89 results about "Parallel processing architecture" patented technology

The invention provides an electronic information data association method and system based on the Internet of Things. The method comprises the steps that a parallel processing architecture is adopted to execute rapid protocol conversion on a high-priority data packet, data recombination is directly carried out through a pre-compiled conversion rule template, a complex format verification link is skipped, and the integrity and format correctness of the converted data packet are determined; response time and error rate statistical information in the data transmission process are obtained, protocol conversion rules and scheduling strategies are optimized through a feedback mechanism, and if the success rate of certain protocol conversion is lower than a standard value, corresponding mapping rules and verification algorithms are updated; and carrying out standardized packaging on all converted data by adopting a unified data exchange standard, adding an equipment identifier, a timestamp and a data type label, establishing a cross-protocol general data format, and carrying out final distribution after judging whether a data packet meets a standard format requirement.

The invention discloses a power supply real-time compensation technology and application based on digital signal processing, and relates to the technical field of power electronics, and the power supply real-time compensation technology comprises a split-phase parallel processing architecture: a three-phase independently configured DSP core, each core integration comprises a real-time harmonic detection module, a dynamic sliding window FFT-RT algorithm is adopted, and the window length is adjustable in a set range; the bandwidth of the self-adaptive variable parameter filter can be dynamically adjusted in a set range; the FPGA coprocessor is specially used for PWM generation; an input stage of the harmonic prediction and decomposition module is synchronously acquired by a multi-source sensor; the processing layer comprises an improved VMD decomposition unit and a variational mode decomposition order; the edge prediction network is used for deploying a lightweight LSTM model, embedding a DSP core and predicting a harmonic spectrum in a future short period; and the multi-target dynamic optimization layer is used for setting a dynamic weight distributor and performing real-time adjustment according to the load sudden change rate. According to the invention, through split-phase parallel architecture, harmonic prediction and dynamic optimization, the response speed, multi-target cooperation and reliability improvement in the field of power supply real-time compensation are realized.

The invention provides an L1-L4 lumbar vertebra segment-oriented multi-feature fusion diagnosis system and method, and the system comprises an image preprocessing module which is used for receiving a lumbar vertebra CT image sequence of a patient; a centrum anatomy partition module; the multi-dimensional image feature extraction module is used for extracting four types of quantitative features from each sub-region; the clinical multi-modal data coding module is used for independently acquiring and processing three types of clinical data: a multi-modal graph attention fusion network; and the segment-level diagnosis output module outputs diagnosis results of three levels. Through a parallel processing architecture and an optimized feature extraction algorithm, the whole diagnosis process only needs 45 seconds from data input to report generation, time is saved compared with manual film reading, and the consistency of diagnosis results is remarkably improved.

The invention provides a method for automatically calibrating collected data of high-precision measurement and control deformation monitoring equipment, and belongs to the technical field of wind tunnel inner profile measurement. Multi-equipment clock synchronization is realized by establishing a unified time reference system, and measurement data of different frequencies are processed by adopting a CPU-GPU heterogeneous parallel processing architecture, so that the measurement accuracy is improved. The method comprises the following steps: performing three-level alignment on time, space and precision dimensions by using a multi-level clustering algorithm, eliminating clock differences among equipment by using a clock drift correction equation set, establishing a space coordinate system conversion matrix to unify a global coordinate system, and performing adaptive data fusion according to a precision weight by using a Kalman filtering fusion algorithm. And a data quality evaluation system is established to ensure the accuracy of a calibration result, a high-precision calibration data set is finally output, and the technical problem that the measurement precision is reduced due to space-time inconsistency of heterogeneous data of multiple devices is solved.

In order to solve the problem that quick response and distortionless recording of ultrahigh-frequency partial discharge signals are difficult to realize when accidental pulses are generated in the prior art, the invention relates to a dual-channel parallel processing architecture for ultrahigh-frequency partial discharge signals, which comprises a signal cloning circuit and a signal processing circuit, the ultrahigh-frequency partial discharge module is used for receiving ultrahigh-frequency partial discharge signals and synchronously outputting the ultrahigh-frequency partial discharge signals as first target signals and second target signals after lossless copying; the low-gain channel circuit is used for performing buffer transmission for maintaining an original bandwidth and a dynamic range on the first target signal to generate a first gain signal; the high-gain channel circuit is used for performing gain amplification on the second target signal to generate a second gain signal; the digital signal processing unit is used for performing synchronous clock control on the low-gain channel circuit and the high-gain channel circuit; and the FPGA chip is integrated in the digital signal processing unit and is used for performing fusion splicing processing on the first gain signal and the second gain signal based on a waveform fusion splicing algorithm and generating a fusion signal.

The invention provides a data error compensation method for a biological net port flowmeter, and belongs to the technical field of ocean detection. A multi-dimensional sensor is arranged at a biological net port to collect fluid parameters, a deep learning model is used for recognizing multi-phase flow characteristics, and a corresponding compensation strategy is selected; based on a correlation analysis algorithm, compensation processing frequency is dynamically adjusted to achieve self-adaptive balance of precision and efficiency, the flow state change degree is monitored through the Euclidean distance, compensation parameters are hierarchically adjusted to ensure system stability, a parallel processing architecture is adopted to execute coarse compensation and fine compensation algorithms at the same time, and results are fused according to weights. Finally, dynamic and accurate compensation of the measurement error of the biological net port flowmeter in the complex multiphase flow environment is achieved, and the technical problem that in the prior art, the biological net port flowmeter cannot achieve dynamic and accurate error compensation in the complex multiphase flow environment is solved.

The invention discloses an acceleration method, device and equipment for partial discharge detection of a switch cabinet and a storage medium, and relates to the technical field of power equipment fault detection, partial discharge signals are collected in real time through a multi-parameter sensor, multi-source signal alignment is realized by combining a hardware trigger mechanism and timestamp synchronization, noise components are separated through multi-scale wavelet transform, and the partial discharge detection of the switch cabinet is realized. The method comprises the following steps: extracting time-frequency joint features through discrete wavelet transform, reducing dimensionality by adopting a principal component analysis method, taking low-dimensional feature vectors as input limited by a Gaussian mixture model (GMM), taking feature vectors limited by the GMM as input of an SVM classifier, outputting a classification result, and performing online judgment according to the classification result. When the output is partial discharge, an alarm prompt is triggered through the interface module; and when the output is normal, real-time monitoring is continued. According to the invention, signal processing and classification are accelerated by using an FPGA parallel processing architecture, and real-time early warning of power equipment faults is realized.

The invention discloses a target detection and tracking method based on a sublimation AI platform, and the method comprises the following steps: reading an input video stream; calling a video decoding module to complete video decoding; calling a visual preprocessing module to zoom the image to a size input by a subsequent network model; an NEON instruction set is adopted to accelerate instruction conversion and a focus operator in the yolov5; converting the format of the model file, and then completing model loading, model execution and model unloading; an efficient template matching algorithm is designed to track a view field center target; various kinds of information are transmitted and received in a message queue and UDP communication mode; calling a visual preprocessing module to zoom the image to the size of an output video stream; calling a video coding module to complete video coding and plug stream output; and processing by the multi-thread parallel processing architecture. According to the invention, the omnibearing performance is obviously improved, the requirements of practical application can be met, and the problems of high calculation cost, insufficient real-time performance and high power consumption in the prior art are improved.

The invention relates to a flow table parallel matching method and system of a programmable switching chip. The method comprises the following steps: determining an input flow table based on the programmable switching chip as a fragment boundary point, and generating a plurality of sub-tables; respectively establishing storage areas of a plurality of sub-tables, configuring matching control units in the storage areas, and connecting the storage areas to obtain a sub-table parallel processing architecture; mathematical modeling is carried out based on the processing unit number constraint, the storage space constraint and the bandwidth constraint of the subtable parallel processing architecture, and a flow table parallel matching mathematical model is generated; performing particle swarm optimization on the system throughput parameter, the resource utilization rate parameter and the matching time delay parameter to obtain a system configuration parameter group; and obtaining an input data packet, distributing the input data packet to the corresponding sub-table unit according to the system configuration parameter group, and performing parallel matching operation in the sub-table unit to obtain a data packet forwarding strategy. According to the invention, accurate matching result sorting is realized, and the accuracy of the matching result is improved.

The invention discloses a robot voice interaction signal identification and processing method and system, and relates to the technical field of electric digital processing, and the method comprises the steps: training a basic acoustic model based on high-resource voice data; freezing bottom layer parameters of the basic acoustic model, and finely adjusting top layer language specificity parameters by using low-resource voice data to obtain a multilingual voice signal recognition model; carrying out acceleration processing on the multilingual speech signal recognition model; splitting processing tasks of the voice signals, and parallelizing the processing tasks by using hardware resources of the robot; and adaptively optimizing the multilingual speech signal recognition model according to the man-machine interaction historical data. Chinese and English mixed input and accurate recognition of dialects and low-resource languages are supported, a lightweight end-to-end speech recognition model and a parallel processing architecture are adopted, and low delay and high real-time performance are achieved.

The invention relates to the technical field of satellite-borne data storage, and discloses a satellite-borne file storage system based on an FPGA (Field Programmable Gate Array), which comprises an FPGA processing unit used for executing file information management, storage control, file indexing, parallel read-write operation and error correction coding; the FPGA processing unit is used for processing file data and file information, the NandFlash storage unit is used for storing file data and file information, and the invention further discloses an FPGA-based satellite-borne file storage method which comprises the following steps: S1, initializing the file storage system, including configuring the FPGA processing unit and the NandFlash storage unit; and S2, creating the file. According to the method, the high-concurrency file operation performance is improved through an FPGA parallel processing architecture, and the efficiency bottleneck of a traditional CPU serial architecture is solved; and meanwhile, by combining triple-modular redundancy and a multi-level error correction mechanism, high-reliability storage of data in a severe environment is ensured, and the problem of data integrity caused by hardware faults and cosmic ray interference is solved.

In order to solve the key problems of difficulty in multi-source information fusion, insufficient spatial topology modeling, attenuation of long-term prediction precision and the like in an existing aircraft ground taxiing trajectory prediction method, a parallel processing architecture of a historical trajectory encoder and a pavement path encoder is designed, and trajectory time sequence features are extracted by using a long-short-term memory network; modeling a spatial topological relation of a control path by adopting a graph attention network, and realizing effective integration of heterogeneous information through a feature fusion layer; a multi-component loss function fusing the position, the speed and the acceleration is provided, and the continuity and the smoothness of a prediction track are constrained; a lightweight data enhancement strategy and an adaptive residual connection mechanism are designed, the model generalization ability is improved, and long-term prediction error accumulation is relieved. According to the method, multi-source trajectory information can be effectively fused, and the accuracy and stability of aircraft ground taxiing trajectory prediction are remarkably improved.

The invention discloses a waveform recognition implementation method based on waveform template matching of an FPGA (Field Programmable Gate Array), which adopts a flexible template configuration mechanism to support self-defined waveform recognition, does not need to depend on deep learning and large-scale data training, and meanwhile, through a configurable frequency parameter and a dynamic zero calibration mechanism, the waveform recognition can be realized. And the method is suitable for different precision, frequency range and direct current bias scenes. On the hardware implementation level, the parallelism and assembly line characteristics of the FPGA are fully exerted, an n-path parallel processing architecture is constructed, a lightweight result processing strategy of intra-group iteration updating to inter-path preferential selection is matched, storage resource occupation is remarkably reduced, and the recognition speed and reliability are guaranteed through a dual-mode judgment mechanism of rapid matching and full-quantity matching. According to the method, the defects of a traditional template matching method are overcome, meanwhile, the problems that a deep learning scheme is difficult to deploy, high in computing power requirement and high in data dependence are solved, and the practical value of the system in multiple scenes is remarkably improved.

The present invention discloses a method and system for optimizing the parallel processing architecture of a test device, and relates to the technical field of test device optimization. The method includes: collecting state parameters of the test device in real time during operation through data monitoring, preprocessing the state parameters, calculating a multi-dimensional architecture optimization index, setting an optimization objective function and constraints, constructing a multi-objective optimization model, optimizing the multi-objective optimization model according to a preset multi-objective optimization algorithm, solving the resource allocation of the processing unit, generating a resource allocation plan, applying it to a parallel processing architecture, adjusting the resource allocation in real time, and continuously monitoring the operating state of the test device through a closed-loop feedback mechanism, dynamically adjusting the optimization strategy, and ensuring that the architecture is optimized under different conditions. By calculating the multi-dimensional architecture optimization index and combining it with a dynamic adaptive optimization mechanism, it is possible to comprehensively consider the key performance indicators of the test device, adjust the resource allocation plan in real time, and flexibly respond to complex test tasks.

The invention discloses an automatic weighing and sorting device integrated on a fastener production line, and relates to the field of fastener sorting, the automatic weighing and sorting device comprises a feeding vibration disc, the feeding vibration disc is provided with a spiral feeding conveying main channel, and the conveying main channel is provided with a plurality of conveying branch channels communicated with the conveying main channel; a plurality of rollers used for assisting in pushing the fasteners to move are rotationally arranged on the conveying branch channels in a penetrating mode, soft magnetic strips used for attracting the fasteners in the conveying main channel into the conveying branch channels are fixedly arranged in the rollers, an auxiliary mechanism is arranged on the conveying branch channels, a feeding mechanism is arranged on the auxiliary mechanism, a U-shaped frame is arranged at the moving end of the feeding mechanism, and the U-shaped frame is fixedly arranged on the conveying branch channels. According to the screw sorting machine, conveying is divided into channels, a multi-channel parallel processing framework is formed, the sorting machine achieves the parallel type working process, the feeding mechanism can enable the U-shaped frame to achieve two-time discharging, the screw weighing and sorting efficiency is improved, and through cooperation of the rotating rollers and the soft magnetic strips, the probability of clamping stagnation caused by interlocking is reduced.

The application discloses a kind of legal document intelligent generation method based on large language model, comprising with: 1) by double-branch parallel processing architecture, the natural language case description of user input is sent into large language model semantic understanding passage and rule engine field extraction passage parallel processing simultaneously, two passages are executed asynchronously and do not block each other, after the result is combined in aggregation node, legal document is generated;2) multi-round dialogue guidance and element check: through multi-round natural language dialogue, user case information is collected, and the filling state of each element is recorded by maintaining dialogue state slot. When detecting the missing of mandatory elements, actively initiate the user to ask questions, and execute element integrity check after all mandatory elements are collected. 3) through the block chain storage mechanism, SHA-256 hash solidification is carried out on the key electronic evidence and is chained, and the electronic data storage certificate with legal effect is generated. At the same time, the legality of the recording evidence is detected, and the evidence list containing the storage state is automatically generated. The application effectively solves the problems of element missing, low efficiency and evidence hidden danger in the existing legal document generation scheme, and provides a low threshold and high reliability legal document intelligent auxiliary generation tool for ordinary users.

This invention discloses a child fall monitoring method based on motion capture, belonging to the field of intelligent protection technology. The method includes: acquiring video of a child's activities and performing posture estimation to form a continuous sequence of activity frames; performing dual-channel parallel processing on the continuous activity frame sequence to obtain skeletal motion features and trunk motion features; performing spectral analysis and instantaneous velocity calculation on the skeletal motion features to obtain the velocity of lower limb bones, and simultaneously performing dense optical flow calculation on the trunk motion features to generate trunk optical flow velocity; based on the child's fall risk value, performing real-time posture tracking and surrounding environment scanning of the child's activity state, and collecting motion trajectory data; and performing a hazard classification response based on the motion trajectory data to output an intelligent protection plan. This invention, through a dual-channel parallel processing architecture and a spatiotemporal behavior analysis model, achieves collaborative perception of multi-dimensional motion features and accurate capture of the imbalance process of fall precursors, improving the timeliness of fall warnings.

The invention provides a high-voltage power supply control system and method based on an FPGA, and relates to the technical field of high-voltage power supplies, and the system comprises a core control module which takes an SoC type FPGA chip integrated with a PS end and a PL end as a control core; the signal acquisition module synchronously acquires analog monitoring signals of an external high-voltage power supply and converts the analog monitoring signals into multiple paths of digital signals; the control signal output module outputs an adjusting signal; the PL end realizes synchronous processing of multiple paths of digital signals through a hardware parallel processing architecture, and directly generates synchronous driving instructions for a plurality of external high-voltage power supplies based on real-time closed-loop control logic; the PS end runs an operating system, processes the advanced control instruction and sets control parameters for the PL end; a real-time protection function is integrated at the PL end, and when the transient fault is judged to occur, the control signal output module is directly controlled to execute a protection action by bypassing the PS end, so that synchronous adjustment and transient fault protection of the multi-path high-voltage power supply are realized, and high reliability and high flexibility are realized.

This invention belongs to the field of digital control systems for power electronic devices. It provides a high-speed communication system and method for multiple units of CHB frequency converter based on FPGA ping-pong buffer. By building an efficient parallel processing architecture at the FPGA hardware level, it realizes zero-wait instruction issuance and instantaneous status acquisition between the controller and dozens of distributed power units, compressing the system communication and processing time from milliseconds to hundreds of microseconds, achieving the technical effect of CHB frequency converters achieving extremely short control cycles and high dynamic performance.

The invention discloses an electronic equipment falling protection system based on multi-dimensional sensing. The system comprises a multi-dimensional sensing array, an edge calculation unit, a self-adaptive magnetic field control system and a continuous learning unit, the multi-dimensional sensor array is integrated with a triaxial accelerometer, a gyroscope, a barometer and a magnetometer, and the edge calculation unit adopts a low-power-consumption FPGA and a customized neural network processor to execute a time sequence feature extraction and trajectory prediction algorithm. According to the invention, the problem of insufficient identification precision of a traditional single sensor method under insufficient illumination or complex scenes is overcome; a 5kHz high-frequency sampling rate is adopted, and a multi-stage assembly line parallel processing architecture is combined, so that efficient extraction and trajectory prediction of time sequence characteristics are realized, and the response speed of the system is greatly improved; according to the method, the system power consumption is effectively reduced, and meanwhile, the real-time performance and accuracy of the model are ensured; the multiple electronic devices are controlled to cooperatively generate the dynamic magnetic field network, accurate magnetic suspension track control is provided for the falling device, and the device damage risk is effectively reduced.

The invention provides a cyclic shift register, a data cyclic shift method and a storage medium, the cyclic shift register comprises a data input port, a data output port, a control port and a multi-stage processing network, and the multi-stage processing network is constructed by mixing basic units of 2 * 2 switching units or 2 * 1 multiplexers. Each stage of processing unit comprises a plurality of basic units, the plurality of basic units of each stage of processing unit are of the same type, at least two stages of processing units in the whole network adopt different types of basic units, and the plurality of basic units of each stage of processing unit are independently controlled by a common shift control word generation stage control signal. The cyclic shift operation can be quickly realized; according to the cyclic shift register, by means of a multi-stage parallel processing architecture, the data processing efficiency is greatly improved, the fan-out load and global connection pressure of a control signal are relieved while ultralow delay and high parallelism are kept, and key path delay and dynamic power consumption are remarkably reduced.

The invention discloses a high dynamic range radar signal acquisition system and method, and relates to the technical field of radars. The system comprises an intensity detection unit, a multi-channel acquisition unit, a channel selection unit and a signal processing unit, the signal processing unit adopts a unique parallel processing architecture: real-time feature extraction is performed on the synchronized data stream on a first processing path to generate a window function selection instruction, meanwhile, accurate delay is performed on the data stream on a second processing path, and the delay time is matched with the instruction generation time; therefore, the time alignment of the instruction and the data stream is realized. Then, the system determines a target window function according to the instruction, and performs windowing operation on the delayed data stream. According to the method, dual self-adaption of the channel and the window function is achieved, it is ensured that signals in the full dynamic range from strong to weak can be optimally processed, and the signal collection quality and the radar detection performance are remarkably improved.

The invention provides a small-view-field temperature-sensitive interferometer antenna real-time calibration method under a multi-node parallel processing architecture, and aims at a small-view-field temperature-sensitive interferometer antenna, and a non-cooperative radiation source signal with a known position is utilized to carry out real-time calibration on the small-view-field temperature-sensitive interferometer antenna. According to the invention, a single phase difference correction mode is adopted to compensate a phase difference error value in a full-view-field range in a certain frequency band under a certain center frequency point of the small-view-field temperature-sensitive interferometer antenna, and the improvement efficiency is extremely high; a time-sharing calibration meter, a real-time calibration meter and a historical calibration meter are provided to adapt to the characteristics that the phase difference of the small-view-field temperature-sensitive interferometer antenna has short-period rapid change at different moments in one day and long-period slow change all year round; a real-time calibration table sharing mechanism is designed, and the problem of calibration parameter sharing under a multi-node parallel processing architecture is solved; a time-sharing calibration table updating mechanism with a fixed size is designed, and the problems of lightweight design of a correction number library table and loading time consumption under a real-time processing condition are solved.

The application relates to a low-orbit satellite orbit determination method, device and medium based on a multi-thread parallel processing architecture, and belongs to the technical field of satellite orbit determination. The method comprises the following steps: adopting a multi-thread parallel architecture to modularly decouple a satellite orbit determination business process, splitting the satellite orbit determination business process into three types of independent threads, i.e., a positioning thread, an orbit integration thread and an orbit prediction thread, and configuring the three types of threads in turn from high to low in priority: the positioning thread has the highest priority, the orbit prediction thread is in the middle, and the orbit integration thread has the lowest priority; meanwhile, relying on a hierarchical priority scheduling mechanism, GNSS observation data is preferentially guaranteed to be processed in real time in the positioning thread, and the latest effective orbit determination basic observation information is continuously output; the second priority is guaranteed to enable the orbit prediction thread to quickly respond to user requests and data retrieval requirements of an inter-satellite link; and finally, the kinetic integration with a large amount of calculation is processed, so that the real-time performance and business continuity of the orbit determination are effectively improved.

The invention discloses an intelligent monitoring method for child falling based on motion capture, and relates to the technical field of intelligent protection, and the method comprises the steps: collecting a child activity video, carrying out the posture estimation, and forming a continuous activity frame sequence; performing dual-channel parallel processing on the continuous activity frame sequence to obtain skeleton motion features and trunk motion features; performing spectral analysis and instantaneous speed calculation on the skeleton motion features to obtain lower limb skeleton point speeds, and synchronously performing dense optical flow calculation on the trunk motion features to generate trunk optical flow speeds; according to the child falling risk value, real-time posture tracking and surrounding environment scanning are carried out on the activity state of the child, and movement track data are collected; and carrying out emergency grading response based on the motion trail data, and outputting an intelligent protection scheme. According to the invention, through a two-channel parallel processing architecture and a time-space behavior analysis model, cooperative perception of multi-dimensional motion features and accurate capture of an unbalance process of a tumble precursor are realized, and timeliness of tumble early warning is improved.

The application discloses a kind of self-adapting focusing microscope equipment, adopt the data processor of ZYNQ heterogeneous multi-core architecture, and realize automatic focusing by step S1 to step S5, based on the parallel processing architecture of PL unit, namely FPGA, image convolution operation of multiple images can be executed in parallel in step S3, will not affect the line-by-line scanning process of microimage in step S1 is executed synchronously, so that while completing microimage scanning in step S1, step S3 has completed the most part of calculation of image definition evaluation function, when microimage scanning is completed, image definition evaluation function can calculate the definition evaluation value of microimage in very short time;Therefore, the application uses camera+ZYNQ heterogeneous multi-core architecture to replace the serial execution architecture of industrial camera+host computer in the prior art, only needs the scanning time of microimage+very short time to obtain definition evaluation value, greatly shorten the time of obtaining definition evaluation value, improve focusing speed.

The present invention provides a GPU-based GMSK parallel modulator, comprising a differential encoding module, a shaping filter module, a data segmentation module, a Gaussian filter module, a Gardener interpolation module, a carrier modulation module, and a data combination module, all electrically connected in sequence. The differential encoding module and the data combination module are disposed on a CPU, while the shaping filter module, the data segmentation module, the Gaussian filter module, the Gardener interpolation module, and the carrier modulation module are all disposed on a GPU. The present invention implements the GMSK modulator using a purely software approach. The upsampling portion after Gaussian filtering employs the Gardener algorithm for interpolation, and interpolated values are calculated in real time, achieving higher data accuracy than modulators using traditional FPGA and DSP dedicated hardware. The use of page-locked memory saves data copying time between the CPU and GPU. Multi-stream data segmentation overlay operations are utilized to dynamically adjust the data throughput of a single parallel modulation, thereby meeting the real-time requirements of engineering applications and realizing a GPU-based multi-stream data segmentation overlay parallel processing architecture.

The purpose of this disclosure is to provide a computer system based on a quantitative pain model, comprising: a preprocessing module, a local temporal feature extraction module, a cross-domain global frequency domain feature extraction module, a cross-domain interactive fusion module, a feature fusion module, and a classification output module. Through a dual-parallel processing architecture of local temporal and cross-domain global pathways, multi-scale convolution and sliding window temporal modeling are used to accurately capture transient and evolutionary temporal features related to pain. Then, through wavelet transform and attention mechanisms, frequency domain oscillation energy closely related to the physiological mechanisms of pain is explicitly extracted. Finally, a cross-domain interactive encoder is used for deep fusion, achieving complementarity and enhancement of temporal and frequency domain features. This disclosure effectively controls the computational scale while maintaining high-precision classification, making it easier to deploy lightweight and perform real-time inference on embedded devices or mobile medical terminals, demonstrating significant practical value.

In order to solve the key problems of the existing aircraft ground sliding trajectory prediction method, such as multi-source information fusion difficulty, space topology modeling deficiency and long-term prediction accuracy decay, the history trajectory encoder and the pavement path encoder parallel processing architecture are designed, the long short-term memory network is used to extract the trajectory time sequence characteristics, the graph attention network is used to model the space topology relationship of the control path, and the heterogeneous information is effectively integrated through the feature fusion layer; The multi-component loss function is proposed to fuse position, speed and acceleration, and the continuity and smoothness of the predicted trajectory are constrained; The lightweight data enhancement strategy and the adaptive residual connection mechanism are designed to improve the model generalization ability and relieve the long-term prediction error accumulation. The method can effectively fuse multi-source trajectory information, and significantly improve the accuracy and stability of the aircraft ground sliding trajectory prediction.

The application discloses an AST-aware hybrid block processing system for RAG system and belongs to the technical field of natural language processing.The application comprises the following steps: parsing a Markdown document and constructing an abstract syntax tree; performing one-time segmentation according to the title level of the abstract syntax tree node type, performing secondary segmentation when the text block length exceeds a threshold value, and setting an overlapping area; cleaning the content of the text block; parsing YAML Frontmatter metadata and traceability information in the header of the Markdown document; connecting with an external server through an end-to-end cloud pipeline; and realizing multi-process parallel and streaming processing through a parallel processing architecture.The application can realize end-to-end automation and high-performance parallel processing under the premise of ensuring block accuracy and semantic integrity.