88 results about "Filter algorithm" patented technology

The application provides a pressure signal fitting method and device, the method comprises: applying pressure to each sensing point in a pressure sensor, measuring the relationship between pressure and resistance of each sensing point, and integrating to obtain the pressure-resistance relationship; the pressure-resistance relationship is corrected through a numerical filtering algorithm, and the corrected pressure-resistance relationship is fitted to obtain an optimized pressure-resistance relationship; the switching of the voltage sensing circuit sensing points is carried out through an integrated analog electronic switch, the voltage of the stressed multiple array sensing points is measured; the resistance of the multiple array sensing points is calculated according to the measured voltage combined with the acquisition circuit; and the pressure of the multiple array sensing points is calculated through the optimized pressure-resistance relationship; based on the area of the stress area and the non-stress area of the multiple array sensing points, the actual stress of the multiple array sensing points is calculated through the relationship between force and pressure.

This invention discloses a method for analyzing the features of crystal electron diffraction patterns based on computer vision, belonging to the field of image processing technology. It solves the problems of low detection efficiency in existing technologies, lack of prior knowledge of crystallographic physics constraints, insufficient noise resistance, and lack of end-to-end quantitative analysis in deep learning methods for diffraction patterns. The method includes: constructing an electron diffraction pattern dataset and training a model; detecting local bounding boxes on sub-patterns of the electron diffraction pattern to be inferred according to a set block size and sliding step size, then superimposing pixel offsets to map back to the global coordinate system; removing redundant boxes through non-maximum suppression to generate an initial set of diffraction spot center coordinates; constructing a lattice filtering algorithm based on the random sampling consensus algorithm to eliminate noise interference and obtain a corrected coordinate set; identifying the local diffraction pattern symmetry based on the number and angular features of diffraction spots in the neighborhood and calculating local geometric parameters to generate quantitative characterization data.

The application discloses a real-time control method, a controller and a control system for a hydraulic variable-damping oil cylinder, wherein the control method comprises the following steps: acquiring in real time piston position information and external driving force information of an oil cylinder piston, calculating actual position information, actual speed information and actual acceleration information of the oil cylinder at a current time of the oil cylinder based on the position information and the external driving force information of the piston by using a filter algorithm, obtaining target position information, target speed information, target acceleration information of the oil cylinder at the current time and output force required for forming a target trajectory based on a target trajectory of the oil cylinder, obtaining a closed-loop target output force of a spool control system based on the above information, and calculating target spool position information of a throttle valve spool at the current time based on an actual flow of the throttle valve at the current time and the closed-loop target output force of the spool control system, so that the stability of target trajectory tracking can be ensured.

This invention relates to a decentralized federated learning optimization method and system based on a consensus Kalman filter, belonging to the fields of privacy computing and data security. The invention aims to optimize and improve the effectiveness of decentralized federated learning methods based on differential privacy. It comprises two core components: a gradient filtering algorithm based on a consensus Kalman filter and an overshoot filtering algorithm based on a consensus Kalman filter. The former uses a Kalman filter to filter gradient information during model training, reducing noise bias in the gradient, increasing gradient accuracy and stability, and thus accelerating the model's convergence speed and accuracy. The latter addresses the overshoot problem caused by momentum optimization methods by using a Kalman filter with exponentially decaying system variance to filter overshoot noise in the model parameters, mitigating the impact of overshoot on model convergence speed and accuracy, ensuring accelerated model convergence in the early stages and stable model convergence in the later stages, thereby achieving higher accuracy.

This invention discloses a distributed WiFi indoor personnel detection and intelligent control method and system with resistance to environmental interference, relating to the fields of wireless sensing and IoT control. The method includes: sensing nodes collecting CSI sequences, constructing frequency domain features through fast Fourier transform, using an adaptive notch filter algorithm to identify and eliminate periodic interference at specific frequencies, and extracting the non-periodic spectral energy of human breathing and micro-movements; the edge gateway employing a lightweight decision fusion strategy, performing a logical "OR" operation on the local features of multiple nodes, and utilizing spatial diversity to eliminate single-link line-of-sight blind spots and multipath dead zones; the controller constructing a dual-threshold three-state state machine based on CSI statistical features, setting a high trigger threshold to prevent false alarms, a low hold threshold to maintain static detection, and introducing an early warning mechanism. This invention effectively solves the problems of traditional wireless sensing being susceptible to interference in complex dynamic environments and having a high rate of missed static personnel detection, achieving highly robust non-contact personnel detection and energy-saving lighting control.

The application discloses a livestock data search method and system based on multi-source heterogeneous fusion. The application adopts a four-layer architecture of data acquisition, data processing, retrieval service and user interaction: through adaptive acquisition integration of active grabbing and passive access, covering public and non-public data; using rule engine and machine learning to clean data and establish meta model, standardizing and converting structured, semi-structured and unstructured data, constructing livestock field knowledge graph and using hybrid storage of relational database, non-relational database and graph database; providing personalized recommendation through user portrait construction and collaborative filtering algorithm; using RBAC permission management and data desensitization to ensure compliance. The application realizes efficient integration, accurate retrieval and intelligent service of multi-source livestock data.

A kind of distributed adaptive target tracking method with communication protection, comprising: after each observation node executes information filtering algorithm, adaptive weight coefficient is calculated according to tracking result, and information interaction is carried out using consistency algorithm with communication protection, and distributed target tracking is completed.The present application can obtain higher cooperative target tracking precision when the number of communication is less, and realizes the communication protection to target tracking result.

This invention relates to a method for optimizing synchronization accuracy in end-to-end transmission mode clock synchronization systems. First, the master and slave nodes within the clock synchronization system exchange PTP messages in end-to-end transmission mode according to the IEEE 1588 Precision Clock Protocol. The slave node obtains the timestamps and transmission delays recorded in each PTP message. Then, the slave node uses a filtering algorithm to process the forward and reverse transmission delays. Combining this with the clock offset calculation formula in the IEEE 1588 Precision Clock Protocol, the clock offset between the slave node and the master clock is calculated using the timestamps in the PTP messages and the processed transmission delays. Finally, the calculated clock offset is used to adjust the local clock. This invention solves the problem of asymmetric delays during message transmission affecting the clock synchronization accuracy between the master and slave nodes in end-to-end transmission mode clock synchronization systems, improving the time synchronization accuracy between devices in network communication systems and having broad application prospects.

This application provides a method for labeling display devices and dynamic gesture data. The method includes: in the data acquisition stage, storing a gesture configuration file containing a specific execution order and time range to provide strong prior constraints for labeling and ensure that the acquired data meets expectations; secondly, using a gesture recognition model for preliminary labeling, and configuring a driving filtering algorithm to match the model inference results with the gesture configuration file to retain high-confidence labeled data; furthermore, introducing an uncertainty measure to distinguish between high- and low-value samples, directly adopting samples with low uncertainty, and manually correcting samples with high uncertainty to ensure the labeling accuracy of difficult samples; finally, verifying the labeling through multi-viewpoints by comparing the consistency of gestures in the same time period under different camera views to eliminate single-viewpoint errors and further improve the labeling accuracy, solving the problem of low labeling accuracy of current gesture video data and improving the accuracy of gesture video data labeling.

This invention discloses a DC-side shunt current acquisition and calibration method, comprising the following steps: acquiring the shunt sampling voltage; converting the ADC voltage into pack current; filtering the sampled signal, including the following two steps: moving average filtering: using a 50-fold moving average filter for data preprocessing, the moving average filter constructs a fixed-length sampling window, calculates the arithmetic mean of continuously acquired DC-side current sampling values ​​within the window, and updates the filtered output value in real time; adaptive envelope filtering: introducing an adaptive envelope real-time filter, through its targeted filtering mechanism, to achieve precise control of current signal amplitude fluctuations; calculating the error between the original and filtered signals; and determining whether the result is smooth. Through the above methods, this invention introduces an adaptive envelope real-time filtering algorithm, significantly improving current detection accuracy, reducing project costs, and enhancing BMS robustness.

The application relates to the technical field of robot motion control, and discloses a high-precision control method and system for robot motion and a storage medium. The method comprises the following steps: after receiving an original instruction, performing encoder calibration processing based on the angle position information recorded by an encoder and the target position deviation, generating a basic control signal quantity of the rotation angle of a motor, converting a distance to be walked into a control signal quantity of the rotation angle of left and right wheels through calibrated wheel diameter parameters and wheelbase parameters, and outputting a driving control signal for controlling the rotation of the left and right wheels after superimposing an error compensation control quantity generated through a filtering algorithm. The application improves the motion control precision and self-adaptability of a robot in a dynamic environment.

This application provides a method, apparatus, device, medium, and product for pseudo-contour removal, relating to the field of image processing technology. The method analyzes the input image according to a preset detection window, using the absolute difference between the maximum and minimum pixel values ​​as the first analysis result and the average of the differences between the center pixel and its neighboring pixels as the second analysis result. This accurately captures the overall difference characteristics and high-frequency detail richness of image regions, effectively distinguishing between flat regions, detail regions, and edge regions. By integrating the analysis results to obtain a hybrid gain, and mapping the hybrid gain to filter weights adapted to different regions, it avoids the problems of incomplete pseudo-contour removal due to insufficient filter strength or blurred real details due to excessive filter strength. Based on the filter weights and a preset filtering algorithm, the input image is subjected to targeted filtering processing, balancing pseudo-contour removal effectiveness with image detail integrity, thus improving image visual quality.

The application provides a bag mouth recognition method during beverage bag filling, comprising: obtaining a bag body image of a beverage bag, and recognizing a bag body region and a suction nozzle candidate region on the bag body; obtaining three-dimensional structure information of the suction nozzle candidate region, and recognizing a real position of the suction nozzle to obtain initial positioning coordinates of the suction nozzle; extracting shoulder feature points of the bag body region, constructing a dynamic constraint model based on a spatial topological relationship between the shoulder feature points and the initial positioning coordinates of the suction nozzle, and verifying and correcting the initial positioning coordinates of the suction nozzle by using the dynamic constraint model to obtain corrected positioning coordinates of the suction nozzle; collecting corrected positioning coordinates of the same suction nozzle in a continuous predetermined number of frames of images to form a motion trajectory of the suction nozzle, and performing position prediction on the motion trajectory of the suction nozzle based on a time sequence filtering algorithm to output final positioning coordinates of the suction nozzle. The recognition method provided by the application realizes stable and accurate positioning of a flexible bag suction nozzle by fusing image analysis, three-dimensional structure recognition and shoulder topological constraints.

The application discloses a kind of based on real-time global illumination and grating hybrid acceleration ultra-high voltage station rendering method, it is related to ultra-high voltage station inspection simulation technical field, including the following steps: in strong electromagnetic interference environment, by adaptive electromagnetic interference filtering algorithm is combined with global illumination probe dynamic interpolation compensation algorithm, dynamic adjustment illumination sampling frequency and insert pre-computed global illumination data.The application is by adaptive electromagnetic interference filtering and illumination sampling dynamic compensation mechanism, effectively suppresses the high-frequency jitter of illumination data caused by strong electromagnetic noise, can according to the noise spectrum characteristics of real-time acquisition intelligent adjustment global illumination probe sampling frequency, and is combined with material perception illumination interpolation algorithm, ensure that metal and insulator surface under different interference intensity all present continuous, smooth illumination performance, so that ultra-high voltage station still can maintain stable visual output under strong electromagnetic interference environment, guarantee the reliability and authenticity of inspection simulation process.

This invention discloses a fault-tolerant control method and system for redundant robot joint deadlock faults, relating to the field of robot control technology. First, a nonlinear optimization objective function incorporating hard constraints on the faulty joint is constructed. The search space of the Jacobian matrix is ​​reconstructed in real time on a reduced-dimensional manifold defined by the joint deadlock. Redundant degrees of freedom are used to compensate for the failed joint, driving the remaining healthy joints to avoid singular configurations and confined regions by maximizing the operability index. Subsequently, a two-layer control architecture of "inverse dynamics feedforward perturbation observer compensation" is adopted, using a torque residual filtering algorithm to offset the static friction and residual torque of the deadlocked joint in real time. Finally, composite control commands are generated concurrently. This invention effectively suppresses dynamic abrupt changes caused by joint faults, significantly improving the operational accuracy and operational stability of redundant robots under extreme conditions.

This invention discloses a data communication protocol adaptation method, device, and medium for mine digital twins, relating to the field of communication protocol adaptation technology. The method includes: constructing a dynamic link reliability control model based on reliable transmission protocol variant types and access call procedure order, employing a reliability state analysis layer, a transmission quality assessment layer, and a time alignment correction layer, and standardizing and encapsulating it to generate reliable transmission control parameters; based on the reliable transmission control parameters, performing dual filtering using a behavior-aware time window filtering algorithm and calculating link reliability indicators for priority scheduling to obtain a real-time high-stability response sequence; and asynchronously pushing the real-time high-stability response sequence to the message bus to drive the digital twin model to generate the twin state of the equipment object. This invention, by employing a behavior-aware time window filtering algorithm to perform dual filtering of effectiveness and continuity, effectively improves the accuracy and continuity of the next-generation information network's response to changes in equipment operating status.

This invention discloses a denoising method for the reflection spectrum of a fiber Bragg grating (FBG) based on physical model guidance and adaptive weight optimization, belonging to the field of fiber optic sensing signal processing technology. Addressing the problem of decreased FBG sensing accuracy caused by noise interference, this invention utilizes complete empirical mode decomposition and an adaptive noise algorithm to obtain the eigenmode functions and residuals of each order of the reflection spectrum, followed by preliminary processing using the Savitzky-Golay filtering algorithm. Subsequently, the theoretical reflection spectrum of the FBG is derived based on the design parameters of the FBG using the transfer matrix method. This theoretical spectrum serves as the optimization benchmark for the weighted differential evolution algorithm, comparing the similarity between the candidate reconstructed spectrum and the theoretical spectrum under multiple indicators to ultimately obtain the globally optimal weights and reconstructed spectrum. This invention overcomes the limitations of existing data-driven mode decomposition methods caused by rigid weight allocation mechanisms. By employing a dual-driven mechanism of physics and data, it suppresses spectral peak distortion and reduces center wavelength positioning deviation, providing valuable reference for improving FBG accuracy.

This invention discloses a privacy-preserving robust federated learning aggregation method to resist backdoor attacks. Based on a dual-server architecture, clustering algorithm, and threshold filtering algorithm, it filters out malicious clients while ensuring the privacy of local updates by federated learning clients. First, this invention uses a dual-server architecture and, through arithmetic sharing and quadratic association pairs, enables the two servers to calculate the cosine distance between local updates of each client without knowing the plaintext of the client's local updates, thus achieving privacy protection for client data. Furthermore, this invention combines clustering operations and threshold filtering, allowing the servers to filter out updates from malicious clients based on both the direction and magnitude of the client's local updates, thereby improving the robustness of federated learning.

The application discloses a method for expanding the instantaneous bandwidth of optical sampling analog-digital conversion of a linear frequency modulation signal, and relates to photoelectric technology. An input linear frequency modulation signal is converted into a digital signal by an optical sampling analog-digital converter. A digital signal processing unit firstly performs interpolation twice on the output digital signal in sequence, then performs fractional Fourier transform on the interpolated signal and performs digital filtering in the fractional domain, and finally performs inverse fractional Fourier transform on the filtered data. Therefore, the input linear frequency modulation signal with a signal bandwidth not greater than twice the instantaneous bandwidth of optical sampling can be received without aliasing. The method uses an interpolation algorithm to improve the equivalent sampling rate of the system, and realizes non-aliasing reception of a full-symbol wideband linear frequency modulation signal through twice interpolation. In addition, the influence of image signals and stray signals can be eliminated by means of the fractional Fourier transform and filtering algorithm, and finally the instantaneous bandwidth of optical sampling analog-digital conversion of the linear frequency modulation signal is expanded to twice the original.

This invention discloses an adaptive compensation method for hydraulic support assembly errors based on laser tracking and positioning, relating to the field of hydraulic support assembly technology. The method includes: deploying laser radars at the four corners of the hydraulic support assembly platform to scan and collect original point cloud data of components and the platform; embedding an adaptive frequency-spatial domain joint filtering algorithm to filter out vibration and dust coupling noise, and collecting effective point cloud data. This invention constructs an accurate three-dimensional digital model of the hydraulic support and assembly platform through multi-laser radar collaborative scanning and high-precision point cloud processing. Combined with real-time coordinate transformation and a closed-loop compensation mechanism, it effectively eliminates subjective errors and random deviations in manual assembly. During dynamic assembly, it can continuously monitor the difference between the actual pose and the virtual model and perform adaptive fine-tuning, significantly improving the positional accuracy and attitude consistency of component assembly, providing a reliable guarantee for the overall structural stability and service life of the hydraulic support.

This invention belongs to the field of target tracking technology based on the fusion of multiple phased array radar data, and specifically relates to a target tracking system and method that utilizes Doppler measurement information for sequential filtering and fusion. A multi-radar sequential fusion target tracking method (SQ-AFSF) based on adaptive information feedback is proposed. In this algorithm, the DMM-LSF filtering algorithm is used, with the estimated result of the sequentially fused position filtering as the input for the sequential filtering. The filtering results of the sequential filtering are further sequentially fused to obtain the estimated result after sequential filtering. Finally, adaptive information feedback is performed according to the corresponding scenario to obtain the final target state estimation result.

This application relates to the technical field of short-circuit detection in electrolytic cells, and in particular to a short-circuit detection method and device for electrolytic cells, which includes power-on preparation, environmental zeroing, signal acquisition, data processing and threshold comparison, and fault determination and alarm output. This application deeply integrates a moving average filtering algorithm to eliminate background stray magnetic field interference, utilizes hysteresis comparison logic and a signal integrity resampling verification mechanism to thoroughly suppress false alarms caused by operating condition fluctuations and strong electromagnetic environments, and is further supplemented by an adaptive compensation model for the operating lever length and intelligent switching between automatic and manual modes. This method significantly improves the accuracy of short-circuit feature extraction and the system's anti-interference robustness, achieving end-to-end automation and data traceability from on-site data acquisition and core algorithm calculation to fault classification and early warning. It largely replaces inefficient manual inspections and easily interfered-prone voltage detection methods, providing a highly reliable process guarantee for the stable operation of electrolytic cells.

The application provides a positioning method, device and readable storage medium based on a Bluetooth beacon, and the method comprises the following steps: generating a Bluetooth indoor ranging model in real time based on LORA and the Bluetooth beacon; performing parameter optimization on the Bluetooth indoor ranging model through a filtering algorithm to obtain an optimized ranging model; obtaining a ranging result through the optimized ranging model; and obtaining a positioning result through a multi-parameter fusion received signal strength RSSI weighted centroid positioning algorithm based on the ranging result. The LoRa and Bluetooth multi-mode beacon system provided by the application can achieve good indoor positioning effect, focuses on online generation and improvement of the Bluetooth indoor ranging model, proposes a centroid weighting improvement algorithm integrating the multi-element idea, and improves indoor positioning result testing, improves the accuracy of the generated indoor ranging model, and reduces indoor ranging errors.

The application provides a remaining life prediction method and system combining a recurrent neural network and a filtering algorithm, comprising the following steps: S1, collecting measured operation observation data and selecting characteristic data; S2, performing product performance degradation state mining and low-dimensional fusion feature extraction on the selected characteristic data; S3, constructing a state space model based on the low-dimensional fusion features, estimating model parameters, and establishing a filtering remaining life prediction model; S4, taking the low-dimensional fusion features as a training set, training a recurrent neural network, and establishing a product degradation state prediction model based on an LSTM algorithm; S5, constructing a remaining life prediction fusion algorithm combining an LSTM recurrent neural network and a filtering algorithm; and S6, realizing online prediction of the remaining life of a product equipment system by using the constructed fusion algorithm. The application improves the accuracy and effectiveness of product degradation data analysis.

The present application relates to resource recommendation technical field, especially to a kind of learning resource recommendation method and system based on user portrait, including establishing user portrait matrix and learning resource feature matrix, and using factor decomposition machine and collaborative filtering algorithm to build the implicit semantic matching model of user and learning resource;The parameter of implicit semantic matching model is learned by alternating least squares method, and user-learning resource implicit semantic association feature is generated;User-learning resource implicit semantic association feature is used as individual code, and the Pareto optimal recommendation solution set of multi-objective learning resource recommendation optimization model is searched using non-dominated sorting genetic algorithm, and learning resource recommendation scheme is generated in combination with optimal recommendation decision model.The present application dynamically adjusts recommendation strategy by combining multi-objective optimization and deep reinforcement learning etc., improves the accuracy of recommendation, while it can balance the diversity and personalization of recommended content, improve the dynamic adaptability and robustness of recommendation system.

The invention relates to the technical field of visual modeling, in particular to a BIM (Building Information Modeling) and VR (Virtual Reality)-based visual modeling reduction method and equipment for an underground powerhouse of a hydropower station. Determining a curvature incongruous characteristic value according to the difference between the polar angle of each data point in each grid unit and the median of the polar angle interval corresponding to the grid unit; and evaluating the spatial distribution regularity of the curvature anisotropic characteristic values of all the grid units in the region where each cluster is located and the energy distribution in the frequency domain, and determining a comprehensive characteristic value so as to optimize a voxel filtering algorithm, filter the three-dimensional point cloud, and perform visual modeling on the hydropower station underground powerhouse based on the filtered three-dimensional point cloud. According to the method, the problems of detail loss and insufficient simplification are solved through self-adaptive recognition of a fine structure and adjustment of filtering, and the modeling precision and efficiency of the underground powerhouse of the hydropower station are improved.

This invention belongs to the field of UAV testing technology, and relates to an airborne integrated monitoring system and method for UAVs under complex flight conditions. The system includes a signal acquisition unit, a main control processing unit, and a data transmission unit. The signal acquisition unit has at least one normalized physical interface for connecting parameter acquisition modules performing different parameter acquisition tasks, and matches corresponding preprocessing strategies based on the characterization signals of the connected parameter acquisition modules. The main control processing unit receives the electrical signals output by the signal acquisition unit, processes the data using corresponding filtering algorithms according to the parameter type, and generates structured data. The data transmission unit encapsulates the structured data into data packets conforming to the airborne wireless protocol and sends them. This significantly improves the flexible scalability of UAV airborne monitoring, the signal measurement accuracy under complex flight conditions, and the reliability of wireless data transmission.

The application belongs to the technical field of deep sea navigation and positioning, and provides an inertial navigation close-coupled correction method and system based on time-sharing cooperative underwater acoustic beacons, which is characterized in that, during the navigation process of an AUV, a pure inertial navigation mode is adopted before the arrival of acoustic auxiliary information, inertial navigation is solved based on collected IMU data and depth sensor data, and the navigation state of the AUV containing accumulated errors is obtained; after the arrival of the acoustic auxiliary information, a combined navigation mode is switched to for solving, acoustic information is broadcasted by time-sharing cooperation of beacons at different positions, the horizontal distance between the AUV and each beacon is solved, and a vertical coordinate constraint is applied; the measured distance and the navigation state of the AUV containing accumulated errors are fused by a close-coupled filtering algorithm, the accumulated errors in the navigation state of the AUV are corrected, and the corrected positioning result of the AUV is obtained. Long-time navigation and high-precision underwater positioning of the AUV are realized.

The application discloses a point cloud ground filtering algorithm combining multi-scale PTD and TPS interpolation, and the method is realized by the following steps: rough error points are preliminarily removed by using statistical outlier filtering; a multi-scale grid structure is constructed according to a proportional attenuation coefficient, the screening principle of original seed points is adjusted, seed points are selected layer by layer in a progressive manner in combination with a TIN filtering method, a self-adaptive boundary expansion strategy is adopted to inhibit non-ground point interference and edge information loss; a TIN is constructed in combination with seed points and extended points, and TIN filtering is performed on potential seed points in a smaller grid size of the next layer; non-ground points after TIN filtering are marked as unclassified points, a local surface is fitted by using TPS interpolation, unclassified points are classified by using the residual error between points and the constructed surface, ground points are recovered, and the real ground is iteratively approximated. The method solves the problems of excessive smoothing and terrain feature loss of traditional PTD algorithms in the edge of discontinuous terrain, steep slopes and mountain top regions.

The application discloses a Parkinson's disease freezing gait early prediction method based on electroencephalogram high-order phase characteristics. The method comprises the following steps: acquiring scalp electroencephalogram signals of multiple target brain regions of a user and performing pretreatment; applying a surface Laplace space filtering algorithm to the pretreated signals to suppress volume conduction effect; extracting multi-dimensional neuroelectrophysiological characteristics from the filtered signals in real time, wherein the characteristics at least include phase locking values of target brain region networks in frequency bands and phase-amplitude coupling strength indexes in target brain regions; inputting the multi-dimensional characteristics into a pre-trained machine learning prediction model to output a probability quantitative value of physical freezing gait of the user in future seconds. By introducing surface Laplace filtering and high-order cross-frequency coupling characteristics, the application effectively improves signal space resolution and specificity of prediction characteristics, realizes low-delay and high-accuracy early prediction, and can be used for wearable electroencephalogram devices.