Fusion monitoring method and system for multi-dimensional data

Abstract

The invention provides a multi-dimensional data fusion monitoring method and system, and relates to the technical field of data processing.The method comprises the steps that a monitoring task is uploaded to a medical information platform through a user side, a transfer device is driven to activate a multi-mode sensing array, same-frequency asynchronous non-uniform sampling and sensing side modulation are executed, and a multi-dimensional data fusion monitoring result is obtained; coding to a same-frequency carrier wave transmitted back by communication; the medical information platform receives the sensing coding data, and for the monitoring task, threshold polling processing under threshold processing link recombination migration is carried out according to a function threshold pool embedded in the platform to generate task monitoring data; and performing early warning matching based on an information early warning rule on the task monitoring data, and executing two-way early warning management of a platform visual port and a mobile terminal. The technical problems of inaccurate and untimely monitoring judgment caused by scattered monitoring data and isolated information in the prior art are solved. The technical effects that real-time and accurate monitoring and early warning are achieved, and the monitoring effect is improved are achieved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method and system for fusion monitoring of multi-dimensional data. Background Technology

[0002] In scenarios such as emergency medical transport, intra-hospital patient transfers, and long-distance medical escort, patients are in a dynamic, mobile state, and their physiological state is easily affected by multiple factors, including the transport environment, vehicle vibrations, and spatial constraints. Traditional monitoring methods typically rely on single or a few physiological parameter monitoring devices, which can only monitor a single indicator. The data is scattered and isolated, making it difficult to comprehensively reflect the patient's overall health status. Due to the lack of effective linkage and data fusion between devices, medical personnel face problems such as incomplete information and data redundancy during transport. Furthermore, due to the bumpy and complex transport environment, medical staff find it difficult to continuously monitor data from multiple devices. This traditional monitoring model cannot ensure a comprehensive and timely analysis of the patient's real-time health status, resulting in inaccurate and untimely patient monitoring and increasing safety risks during patient transport.

[0003] Existing technologies suffer from problems such as scattered monitoring data and isolated information, leading to inaccurate and untimely monitoring and judgment. Summary of the Invention

[0004] The purpose of this application is to provide a multi-dimensional data fusion monitoring method and system to solve the technical problems of scattered monitoring data and isolated information in the existing technology, which leads to inaccurate and untimely monitoring and judgment.

[0005] In view of the above problems, this application provides a method and system for multi-dimensional data fusion monitoring.

[0006] The first aspect of this application provides a method for fusion monitoring of multi-dimensional data. This method includes: uploading a monitoring task to a medical information platform from the user side; driving a transport device to activate a multimodal sensor array; performing synchronous asynchronous non-uniform sampling and sensor-side modulation; and encoding the data onto a synchronous carrier wave for communication return. The medical information platform receives the sensor-encoded data and, for the monitoring task, performs threshold polling processing under link reconfiguration and migration based on a threshold pool embedded in the platform to generate task monitoring data. The task monitoring data is then matched for early warning based on information warning rules, and dual-path early warning management is implemented via the platform's visualization port and the mobile terminal.

[0007] Optionally, the task parsing engine receives the monitoring task, performs task intent interpretation and decomposition coupling of the smallest task unit, and generates a task combination; for the task combination, it generates a dynamic threshold decision package based on data admission, threshold function and execution order; according to the dynamic threshold decision package, it performs microprocessing threshold invocation and reorganization orchestration in the functional threshold pool to determine the threshold processing link; and it migrates the threshold processing link to the data processing center.

[0008] Optionally, monitoring records are retrieved and reconstructed into M monitoring sequences representing the data source and functional logic; the M monitoring sequences are traversed, and decomposition based on the smallest functional logic unit is performed to determine N unit monitoring sequences, where N is a positive integer greater than or equal to M; by clustering the N unit monitoring sequences, X functional monitoring sequences are determined, where X is a positive integer less than or equal to M; based on the X functional monitoring sequences, microprocessor thresholds are constructed and integrated to form the functional threshold pool.

[0009] Optionally, based on the dynamic threshold decision package, a matching call is made in the functional threshold pool using the threshold function to determine the candidate microprocessor thresholds; according to data admission, the input end of each candidate microprocessor threshold is initialized; according to the execution order, the initialized candidate microprocessor thresholds are associated and reassembled to form the threshold processing link, wherein the execution order is used to associate the data interfaces between each candidate microprocessor threshold through serial or parallel relationships.

[0010] Optionally, the sensor-encoded data is transmitted to a data processing center to activate the migration threshold processing link; according to the threshold processing link, threshold polling processing under threshold admission filtering is performed on the sensor-encoded data to determine task monitoring data, wherein the task monitoring data includes threshold monitoring chain and comprehensive monitoring data.

[0011] Optionally, the threshold processing link is a directed acyclic type; after the monitoring task ends, the threshold processing link is decomposed and migrated back to the functional threshold pool.

[0012] Optionally, the medical information platform analyzes the monitoring task and allocates a time-priority-based periodic asynchronous sampling window for the multimodal sensors assembled on the transport equipment; it issues a task sampling instruction according to the periodic asynchronous sampling window, drives the multimodal sensors to perform source-end sensing acquisition, and determines the multimodal sensing signals; it modulates the original waveform of the multimodal sensing signals and encodes them onto the same-frequency carrier for communication backhaul.

[0013] Optionally, according to lightweight signal processing rules, real-time feature processing is performed on the original waveform of the multi-threaded system to determine low-dimensional feature vectors; the low-dimensional feature vectors are modulated onto a carrier of the same frequency for asynchronous backhaul through a pre-allocated orthogonal physical layer coding sequence, wherein the carrier of the same frequency is a carrier of the same frequency as the platform command but whose phase and amplitude are distinguishable, and the asynchronous backhaul method is to backhaul through the same communication channel via time division or code division.

[0014] Optionally, information early warning rules are deployed within the medical information platform, wherein the information early warning rules are jointly defined by multiple transfer impact types and multiple early warning levels; based on the task monitoring data, a matching judgment based on the information early warning rules is performed to generate targeted early warning instructions; on the visualization port of the medical information platform, the task monitoring data and targeted early warning instructions are displayed in a pop-up window, and the targeted early warning instructions are sent to the mobile terminals of the transfer personnel for early warning management.

[0015] A second aspect of this application provides a multi-dimensional data fusion monitoring system, comprising: a sensor activation module for uploading monitoring tasks from the user side to a medical information platform, driving the transport equipment to activate a multimodal sensor array, performing synchronous asynchronous non-uniform sampling and sensor-side modulation, and encoding the data into a synchronous carrier wave for communication return; a data processing module for receiving sensor-encoded data on the medical information platform, performing threshold polling processing under link reorganization migration based on the platform's embedded functional threshold pool for the monitoring task, and generating task monitoring data; and an early warning management module for performing early warning matching based on information early warning rules on the task monitoring data, and performing dual-path early warning management on the platform's visualization port and mobile terminal.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages: The method provided in this application uploads monitoring tasks to a medical information platform from the user side, drives the transport device to activate a multimodal sensor array, performs synchronous asynchronous non-uniform sampling and sensor-side modulation, and encodes the data onto the synchronous carrier of the communication return. The medical information platform receives the sensor-encoded data and, for the monitoring task, performs threshold processing and threshold polling under link reassembly migration based on the platform's embedded functional threshold pool to generate task monitoring data. The task monitoring data is then matched with early warning rules, and dual-channel early warning management is implemented via the platform's visualization port and the mobile terminal. This achieves real-time and accurate monitoring and early warning, improving the technical effect of monitoring.

[0017] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating a multi-dimensional data fusion monitoring method provided in this application.

[0020] Figure 2 This is a schematic diagram of the structure of a multi-dimensional data fusion monitoring system provided in this application.

[0021] Explanation of reference numerals in the attached diagram: Sensing activation module 11, data processing module 12, and early warning management module 13. Detailed Implementation

[0022] This application provides a multi-dimensional data fusion monitoring method and system to address the technical problems of scattered and isolated monitoring data in existing technologies, which lead to inaccurate and untimely monitoring and judgment. It achieves real-time and accurate monitoring and early warning, thereby improving the effectiveness of monitoring.

[0023] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be understood that the present invention is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. It should also be noted that, for ease of description, only the parts related to the present invention are shown in the accompanying drawings, not all of them.

[0024] Example 1, as Figure 1 As shown, this application provides a method for fusion monitoring of multi-dimensional data, which includes: The user side uploads a monitoring task to the medical information platform, driving the transfer device to activate the multi-modal sensing array, perform co-frequency asynchronous non-uniform sampling and sensing-side modulation, and encode it into the co-frequency carrier for communication backhaul.

[0025] Furthermore, driving the transfer device to activate the multi-modal sensing array, perform co-frequency asynchronous non-uniform sampling and sensing-side modulation, and encode it into the co-frequency carrier for communication backhaul includes: the medical information platform analyzes the monitoring task, assigns a periodic asynchronous sampling window based on time-priority to the multi-modal sensors equipped on the transfer device; issues a task sampling instruction according to the periodic asynchronous sampling window, driving the multi-modal sensors to perform source-end sensing acquisition to determine multi-modal sensing signals; performs original waveform modulation on the multi-modal sensing signals and encodes them into the co-frequency carrier for communication backhaul.

[0026] Specifically, the user side inputs a monitoring task through the user interface of the medical information platform, such as a mobile APP, a web page, etc. The user side is a doctor, a nurse or relevant medical staff, responsible for submitting the task content to be monitored. The monitoring tasks include, for example, continuous vital sign monitoring, key index tracking, abnormal event capture, etc. The medical information platform uses natural language processing (NLP) technology to semantically analyze the monitoring task, clarify the task requirements and objectives, and obtain the analysis result. For example, use word segmentation tools, such as Jieba, HanLP, etc. to segment the task text input by the user into individual words or phrases. For example, for the task text of continuous vital sign monitoring: perform electrocardiogram monitoring on a patient with hypertension every 10 seconds. After word segmentation, words such as hypertension, electrocardiogram monitoring, every 10 seconds, etc. are obtained, and meaningless words in the text are removed, such as common conjunctions like "of", "and". Key information is identified from the task text, such as task type, patient status, etc. Key entities in the text, such as monitoring items, types, etc. are identified through named entity recognition (NER). An intention recognition algorithm, such as a BERT-based model, is used to classify the text to identify the type of the task, such as electrocardiogram monitoring, blood oxygen monitoring. The analysis result is obtained. For example, for the monitoring of electrocardiogram, it is analyzed as: the monitoring item is continuous electrocardiogram monitoring.

[0027] Based on the analysis results, for the multimodal sensors equipped in transport equipment, time-priority-based periodic asynchronous sampling windows are allocated to the multimodal sensors according to the importance and real-time requirements of the data collected by different sensors. Transport equipment refers to mobile devices used to carry and support patients during medical transport, such as ambulances, ambulances, and transport beds. These devices are equipped with various multimodal sensors to monitor patients' vital signs and environmental changes in real time, ensuring patient safety during transport. Multimodal sensors include, but are not limited to, electrocardiogram (ECG) sensors, blood oxygen sensors, blood pressure sensors, body temperature sensors, respiratory rate sensors, and ambient temperature and humidity sensors. These multimodal sensors work collaboratively to collect various physiological data of the patient in real time. When allocating sampling windows, sensors reflecting key physiological indicators of the patient's life-threatening condition, such as heart rate and blood oxygen sensors, are given higher priority, while relatively less important environmental parameter sensors, such as temperature and humidity sensors, are given lower priority. For example, a priority scoring system is established. Based on the importance of different indicators in assessing the patient's condition and providing emergency treatment, each sensor is scored, with higher scores indicating higher priority. For instance, the heart rate sensor is crucial for assessing cardiac function and receives a score of 9; the blood oxygen sensor is critical for assessing respiratory and circulatory function and receives a score of 8; and the ambient humidity sensor has a relatively small direct impact on the patient's condition and receives a score of 2. According to the monitoring duration and real-time requirements specified in the monitoring task, corresponding sampling periods are set for sensors of different priorities. High-priority sensors use shorter sampling periods to ensure timely capture of subtle changes in the patient's physiological indicators, providing accurate information for emergency treatment. Low-priority sensors use longer sampling periods to reduce data redundancy and equipment burden. For example, for a high-priority heart rate sensor, the sampling period is set to 3 seconds, meaning data is collected every 3 seconds; for a low-priority ambient humidity sensor, the sampling period is set to 30 seconds. After determining the priority and sampling period of each sensor, an asynchronous sampling window is allocated to each sensor in chronological order. Using time-division multiplexing, the entire monitoring timeline is divided into multiple time segments, and a specific time segment is allocated as its periodic asynchronous sampling window for each sensor based on its priority and sampling period. During the allocation process, sampling windows for high-priority sensors are prioritized, and there is a certain time interval between adjacent sampling windows to avoid sampling conflicts between sensors. This allocation method can make rational use of sensor resources, ensure timely acquisition of critical data, avoid unnecessary data redundancy, and improve data acquisition efficiency.

[0028] Based on a periodic asynchronous sampling window, the medical information platform issues task sampling commands to the multimodal sensors installed in the transport equipment through a dedicated control protocol. This dedicated control protocol includes sensor interface protocols and data transmission protocols to ensure effective command transmission. Sensor interface protocols include Bluetooth and Zigbee, while data transmission protocols include MQTT. After receiving the task sampling command, the multimodal sensors perform corresponding source-end sensing acquisition according to the same-frequency asynchronous non-uniform sampling window. During the acquisition process, the sensors generate corresponding data, such as the waveform of an electrocardiogram (ECG) signal, blood oxygen saturation, body temperature, and ambient temperature, forming multimodal sensing signals. These multimodal sensing signals encompass various aspects, including the patient's vital signs and the transport environment. The acquired multimodal sensing signals are then modulated using the original waveform, and key feature information is extracted. The complex high-dimensional original waveform data is transformed into low-dimensional feature vectors. For example, for an ECG signal, key feature parameters such as heart rate and rhythm are extracted after processing, forming a low-dimensional feature vector. The processed features are then encoded into the same-frequency carrier for communication backhaul. The same-frequency carrier uses a frequency band shared by the platform and the device. Time division multiplexing or code division multiplexing is used to distinguish the data from different sensors, ensuring that different signals do not interfere with each other while sharing the same frequency.

[0029] By precisely prioritizing time allocation and issuing sampling instructions, the timely acquisition and transmission of data from various sensors during transport are ensured, providing comprehensive, timely, and reliable data support for patient health monitoring and safety assurance during medical transport.

[0030] Furthermore, the original waveform of the multimodal sensing signal is modulated and encoded onto the same-frequency carrier for communication backhaul, including: performing real-time feature processing on the original waveform of the multi-threaded signal according to lightweight signal processing rules to determine a low-dimensional feature vector; and modulating the low-dimensional feature vector onto the same-frequency carrier for asynchronous backhaul through a pre-allocated orthogonal physical layer encoding sequence, wherein the same-frequency carrier is a carrier with the same frequency as the platform command but whose phase and amplitude are distinguishable, and the asynchronous backhaul method is to backhaul through the same communication channel via time division or code division.

[0031] Specifically, lightweight signal processing rules are a collection of signal processing algorithms with low computational cost and high processing speed, including at least normalization, filtering, and dimensionality reduction. These algorithms aim to minimize system resource consumption while maintaining a certain level of processing accuracy, meeting the real-time requirements of medical monitoring scenarios. The raw waveforms from multiple threads contain data from different multimodal sensors, such as ECG waveforms acquired by an ECG sensor and blood pressure fluctuation waveforms acquired by a blood pressure sensor. Taking ECG waveform processing as an example, the min-max normalization method is used to adjust the raw waveform data from multiple threads to the same order of magnitude. Filtering algorithms are then used to remove noise interference, such as median filtering. This effectively removes impulse noise by replacing the value of each point in the ECG signal sequence with the median value of all points in its neighborhood. For example, if there is an impulse noise point with an amplitude of 5 in the original ECG signal, and its neighborhood contains 10 normal signal points with amplitudes between 0 and 2, after median filtering, the amplitude of the noise point will be replaced with the median value of the neighboring signal points, thus eliminating the noise effect. Then, dimensionality reduction is performed. Dimensionality reduction techniques, such as PCA, are used to extract key feature parameters from the filtered waveform. First, the filtered waveform data is arranged into a matrix, where each row represents a sample and each column represents a feature dimension. The matrix is ​​then standardized by subtracting the mean from each feature dimension and dividing by the standard deviation, resulting in zero mean and unit variance. The covariance matrix of the standardized matrix is ​​calculated, reflecting the correlation between the feature dimensions. Eigenvalue decomposition is performed on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors. The eigenvectors are then sorted from largest to smallest eigenvalue. Based on the preset dimensionality reduction dimension, the first k eigenvectors are selected to form a projection matrix. The standardized original data matrix is ​​multiplied by the projection matrix to obtain the dimensionality-reduced low-dimensional feature vectors. For ECG signals, features such as heart rate, heart rhythm, and QRS complex width are extracted. For blood pressure signals, features such as systolic blood pressure, diastolic blood pressure, and mean arterial pressure are extracted.

[0032] Orthogonal physical layer coding sequences (OPCGS) are coding sequences designed and allocated using mathematical orthogonality principles in wireless communication to effectively distinguish different sensors or data streams within the same frequency band. OPCGS modulates low-dimensional eigenvectors, enabling multiple signals transmitted on the same frequency carrier to be distinguishable and avoiding mutual interference. Orthogonality refers to a specific mathematical property of coding sequences, ensuring that the inner product between them is zero, i.e., they are uncorrelated, thus achieving interference-free signal separation. The number of OPCGS is determined based on the number of multimodal sensors assembled in the transport equipment and the capacity of the communication channel. Each sensor corresponds to one coding sequence, ensuring that the data from each sensor can be independently transmitted back within the same frequency band without interfering with other signals. Coding sequences are generated based on orthogonal polynomials, such as Huffman coding, orthogonal frequency division multiplexing (OFDM), and linear block coding, resulting in pre-allocated OPCGS. Commonly used coding methods include Walsh-Hadamard sequences, Gray codes, and Zadoff-Chu sequences, which possess orthogonality and ensure that different sequences do not interfere with each other on the same frequency resource. The orthogonality principle ensures that the inner product between different coded sequences is zero, indicating that they are orthogonal to each other in the signal space and can be transmitted in parallel in the same frequency band without interference. That is, for two sets of coded sequences C1 and C2, C1×C2=0 is satisfied, indicating that they do not interfere with each other at the physical layer.

[0033] By using pre-allocated orthogonal physical layer coding sequences, the low-dimensional feature vectors of each sensor are mapped onto different carriers. These vectors are then transmitted back through the same communication channel using either time-division or code-division multiplexing (TDM) methods. This asynchronous transmission modulates the low-dimensional feature vectors onto carriers of the same frequency for information transmission. A carrier is a signal carrier used to transmit information; a carrier of the same frequency means all sensors use the same frequency, but the phase and amplitude of each sensor's carrier can be distinguished, thus avoiding interference between different signals. Asynchronous transmission uses either time-division or code-division multiplexing to transmit data through the same communication channel. Time-division multiplexing divides time into different time slots, with each sensor transmitting data within its pre-allocated slot. For example, in a 1-second monitoring period with 5 sensors, the 1-second interval is divided into 5 0.2-second time slots, with each sensor transmitting its own low-dimensional feature vector data within its corresponding slot. Code-division multiplexing assigns a unique orthogonal coding sequence to each sensor. When transmitting data, the sensor multiplies its low-dimensional feature vector by the coding sequence, and the medical information platform extracts the data from each sensor through decoding. For example, sensor A's encoding sequence is [1,-1,1,-1], and sensor B's encoding sequence is [1,1,-1,-1]. These two sets of encoding sequences can be transmitted in parallel within the same frequency band without interfering with each other, based on the orthogonality principle. When sensor A transmits the low-dimensional feature vector [2,3], the actual transmitted signal is [2×1,2×(-1),3×1,3×(-1)]=[2,-2,3,-3]. When sensor B transmits the low-dimensional feature vector [4,5], the actual transmitted signal is [4×1,4×1,5×(-1),5×(-1)]=[4,4,-5,-5]. The medical information platform decodes the received mixed signal according to the different encoding sequences and extracts the data from sensor A and sensor B respectively.

[0034] By converting the high-dimensional original waveform into a low-dimensional feature vector and then performing modulation coding, the amount of data is reduced, the efficiency and reliability of data transmission are improved, and a large amount of monitoring data can be transmitted in a timely and accurate manner under limited communication bandwidth, thereby improving the effectiveness and reliability of data fusion monitoring.

[0035] The medical information platform receives sensor-coded data and, for the monitoring task, performs threshold polling processing under the threshold processing link reorganization and migration based on the platform's embedded functional threshold pool to generate task monitoring data.

[0036] Furthermore, based on the platform's embedded functional threshold pool, the threshold processing link is reorganized and migrated, including: the task parsing engine receives the monitoring task, performs task intent interpretation and decomposition coupling of the smallest task unit, and generates a task combination; for the task combination, a dynamic threshold decision package is generated based on data access, threshold functions, and execution order; based on the dynamic threshold decision package, microprocessing threshold calls and reorganization orchestration are performed in the functional threshold pool to determine the threshold processing link; and the threshold processing link is migrated to the data processing center.

[0037] Specifically, continuous vital sign monitoring refers to the uninterrupted monitoring of basic vital signs such as heart rate, blood pressure, and respiratory rate; key indicator tracking refers to the continuous monitoring of critical physiological indicators, such as blood glucose levels; and abnormal event capture refers to the acquisition of abnormal situations occurring in the physiological state, such as severe arrhythmias. The task parsing engine utilizes natural language processing technology and semantic analysis methods based on medical knowledge graphs to perform coupled operations of task intent interpretation and minimum task unit decomposition. Natural language processing technology can understand the semantic information in the monitoring task text and identify key task elements, while the medical knowledge graph contains rich medical knowledge and experience, providing professional support for task parsing. Natural Language Processing (NLP) technology preprocesses the user-input monitoring task text, including word segmentation, part-of-speech tagging, and named entity recognition, extracting key information from the task and identifying semantic information such as monitoring type and target monitoring indicators through contextual analysis. During task interpretation, a medical knowledge graph provides comprehensive professional knowledge of medical terminology, standards, and treatment procedures. The medical knowledge graph is constructed using nodes and edges. Nodes represent various medical concepts, such as medical terminology, standards, and treatment procedures, while edges represent relationships between concepts, such as causal and synergistic relationships. It can be constructed using standardized medical terminology and conceptual structures provided by public medical databases, combined with expert knowledge. NLP technology, combined with entity relationships in the medical knowledge graph and the medical knowledge base, performs precise semantic matching and understanding of medical terms in the monitoring task, such as heart rate and blood oxygen. For example, through a knowledge graph query, it confirms that the heart rate indicator corresponds to an electrocardiogram (ECG) sensor, and identifies the normal range and abnormal thresholds of the indicator.

[0038] By combining the results of natural language processing (NLP) analysis with a medical knowledge graph, the task parsing engine can accurately interpret task intent and identify monitoring targets and specific operational requirements within the task. Through parsing, the monitoring task is broken down into multiple smallest subtasks, each corresponding to an independently executable monitoring target or processing operation. For example, continuous vital sign monitoring is decomposed into subtasks for continuous monitoring of individual indicators such as heart rate, blood pressure, and blood oxygen saturation, thus generating task combinations. Each subtask in the task combination has its corresponding threshold requirements. For example, a heart rate threshold: if the heart rate is higher or lower than a set threshold, such as a heart rate exceeding 120 beats per minute, a warning is triggered. A blood oxygen saturation threshold: when blood oxygen saturation is below 90%, an alarm is triggered. For each task combination, a corresponding dynamic threshold decision package is generated based on data admission, threshold functions, and execution order. Data admission specifies data quality, i.e., which data can enter the subsequent processing flow. For example, for a heart rate monitoring task, only heart rate data that meets the frequency range and signal quality requirements is allowed, excluding abnormal data caused by sensor malfunction or interference. Threshold functions refer to setting different threshold values ​​for different task units, such as heart rate threshold and blood oxygen threshold. For heart rate monitoring, the heart rate threshold is set to 60-100 beats / minute. When the heart rate is below 60 beats / minute or above 100 beats / minute, an early warning mechanism is triggered. The execution order determines the order in which each task unit is processed. For example, critical vital signs are processed first, followed by other relatively minor indicators. Taking the key indicator tracking task as an example, assuming it is blood pressure monitoring for hypertensive patients, the task combination includes systolic blood pressure monitoring and diastolic blood pressure monitoring. In terms of data access, the systolic blood pressure data range is specified as 80-220 mmHg, and the diastolic blood pressure data range is 50-140 mmHg. The threshold function sets the normal range for systolic blood pressure to 90-139 mmHg and the normal range for diastolic blood pressure to 60-89 mmHg. When systolic blood pressure ≥140 mmHg or diastolic blood pressure ≥90 mmHg, a hypertension early warning is triggered, and the execution order is to process the systolic blood pressure data first, followed by the diastolic blood pressure data. By setting parameters, a dynamic threshold decision package is generated for the continuous blood pressure monitoring task.

[0039] Based on the dynamic threshold decision package, microprocessor thresholds are invoked from the functional threshold pool and reorganized to form a threshold processing chain. The functional threshold pool is a resource library storing various threshold processing functions and rules to ensure rapid and accurate data processing and threshold judgment when monitoring data input. Microprocessor threshold invocation refers to matching and invoking candidate microprocessor thresholds from the functional threshold pool according to the requirements in the dynamic threshold decision package. Reorganization and orchestration combine the matched and invoked candidate unprocessed thresholds in the execution order to form a complete threshold processing chain. For example, in the aforementioned blood pressure monitoring task for hypertensive patients, the normal range judgment thresholds for systolic and diastolic blood pressure, hypertension warning thresholds, etc., are invoked from the functional threshold pool. The chain is reorganized and orchestrated in the order of first judging systolic blood pressure, then judging diastolic blood pressure, and finally triggering the corresponding warning based on the judgment result, forming the threshold processing chain. After determining the threshold processing chain, it is migrated to the data processing center to achieve efficient threshold processing of sensor-coded data and generate task monitoring data.

[0040] By efficiently managing monitoring tasks through intent interpretation and decomposition into the smallest task units, and generating dynamic threshold decision packages, the system ensures accurate and timely task execution and data processing. Through microprocessor threshold invocation and reorganization orchestration in the threshold pool, it can flexibly adapt to different monitoring needs and ensure smooth migration of the data processing chain.

[0041] Furthermore, before microprocessor threshold invocation in the functional threshold pool, the construction of the functional threshold pool includes: retrieving monitoring records and reconstructing them into M monitoring sequences representing the data source-functional logic; traversing the M monitoring sequences and performing decomposition based on the smallest functional logic unit to determine N unit monitoring sequences, where N is a positive integer greater than or equal to M; determining X functional monitoring sequences by clustering the N unit monitoring sequences, where X is a positive integer less than or equal to M; and constructing and integrating microprocessor thresholds based on the X functional monitoring sequences to form the functional threshold pool.

[0042] Specifically, past monitoring records are retrieved from the medical information database of the medical information platform. These records contain various data from different monitoring tasks at different time periods and in different monitoring scenarios, such as monitoring data for indicators like heart rate, blood pressure, and blood oxygen saturation. Each monitoring record includes a timestamp, the data acquisition device, and the data value. Each monitoring record is labeled according to its data source. For example, data from an electrocardiogram (ECG) monitoring device is labeled as ECG data, and data from a pulse oximeter is labeled as blood oxygen data. Through reconstruction technology, the labeled monitoring records are reconstructed into M monitoring sequences according to the relationship between data source and functional logic. The data source refers to the specific monitoring device or monitoring item that generates the data, such as an ECG monitoring device or a pulse oximeter. The functional logic describes the physiological function or monitoring purpose reflected by the data, such as monitoring cardiac electrical activity or monitoring the oxygen-carrying capacity of the blood. Each monitoring sequence represents data collected by a specific sensor within a specific time period and its corresponding functional logic, such as heart rate collected by an ECG or blood oxygen saturation collected by a pulse oximeter. M is a positive integer.

[0043] The process iterates through M monitoring sequences, using minimum functional logic units (FLUs) to break each sequence down into smaller task units, ensuring each unit can independently perform data acquisition, processing, or threshold judgment. Taking an electrocardiogram (ECG) monitoring sequence as an example, the ECG signal can be decomposed into minimum functional logic units such as P waves, QRS complexes, and T waves. Each unit reflects a different stage of cardiac electrical activity. After decomposition, N unit monitoring sequences are determined, where N is a positive integer greater than or equal to M. Clustering algorithms, such as K-means clustering or hierarchical clustering, are used to group the unit monitoring sequences based on feature similarity. Using K-Means as an example, the number of clusters K is determined using the elbow method. By calculating the distance between each unit monitoring sequence and its cluster center, the sequence is assigned to the nearest cluster. The cluster centers are then continuously updated until the clustering results stabilize. After clustering, X functional monitoring sequences are obtained, where X is a positive integer less than or equal to M. For example, with N = 6 unit monitoring sequences, clustering divides each unit monitoring sequence into 3 functional monitoring sequences, used for heart rate monitoring, blood oxygen monitoring, and body temperature monitoring, respectively. After clustering, microprocessor thresholds are constructed and integrated based on the needs of each functional monitoring sequence. These thresholds, tailored to specific monitoring sequences and functional logic, incorporate medical expertise and clinical experience to set corresponding threshold values ​​and processing rules. For example, heart rate monitoring sets a threshold of 60-100 beats / minute; an alarm is triggered when the heart rate exceeds this range. By integrating microprocessor thresholds for different functional monitoring sequences into a unified threshold pool, these thresholds can be dynamically invoked, ensuring that appropriate threshold rules are applied in different monitoring tasks.

[0044] By decomposing monitoring tasks into the smallest units and performing clustering and threshold construction, precise control and efficient management of threshold rules can be achieved, avoiding redundant data processing. Different threshold processing chains can be dynamically invoked according to the actual needs of the task, ensuring that the task can respond accurately based on real-time data, thereby improving the accuracy and timeliness of medical data fusion monitoring during medical transport.

[0045] Furthermore, microprocessor threshold invocation and reorganization orchestration are performed in the functional threshold pool to determine the threshold processing link, including: based on the dynamic threshold decision package, matching invocation is performed in the functional threshold pool using threshold functions to determine candidate microprocessor thresholds; input initialization is performed on each candidate microprocessor threshold according to data admission; and the initialized candidate microprocessor thresholds are associated and reorganized according to the execution order to form the threshold processing link, wherein the execution order is determined by associating the data interfaces between each candidate microprocessor threshold through serial or parallel relationships.

[0046] Specifically, based on a dynamic threshold decision package and indexed by threshold functions, efficient retrieval algorithms, such as hash retrieval or tree-based retrieval, are used to quickly and accurately find candidate microprocessor thresholds that match the threshold functions in the dynamic threshold decision package from the functional threshold pool. Taking hash retrieval as an example, a hash table is constructed based on the threshold functions of each microprocessor threshold, such as heart rate monitoring and blood oxygen monitoring. In the hash table, the threshold function serves as the hash key, and each hash key corresponds to a set of microprocessor thresholds. Each microprocessor threshold contains corresponding threshold judgment rules and execution logic, and each threshold function has a unique hash value. When processing monitoring tasks, a threshold function, such as heart rate monitoring, is extracted from the dynamic threshold decision package. A hash algorithm, such as MD5 or SHA, is used to perform hash calculation on the threshold function to obtain a hash value. Using the calculated hash value, the microprocessor threshold used for heart rate judgment can be quickly located in the hash table and used as a candidate microprocessor threshold, achieving efficient threshold invocation and processing.

[0047] Data admission refers to checking whether the data meets the processing requirements according to preset conditions. Based on data admission, the input end of each candidate microprocessor threshold is initialized. For example, for candidate microprocessor thresholds in heart rate monitoring, the data admission rule requires that the input heart rate data must be an integer between 30 and 200 beats per minute. If the heart rate data obtained from the sensor is a floating-point number, the data format is converted to an integer. At the same time, the data is range-validated. If the data is not within the 30-200 beats per minute range, it is marked as invalid data and will not proceed to subsequent threshold processing. The initialized microprocessor thresholds are associated and reorganized according to the execution order. The execution order clarifies the order in which each candidate microprocessor threshold is processed. Data interface associations between candidate microprocessor thresholds are established through serial or parallel relationships. When there is a data dependency between multiple microprocessor thresholds, they must be executed sequentially, using a serial relationship. For example, heart rate monitoring performs feature extraction and threshold judgment after data preprocessing before triggering alarms. When certain physiological indicators can be processed simultaneously without a strict order, a parallel processing approach is adopted. For example, simultaneously monitoring blood oxygen saturation and body temperature. After associating and recombining the initialized candidate microprocessor thresholds through serial or parallel processing, a complete threshold processing chain is formed. This chain defines the entire process from data acquisition to alarm or decision output. For example, in a comprehensive vital sign monitoring task, the threshold processing chain first processes heart rate and blood pressure serially, then processes blood oxygen saturation and body temperature in parallel, and finally integrates all processing results to determine whether to trigger an early warning mechanism. For instance, a dynamic threshold decision package requires monitoring heart rate, blood pressure, and blood oxygen saturation. In the functional threshold pool, three candidate microprocessor thresholds—heart rate judgment, blood pressure judgment, and blood oxygen saturation judgment—are matched and invoked. According to the data admission rules, heart rate data must be an integer between 30 and 200 beats per minute; blood pressure data must be systolic between 80 and 220 mmHg and diastolic between 50 and 140 mmHg; and blood oxygen saturation data must be between 70% and 100%. The inputs to these three candidate microprocessor thresholds are initialized to ensure the input data meets the requirements. In terms of execution order, heart rate and blood pressure are processed sequentially first; that is, the heart rate is checked for normality first, and if normal, the blood pressure is checked next. Then, blood oxygen saturation is processed in parallel, and its status is checked simultaneously. Finally, the results of the heart rate, blood pressure, and blood oxygen saturation checks are combined. If any of these indicators is abnormal, an early warning mechanism is triggered.

[0048] Based on the specific monitoring task, appropriate microprocessor thresholds are precisely called from the functional threshold pool, and through reasonable initialization settings and correlation reorganization, an efficient threshold processing link is formed, enabling the medical information platform to process sensor coded data quickly, comprehensively, and accurately, detect abnormalities in a timely manner, and improve the quality and efficiency of medical monitoring during medical transport.

[0049] Furthermore, threshold polling processing is performed to generate task monitoring data, including: transmitting the sensor-encoded data to the data processing center and activating the migrated threshold processing link; according to the threshold processing link, performing threshold polling processing under threshold admission filtering on the sensor-encoded data to determine task monitoring data, wherein the task monitoring data includes threshold monitoring chain and comprehensive monitoring data.

[0050] Furthermore, the threshold processing link is a directed acyclic type; after the monitoring task ends, the threshold processing link is decomposed and migrated back to the functional threshold pool.

[0051] Specifically, the collected sensor-encoded data is transmitted to the data processing center. At the data processing center, the migrated threshold processing link is activated. This threshold processing link is a directed acyclic structure (DAG) containing the nodes for threshold judgment and processing, along with their execution order. In other words, the threshold processing link consists of multiple microprocessor thresholds connected according to a specific execution order and data flow direction, and there are no loops, thus avoiding deadlocks and ensuring the efficiency and logic of data processing. For example, in a blood pressure monitoring task, the threshold processing link sequentially includes a data preprocessing threshold, a systolic blood pressure judgment threshold, a diastolic blood pressure judgment threshold, and a comprehensive warning threshold. The migration operation involves retrieving the pre-built threshold processing link from the functional threshold pool and transmitting it to the data processing center via secure communication protocols such as HTTPS and MQTT. The threshold processing link is then deployed into the data processing center's runtime environment. During deployment, modular deployment of the link is ensured, meaning each microprocessor threshold runs on an independent node in the data processing center, achieving efficient processing and computation of the sensor-encoded data.

[0052] Based on the activated threshold processing chain, threshold polling processing under threshold admission filtering is performed on the sensor-coded data. Threshold admission filtering refers to filtering out data that meets the processing requirements according to thresholds, and removing invalid data, such as abnormal data or data exceeding the preset range. During processing, each microprocessor threshold judges the input sensor-coded data according to its corresponding data admission rules and outputs the judgment result. For example, for heart rate monitoring, the accuracy rule stipulates that the heart rate data should be between 30 and 200 beats / minute. If the collected heart rate data is 25 beats / minute, it will be filtered out and will not enter the subsequent heart rate judgment threshold processing. Through threshold admission filtering, the sensor-coded data can be comprehensively analyzed, reducing the interference of invalid data.

[0053] After passing the threshold admission filtering, the process enters the threshold polling stage. Following the execution order specified in the threshold processing chain, the sensor-coded data is sequentially subjected to threshold judgments. Each threshold judgment task, such as heart rate judgment or blood oxygen judgment, is evaluated based on the threshold conditions set for the task. For example, if the heart rate exceeds 100 beats / minute, an alarm is triggered. The filtered sensor-coded data is then judged, and the corresponding processing result is output. The entire process is completed sequentially through polling, ensuring that each data point undergoes comprehensive evaluation. The task monitoring data includes the threshold monitoring chain and comprehensive monitoring data. The threshold monitoring chain records the judgment results and processing information of each microprocessor threshold that the sensor-coded data passes through in the threshold processing chain, such as the input data and output results for each threshold. Output results include, for example, heart rate exceeding the threshold or blood oxygen being normal. Comprehensive monitoring data is an overall monitoring conclusion obtained by comprehensively analyzing the judgment results in the threshold monitoring chain: when all judgment results are within their respective threshold ranges, the current vital signs are judged to be in a normal state; conversely, if any judgment result exceeds the predetermined threshold range, it indicates an abnormal situation. By integrating the threshold monitoring chain and comprehensive monitoring data into task monitoring data, a comprehensive and detailed monitoring report can be provided.

[0054] Once the monitoring task is completed, the executed threshold processing chain is decomposed into individual microprocessor thresholds and migrated back to the functional threshold pool. The migration process involves deconstructing the executed microprocessor thresholds and chains and storing them again in the threshold pool so that they can be called again in other monitoring tasks, thereby achieving resource sharing and reuse and improving the overall efficiency and flexibility of data processing.

[0055] By transmitting coded data to the data processing center and activating the threshold processing link, data from various sensors can be processed in real time and accurately. By forming a threshold monitoring chain and comprehensive monitoring data, the accuracy and reliability of task monitoring are improved, thereby effectively enhancing the monitoring effect. Through the decomposition and relocation of the threshold processing link, the redundant process of repeatedly constructing the threshold link is avoided, improving resource utilization efficiency.

[0056] The task monitoring data is matched with early warnings based on information early warning rules, and dual-channel early warning management is implemented on the platform's visual port and mobile terminal.

[0057] Furthermore, the task monitoring data is matched with early warning rules based on information warning rules, and dual-path early warning management is implemented on the platform's visual interface and mobile terminals. This includes: deploying information warning rules within the medical information platform, wherein the information warning rules are jointly defined by multiple transfer impact types and multiple early warning levels; performing matching judgment based on the information warning rules according to the task monitoring data, and generating targeted early warning instructions; displaying the task monitoring data and targeted early warning instructions in a pop-up window on the visual interface of the medical information platform, and sending the targeted early warning instructions to the mobile terminals of the transfer personnel for early warning management.

[0058] Specifically, within the medical information platform, information early warning rules are set and deployed. These rules are jointly defined by multiple types of transport impacts and multiple early warning levels. The multiple types of transport impacts refer to the effects of different transport states or conditions on the patient's health during medical transport, such as abnormal vital signs, unsuitable temperature or humidity during transport, and other transport-related factors. Multiple early warning levels are different levels based on the severity of the transport impact types, and can be divided into low-level, medium-level, and high-level early warnings. Different early warning levels are set according to the severity of the monitored data to ensure corresponding response measures for different situations. When vital signs fluctuate slightly, such as a temperature increase within 1 degree Celsius, it is defined as a low-level early warning; when vital signs are significantly abnormal but not yet life-threatening, it is defined as a medium-level early warning, such as a blood pressure increase of 140 / 90 mmHg, which is in the early stages of hypertension and may cause discomfort but is not life-threatening. Extremely unstable vital signs that could endanger life at any moment are defined as a high-level warning. For example, severe hypoxia may lead to multiple organ damage or shock, requiring immediate intervention. A blood oxygen saturation level below 90% is also defined as a high-level warning, as is an extremely high or low heart rate. A heart rate greater than 150 beats / minute or less than 40 beats / minute indicates severe cardiac dysfunction and may require immediate cessation of transport, emergency treatment, or other urgent medical interventions. The deployment and management of information warning rules are achieved using rule engine technology. A rule engine is a component embedded in an application that separates business rules from application code, allowing for independent storage and management. For example, using the Drools rule engine, information warning rules can be flexibly defined, modified, and updated according to actual needs, improving the maintainability and scalability of analysis and processing.

[0059] The system matches task monitoring data with information early warning rules one by one, analyzes the status of each monitoring indicator, and determines whether an early warning is triggered based on the defined rules. Targeted early warning instructions are then generated, clearly indicating the specific issues requiring attention and the warning level. For example, a high-priority alert is triggered when the heart rate exceeds the normal range and blood oxygen levels are low; a mild warning is generated if body temperature is slightly elevated but other physiological indicators are normal. On the visualization interface of the medical information platform, pop-up windows display the task monitoring data and targeted early warning instructions, showing abnormal information in the task monitoring data and the corresponding warning level. Targeted early warning instructions are also sent to the mobile terminals of transport personnel for early warning management, such as via SMS or app push notifications, reminding them to take appropriate measures.

[0060] By deploying information early warning rules, scientific and reasonable early warning judgments can be made based on the patient's actual situation and various factors during the transfer process. Matching judgments based on information early warning rules can promptly identify problems and potential risks in patients. The dual-path early warning management of the platform's visual portal and mobile terminals ensures that medical personnel and transfer personnel can obtain early warning information in a timely manner and take corresponding measures. This further improves the timeliness, accuracy, efficiency, and pertinence of multi-dimensional data fusion monitoring during the transfer process, enhances monitoring effectiveness, and ensures patient safety.

[0061] Example 2, based on the same inventive concept as the multi-dimensional data fusion monitoring method in the foregoing examples, such as... Figure 2 As shown, this application provides a multi-dimensional data fusion monitoring system, wherein the multi-dimensional data fusion monitoring system includes: The sensor activation module 11 is used to upload monitoring tasks from the user side to the medical information platform, drive the transport equipment to activate the multimodal sensor array, perform synchronous asynchronous non-uniform sampling and sensor-side modulation, and encode the data into the synchronous carrier of the communication return. The data processing module 12 is used for the medical information platform to receive the sensor-encoded data, and for the monitoring task, perform threshold polling processing under the link reorganization migration based on the threshold pool embedded in the platform to generate task monitoring data. The early warning management module 13 is used to perform early warning matching based on information early warning rules on the task monitoring data, and perform dual-path early warning management on the platform visualization port and mobile terminal.

[0062] Furthermore, the data processing module 12 is also used for: receiving the monitoring task by the task parsing engine, performing task intent interpretation and minimum task unit decomposition coupling to generate a task combination; generating a dynamic threshold decision package for the task combination based on data access, threshold functions and execution order; performing microprocessing threshold invocation and reorganization orchestration in the functional threshold pool according to the dynamic threshold decision package to determine the threshold processing link; and migrating the threshold processing link to the data processing center.

[0063] Furthermore, the data processing module 12 is also used to: retrieve monitoring records and reconstruct them into M monitoring sequences representing the data source-functional logic; traverse the M monitoring sequences, perform decomposition based on the smallest functional logic unit, and determine N unit monitoring sequences, where N is a positive integer greater than or equal to M; determine X functional monitoring sequences by clustering the N unit monitoring sequences, where X is a positive integer less than or equal to M; and construct and integrate microprocessor thresholds based on the X functional monitoring sequences to form the functional threshold pool.

[0064] Furthermore, the data processing module 12 is also used to: determine candidate microprocessor thresholds by matching and calling the functional threshold pool based on the dynamic threshold decision package and using the threshold function; initialize the input end of each candidate microprocessor threshold according to data admission; and associate and reorganize the initialized candidate microprocessor thresholds according to the execution order to form the threshold processing link, wherein the execution order associates the data interfaces between each candidate microprocessor threshold through serial or parallel relationships.

[0065] Furthermore, the data processing module 12 is also used to: transmit the sensor-encoded data to the data processing center and activate the migration threshold processing link; perform threshold polling processing under threshold admission filtering on the sensor-encoded data according to the threshold processing link to determine task monitoring data, wherein the task monitoring data includes threshold monitoring chain and comprehensive monitoring data.

[0066] Furthermore, the data processing module 12 is also used to: the threshold processing link is a directed acyclic type; after the monitoring task ends, the threshold processing link is decomposed and migrated back to the functional threshold pool.

[0067] Furthermore, the sensing activation module 11 is also used for: the medical information platform, by parsing the monitoring task, allocating a time-priority-based periodic asynchronous sampling window for the multimodal sensors assembled on the transport equipment; issuing a task sampling instruction according to the periodic asynchronous sampling window, driving the multimodal sensors to perform source-end sensing acquisition, and determining the multimodal sensing signal; and performing original waveform modulation on the multimodal sensing signal and encoding it to the same-frequency carrier of the communication return.

[0068] Furthermore, the sensing activation module 11 is also used to: perform real-time feature processing on the original waveform of the multi-threaded system according to lightweight signal processing rules to determine a low-dimensional feature vector; and modulate the low-dimensional feature vector onto a carrier of the same frequency for asynchronous backhaul through a pre-allocated orthogonal physical layer coding sequence, wherein the carrier of the same frequency is a carrier of the same frequency as the platform command but whose phase and amplitude are distinguishable, and the asynchronous backhaul method is to backhaul through the same communication channel via time division or code division.

[0069] Furthermore, the early warning management module 13 is also used to: deploy information early warning rules within the medical information platform, wherein the information early warning rules are jointly defined by multiple transfer impact types and multiple early warning levels; perform matching judgment based on the information early warning rules according to the task monitoring data, and generate targeted early warning instructions; display the task monitoring data and targeted early warning instructions in a pop-up window on the visualization port of the medical information platform, and send the targeted early warning instructions to the mobile terminals of the transfer personnel for early warning management.

[0070] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The multi-dimensional data fusion monitoring method and specific examples in the foregoing embodiment one are also applicable to the multi-dimensional data fusion monitoring system in this embodiment. Through the foregoing detailed description of the multi-dimensional data fusion monitoring method, those skilled in the art can clearly understand the multi-dimensional data fusion monitoring system in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0071] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0072] Obviously, those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for fusion and monitoring of multi-dimensional data, characterized in that, The method includes: The user uploads the monitoring task to the medical information platform, which drives the transport equipment to activate the multimodal sensor array, performs synchronous asynchronous non-uniform sampling and sensor-side modulation, and encodes it into the synchronous carrier of the communication return. The medical information platform receives sensor-coded data and, for the monitoring task, performs threshold polling processing under the threshold processing link reorganization and migration based on the platform's embedded functional threshold pool to generate task monitoring data. The task monitoring data is matched with early warnings based on information early warning rules, and dual-channel early warning management is implemented on the platform's visual port and mobile terminal.

2. The multi-dimensional data fusion monitoring method as described in claim 1, characterized in that, Based on the platform's embedded functional threshold pool, threshold processing links are reorganized and migrated, including: The task parsing engine receives the monitoring task, performs task intent interpretation and decomposition of the smallest task unit, and generates a task combination; For the aforementioned task combination, a dynamic threshold decision package is generated based on data admission, threshold functions, and execution order. Based on the dynamic threshold decision package, microprocessor threshold invocation and reorganization are performed in the functional threshold pool to determine the threshold processing link; The threshold processing link will be migrated to the data processing center.

3. The multi-dimensional data fusion monitoring method as described in claim 2, characterized in that, Before microprocessor threshold calls are made in the function threshold pool, the construction of the function threshold pool includes: Retrieve monitoring records and reconstruct them into M monitoring sequences that represent the data source and functional logic; Traverse the M monitoring sequences, perform decomposition based on the smallest functional logic unit, and determine N unit monitoring sequences, where N is a positive integer greater than or equal to M; By clustering the N unit monitoring sequences, X functional monitoring sequences are determined, where X is a positive integer less than or equal to M; Based on the X functional monitoring sequences, microprocessor thresholds are constructed and integrated to form the functional threshold pool.

4. The multi-dimensional data fusion monitoring method as described in claim 3, characterized in that, Microprocessor threshold invocation and reorganization orchestration are performed in the functional threshold pool to determine the threshold processing chain, including: Based on the dynamic threshold decision package, the threshold function is used to match and call in the functional threshold pool to determine the candidate microprocessor threshold; Based on the data admission criteria, the input terminals of each candidate microprocessor threshold are initialized. According to the execution order, the initialized candidate microprocessor thresholds are associated and reorganized to form the threshold processing link. The execution order associates the data interfaces between the candidate microprocessor thresholds through serial or parallel relationships.

5. The multi-dimensional data fusion monitoring method as described in claim 4, characterized in that, Perform threshold polling to generate task monitoring data, including: The sensor-encoded data is transmitted to the data processing center to activate the migration threshold processing link; According to the threshold processing chain, threshold polling processing under threshold admission filtering is performed on the sensor encoded data to determine the task monitoring data, wherein the task monitoring data includes the threshold monitoring chain and the comprehensive monitoring data.

6. The multi-dimensional data fusion monitoring method as described in claim 5, characterized in that, The threshold processing link is a directed acyclic type; Once the monitoring task is completed, the threshold processing link is decomposed and migrated back to the functional threshold pool.

7. The multi-dimensional data fusion monitoring method as described in claim 1, characterized in that, The driving transfer equipment activates the multimodal sensor array, performs synchronous asynchronous non-uniform sampling and sensor-side modulation, and encodes the synchronous carrier transmitted back to the communication, including: The medical information platform analyzes the monitoring task and allocates time-priority-based periodic asynchronous sampling windows for the multimodal sensors installed on the transport equipment. According to the periodic asynchronous sampling window, a task sampling instruction is issued to drive the multimodal sensor to perform source-end sensing acquisition and determine the multimodal sensing signal; The original waveform of the multimodal sensing signal is modulated and encoded onto the same frequency carrier for communication backhaul.

8. The multi-dimensional data fusion monitoring method as described in claim 7, characterized in that, Modulating the original waveform of the multimodal sensing signal and encoding it onto the same-frequency carrier for communication return includes: Based on lightweight signal processing rules, real-time feature processing is performed on the original waveforms of multi-threaded signals to determine low-dimensional feature vectors. The low-dimensional feature vector is modulated onto a carrier of the same frequency for asynchronous backhaul using a pre-allocated orthogonal physical layer coding sequence. The carrier of the same frequency is a carrier of the same frequency as the platform command but whose phase and amplitude are distinguishable. The asynchronous backhaul method is to backhaul through time division or code division on the same communication channel.

9. The multi-dimensional data fusion monitoring method as described in claim 1, characterized in that, The task monitoring data is matched for early warning based on information warning rules, and dual-channel early warning management is implemented on both the platform's visual interface and the mobile terminal, including: Information early warning rules are deployed within the medical information platform, wherein the information early warning rules are jointly defined by multiple types of transport impacts and multiple early warning levels; Based on the task monitoring data, a matching judgment based on the information early warning rules is performed to generate a targeted early warning instruction; The task monitoring data and targeted early warning instructions are displayed in a pop-up window on the visualization port of the medical information platform, and the targeted early warning instructions are sent to the mobile terminals of the transport personnel for early warning management.

10. A multi-dimensional data fusion monitoring system, characterized in that, The step of implementing the multi-dimensional data fusion monitoring method according to any one of claims 1 to 9, wherein the multi-dimensional data fusion monitoring system comprises: The sensor activation module is used to upload monitoring tasks from the user side to the medical information platform, drive the transport equipment to activate the multimodal sensor array, perform synchronous asynchronous non-uniform sampling and sensor-side modulation, and encode it into the synchronous carrier of the communication return. The data processing module is used by the medical information platform to receive sensor-coded data and, for the monitoring task, to perform threshold polling processing under the threshold processing link reorganization and migration based on the functional threshold pool embedded in the platform to generate task monitoring data. The early warning management module is used to perform early warning matching based on information early warning rules on the task monitoring data, and to perform dual-channel early warning management on the platform's visual port and mobile terminal.

Images