. This invention, through the design of a hierarchical attention enhancement architecture, fuses multi-scale motion pattern information in shallow features and strengthens the time-frequency texture features of human motion in deep features, effectively solving the problems of low recognition accuracy in existing technologies caused by the limitations of the local receptive field of convolution, loss of details in deep features, and lack of adaptive feature calibration.

This application discloses a method and system for collaboratively enhancing the signal-to-noise ratio (SNR) of UAV signals through multi-station time-frequency map fusion, relating to the field of UAV radio detection and signal processing technology. The method includes: acquiring UAV time-domain signals collected by several time-synchronized radio monitoring stations, where ≥2; performing a short-time Fourier transform on the UAV time-domain signals collected by each radio monitoring station to obtain the corresponding two-dimensional time-frequency map; using a preset reference radio monitoring station as a benchmark and aiming for optimal time-frequency map fusion quality, obtaining the optimal time delay vectors of the remaining stations relative to the reference station through global optimization; aligning and fusing the two-dimensional time-frequency maps according to the optimal time delay vectors, and outputting a fused time-frequency map. This invention can effectively improve the SNR of UAV signals in low SNR environments, enhance the reliability of UAV detection in complex electromagnetic environments, and provide reliable data support for subsequent UAV identification and positioning.

This invention discloses a method for diagnosing vibration anomalies in continuous casting rolls based on time-series analysis of process parameters, belonging to the field of intelligent operation and maintenance technology for industrial equipment. The method includes synchronously acquiring time-series data of multi-source process parameters of the drive motor, fusing them to generate a composite electrical feature sequence; performing time-frequency joint analysis to obtain a time-frequency energy distribution map and extracting electrical state patterns; adaptively segmenting the time-frequency map based on these patterns to pinpoint abnormal regions and trace back to the corresponding original signal segments; performing dual analysis of these segments using high-order statistical features and nonlinear dynamic features, fusing the results to construct a multi-dimensional feature vector describing the instantaneous health of the system; inputting this vector into a pre-trained vibration anomaly diagnosis network for state identification, and generating a diagnostic report containing anomaly type and location information. This invention achieves precise location and deep nonlinear analysis of abnormal periods, improving diagnostic accuracy and early warning capabilities.

This invention discloses a method and system for multimodal fusion recognition of pedestrian gait based on 4D imaging radar, comprising the following steps: target detection using 4D imaging radar; multi-target tracking and association, retaining pedestrian target trajectories and removing ghost target trajectories; extracting pedestrian target 4D point cloud sequences and micro-Doppler time-frequency maps; constructing a dual-modal gait recognition network model; training and updating the parameters of the dual-modal gait recognition network model; and using the trained dual-modal gait recognition network model to achieve gait identity recognition. This invention aims to eliminate the influence of ghost targets on pedestrian gait data extraction through multi-target tracking, and to fuse 4D point cloud sequences and micro-Doppler time-frequency maps related to pedestrian gait, combining spatiotemporal variation information and micro-motion information to solve the problem of low gait data representation capability of single-modal radar, thereby improving the robustness of gait recognition.

The application discloses a bridge health state detection method and system based on a double-branch mamba network. The method comprises the following steps: collecting bridge vibration field data and processing, generating a vibration field time profile and an instantaneous frequency diagram, obtaining a data set and a training set; taking the data set as input, constructing a double-path lightweight network model for bridge health state identification; using the training set to train and optimize the double-path lightweight network model; inputting a test sample into the trained double-path lightweight network model, outputting a bridge health state classification result, and evaluating the model performance. The system comprises a data acquisition module, a model construction module, a training module and a performance evaluation module. The application uses distributed sensors arranged on both sides of the bridge pavement, and uses vehicle driving as an excitation source, so that the bridge health condition detection can be realized without affecting the normal operation of traffic.

This invention provides a diagnostic method for new energy vehicle bearings based on synchronous compressed wavelet and Markov models, belonging to the field of fault diagnosis technology for new energy vehicle components. It involves collecting raw vibration data of new energy vehicle bearings on a test bench, converting the collected raw vibration signals into high-resolution time-frequency maps using synchronous compressed wavelet transform, constructing a transfer learning model integrating a MobileNetV2 network and a Markov chain, training and fine-tuning the transfer learning model, and inputting the high-resolution time-frequency map generated by the synchronous compressed wavelet transform of the vibration signal of the bearing to be tested into the trained and fine-tuned transfer learning model. The model automatically outputs the corresponding bearing fault type and fault severity level. Based on synchronous compressed wavelet transform, MobileNetV2 network, and Markov chain model, it achieves sufficient extraction of fault features and accurate cross-domain diagnosis, improving the model's generalization ability and diagnostic accuracy under complex working conditions.

This invention provides a fault diagnosis method for nuclear power plant circulating water pumps based on imbalanced learning, comprising: acquiring signals from fault-prone components of the circulating water pump to obtain raw vibration signals; performing data augmentation using a sliding window method to obtain augmented signal samples; extracting time-frequency domain features and generating wavelet time-frequency maps to construct an imbalanced dataset that meets the requirements of real-world operation, and dividing the imbalanced dataset into training and testing sets; building and training a neural network model using the training set until the model meets the usage requirements, then saving the parameters of the neural network model to obtain the trained neural network model; testing the fault diagnosis effect of the trained neural network model and visualizing the results. This invention significantly improves the model's generalization ability on minority class samples and effectively solves the impact of data imbalance on the model's decision boundary.

The invention discloses a transformer acoustic small sample fault diagnosis method and system based on mahalanobis distance, and the method comprises the steps: simulating a plurality of fault states of a transformer, collecting voiceprint signals in corresponding states as a data set, and dividing the data set into a training set and a verification set; respectively generating corresponding Mel time-frequency diagrams for the voiceprint signals of the training set and the verification set; extracting feature vectors from the training set Mel time-frequency graph and the verification set Mel time-frequency graph by using a convolutional neural network; calculating an average feature vector and a covariance matrix of each fault type based on the feature vectors extracted from the training set; on the basis of each sample in the verification set, the mahalanobis distance between the feature vector of the sample and the distribution of each fault type is obtained, and the fault diagnosis result with the minimum mahalanobis distance is judged to be the fault diagnosis result so as to evaluate the diagnosis accuracy; and for the to-be-diagnosed transformer, acquiring voiceprint signals of the to-be-diagnosed transformer in an operation state, and outputting a fault type according to a Mahalanobis distance minimum principle to complete fault diagnosis. According to the scheme, high-precision fault diagnosis is realized.