Contextual directional information casting method and system based on patient real-time location

Abstract

The application discloses a situational directional information delivery method and system based on real-time patient position, and relates to the technical field of wireless communication.The method comprises the following steps: uniformly accessing edge nodes of a multi-source positioning device in a target area to obtain an edge node set; connecting the edge node set with a central cloud node to construct a cloud-edge collaborative positioning network; obtaining an information delivery task according to a preset index by the central cloud node; performing situational positioning on a target patient in combination with the network to obtain position and situational information; aggregating the patients based on the positioning information to obtain an aggregation set; constructing a relationship graph in combination with the aggregation set and analyzing a strategy; and delivering information according to the strategy based on the cloud-edge collaborative network.The technical problem of delivery conflict and low efficiency caused by rule conflict and lack of adaptive mechanism in the prior art is solved, and the technical effects of reducing the misdelivery rate and environmental interference and improving the patient treatment efficiency and medical service quality are achieved.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, and more specifically to a method and system for contextualized directional information delivery based on the real-time location of a patient. Background Technology

[0002] With the rapid development of smart healthcare and IoT technologies, hospitals are continuously improving their digital management capabilities. However, numerous challenges remain in the delivery of information within hospitals. Traditional information dissemination methods rely primarily on manual broadcasting, fixed-terminal push notifications, or unified message notifications, making it difficult to provide personalized and precise information delivery based on patients' real-time locations, stages of their medical visits, and surrounding environments. Especially in complex environments with multiple floors, departments, and terminals, information is prone to delays, duplication, misbroadcasting, or interference, impacting efficiency and patient experience. Meanwhile, advancements in in-hospital positioning technology and wireless communication capabilities have made real-time location-based data sensing possible. Dynamic positioning of patients, medical staff, and equipment allows for comprehensive perception of multi-dimensional information such as hospital operational status, departmental load, waiting area congestion, and noise thresholds, providing a data foundation for accurate information delivery. However, existing solutions largely remain at the static rule level, lacking contextual adaptation and strategy optimization capabilities. They cannot dynamically adjust broadcast strategies based on real-time feedback, making it difficult to ensure information delivery rates while avoiding excessive disruption. Summary of the Invention

[0003] This application provides a contextualized targeted information delivery method and system based on the real-time location of patients, which solves the technical problems of delivery conflicts and low efficiency caused by rule conflicts and lack of adaptive mechanisms in existing in-hospital information delivery.

[0004] The first aspect of this application provides a method for contextualized targeted information delivery based on the real-time location of a patient, the method comprising:

[0005] A unified edge node access is performed on a set of multi-source indoor positioning devices in the target area to obtain an edge node set. This edge node set is then connected to a central positioning cloud node to construct a cloud-edge collaborative positioning network. The central positioning cloud node interacts with the network, and delivery requests are retrieved according to preset delivery indicators to obtain K information delivery tasks, where K is a positive integer. K target patients for these K information delivery tasks are identified, and contextualized positioning is performed on these K target patients using the cloud-edge collaborative positioning network to obtain their locations and contextualized positioning information. Based on the contextualized positioning information and locations of the K target patients, they are aggregated into M aggregated target patient sets, where M is a positive integer less than or equal to K. An interaction relationship graph structure is constructed for the K information delivery tasks using the M aggregated target patient sets, and information delivery strategies are analyzed based on the construction results. Finally, the cloud-edge collaborative positioning network delivers targeted information to the K information delivery tasks according to the information delivery strategy.

[0006] A second aspect of this application provides a contextualized, directional information delivery system based on the real-time location of a patient, the system comprising:

[0007] Communication Connection Module: Unifies edge node access for the multi-source indoor positioning device set in the target area to obtain an edge node set, and connects the edge node set to the central positioning cloud node to construct a cloud-edge collaborative positioning network; Delivery Request Retrieval Module: Interacts with the central positioning cloud node to retrieve delivery requests according to preset delivery indicators, obtaining K information delivery tasks, where K is a positive integer; Contextualized Positioning Module: Obtains K target patients for the K information delivery tasks, and performs contextualized positioning of the K target patients in conjunction with the cloud-edge collaborative positioning network to obtain K target patient... The system includes: a patient location and contextualized location information for K target patients; a similar aggregation module: based on the contextualized location information and the location of the K target patients, the system performs similar aggregation on the K target patients to obtain M aggregated target patient sets, where M is a positive integer less than or equal to K; and a targeted information delivery module: combining the M aggregated target patient sets, the system constructs an interaction relationship graph structure for the K information delivery tasks, and parses the information delivery strategy based on the construction result, and performs targeted information delivery for the K information delivery tasks based on the cloud-edge collaborative positioning network according to the information delivery strategy.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] First, multi-source indoor positioning devices within the hospital are uniformly accessed and managed. A cloud-edge collaborative positioning network is constructed through communication between edge nodes and the central cloud node, achieving high-precision positioning of personnel and equipment within the hospital. Then, the central cloud node retrieves and generates information delivery tasks based on preset delivery indicators. After acquiring the target patient corresponding to the task, the cloud-edge collaborative positioning network is used to perform contextualized positioning of the patient, determining not only their specific location but also their environment and state. Next, based on the patient's positioning and contextual information, patients with similar characteristics are aggregated and classified, forming several aggregated target patient sets. Finally, the interaction graph of the information tasks is constructed by combining these aggregation results, the optimal information delivery strategy is analyzed, and the information is pushed to the corresponding terminals through the cloud-edge collaborative network, achieving personalized and contextualized precise information delivery. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic flowchart of a contextualized targeted information delivery method based on the real-time location of a patient, provided in an embodiment of this application.

[0012] Figure 2 This is a schematic diagram of the structure of a contextualized orientation information delivery system based on the real-time location of a patient, provided in an embodiment of this application.

[0013] Explanation of reference numerals in the attached diagram: 11. Communication connection module; 12. Demand retrieval module; 13. Contextualized positioning module; 14. Similar aggregation module; 15. Targeted information delivery module. Detailed Implementation

[0014] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0015] Example 1, as Figure 1 As shown, this application provides a method for contextualized targeted information delivery based on the real-time location of a patient, the method including:

[0016] A unified edge node access is performed on the set of multi-source indoor positioning devices in the target area to obtain an edge node set. The edge node set is then connected to the central positioning cloud node to construct a cloud-edge collaborative positioning network.

[0017] In this embodiment, various indoor positioning devices within the multi-source indoor positioning device set in the target area, such as Wi-Fi access points, Bluetooth base stations, and UWB devices, are first uniformly connected as edge nodes. These edge nodes, acting as intermediate layer devices, are primarily responsible for signal reception and processing with the positioning devices. By connecting this set of edge nodes to the central positioning cloud node, a collaborative positioning network structure is established, forming a cloud-edge collaborative positioning architecture. In this cloud-edge collaborative positioning network architecture, the central positioning cloud node is not only responsible for receiving real-time data from the edge nodes but also dynamically allocates tasks according to the region and needs, ensuring efficient operation of the entire positioning system in a large-scale, complex environment and ensuring efficient cooperation and accurate positioning of multiple terminal devices in different areas within the facility.

[0018] The centrally located cloud node is interacted with to retrieve delivery requests according to preset delivery indicators and obtain K information delivery tasks, where K is a positive integer.

[0019] In one embodiment, the central positioning cloud node acts as the core of the system's information delivery management, responsible for intelligently retrieving information requests and generating tasks based on preset delivery indicators. Specifically, the central positioning cloud node establishes interactive connections with the hospital information system, electronic medical record system, and in-hospital workflow system through data interfaces to acquire multi-source data in real time, including patient status, treatment process, appointment information, and terminal device status. Subsequently, it reads preset delivery indicators, such as target patients, geofences, and waiting area load. When a new information delivery request arises, the central positioning cloud node uses these delivery indicators to match and filter data through key-name comparison, automatically retrieving the corresponding delivery indicator data from the delivery requests. It then aggregates delivery indicator data belonging to the same patient, abstracting it into K independent information delivery tasks, where K is a positive integer. Each information delivery task contains core elements such as the corresponding target patient identifier, location tag, and content type. These information delivery tasks provide standardized inputs for subsequent contextualized positioning, task aggregation, and strategy generation, ensuring that information delivery is logically traceable and execution is controllable, thereby achieving intelligent, manageable, and efficient information transmission.

[0020] Furthermore, the preset delivery indicators include target patients, geofence, waiting load, time window, and terminal type.

[0021] Preferably, preset delivery indicators are used to guide the retrieval, filtering, and allocation of information tasks to achieve precise information delivery and dynamic control. These indicators mainly include target patients, geofencing, waiting room load, time windows, and terminal types. Among them, the target patient indicator is used to determine the recipient of information delivery, which can be identified and bound based on the patient's unique identification, such as hospital number, consultation number, or wristband ID. The geofencing is used to limit the spatial range of information delivery, and can dynamically generate corresponding virtual geographical boundaries based on the coordinate data of the patient's location, thereby controlling the spatial orientation of the information. The waiting room load is used to reflect the real-time flow density and service pressure of each department or area. The priority and frequency of information push can be adjusted according to this parameter to avoid interference in high-load areas. The time window is used to limit the time conditions of information delivery, such as triggering tasks only during specific time periods before patients wait for their appointments, before examinations, or after treatment, to ensure the timeliness and relevance of the information. The terminal type is used to determine the information delivery carrier, including different forms such as bedside screens, corridor speakers, nurse station terminals, doctor mobile terminals, or patient mobile phones.

[0022] K target patients for K information delivery tasks are obtained, and the K target patients are contextualized and located using the cloud-edge collaborative positioning network to obtain the location of the K target patients and the contextualized positioning information of the K target patients.

[0023] In one embodiment, after acquiring K information delivery tasks, K corresponding target patients are extracted from these tasks. These target patients are selected and matched by the central positioning cloud node according to preset delivery indicators. Subsequently, the K target patients are contextually located using a cloud-edge collaborative positioning network. That is, the positioning signals of the patient's location are collected in real time through an edge node set, and then the positioning signals are corrected using a scene fingerprint database to determine the location of the K target patients. Afterward, the location of the K target patients is matched and retrieved using the target area scene model of the central positioning cloud node to obtain more detailed contextual positioning information of the K target patients, providing comprehensive and accurate decision support for subsequent targeted information delivery strategies.

[0024] Furthermore, for K target patients in the K information delivery task, the cloud-edge collaborative positioning network is used to perform contextualized positioning of the K target patients, obtaining the locations of the K target patients and their contextualized positioning information, including:

[0025] The edge node set of the cloud-edge collaborative positioning network is used to collect and preprocess the terminal signals of the K target patients to obtain K initial positioning information; the K initial positioning information is corrected by combining the scene fingerprint database to determine the location of the K target patients; the location of the K target patients is matched and retrieved with the target area scene model of the central positioning cloud node to obtain the contextualized positioning information of the K target patients.

[0026] Preferably, the terminal signals of K target patients are collected in real time through a set of edge nodes in a cloud-edge collaborative positioning network. These terminal signals can originate from terminal devices carried by the target patients, such as mobile phones, wearable devices, or bed monitoring devices. Preliminary positioning information for the K target patients is obtained through preliminary processing of the collected terminal signals, including denoising, signal strength standardization, and time alignment. This information is typically presented in two-dimensional coordinates. Denoising can be performed using filtering algorithms such as Kalman filtering and mean filtering; signal strength standardization can be performed by normalizing the signal strength values; and time alignment can be performed using time synchronization algorithms such as timestamp synchronization and interpolation methods. Subsequently, since the initial positioning information often contains some errors, especially in complex indoor environments, a scene fingerprint database is used for correction to improve positioning accuracy. This scene fingerprint database stores signal characteristics of various areas of the hospital, including the signal strength distribution and spatial layout of each positioning device. This data helps eliminate positioning errors caused by signal reflection, obstruction, or multipath effects. Specifically, the KNN algorithm is used to calculate feature information in the initial positioning information, such as the signal strength and the Euclidean distance between the signal features of each region in the scene fingerprint database. The K nearest reference points are selected as candidate positioning points, and each reference point is assigned a weight inversely proportional to its distance. The coordinates of these reference points are then weighted and averaged to obtain the actual locations of the K target patients. Next, the locations of the K target patients are matched against the target area scene model of the central positioning cloud node. This scene model contains detailed spatial information about various areas within the hospital, such as the functional areas of each department, waiting areas, operating rooms, wards, the spatial division of areas, equipment distribution, and the status of specific areas. By matching with these scene models, the specific scene and context of the patient's location can be identified, obtaining contextualized positioning information for the K target patients. For example, is the patient in a noisy waiting area or in an examination room with frequent equipment operation? Through these steps, while accurately obtaining the patient's location information, a deeper understanding of the patient's environmental context can be gained, providing accurate basic data for subsequent information delivery and ensuring the relevance and timeliness of information transmission.

[0027] Based on the contextualized location information and location of the K target patients, the K target patients are aggregated into similar categories to obtain M aggregated target patient sets, where M is a positive integer less than or equal to K.

[0028] In one embodiment, after contextualizing the locations of K target patients, the contextualized location information and the locations of the K target patients are used to perform homogeneous aggregation on the K target patients through weighted summation and overall profile coefficient analysis. This identifies patient groups with similar characteristics in space, time, and context, forming M aggregated target patient sets, where M is a positive integer less than or equal to K. These aggregated target patient sets reflect the high correlation between patients in geospatial, contextual characteristics, and behavioral states, providing a grouping basis for subsequent information delivery strategies, thereby improving the accuracy of information dissemination and the overall operational efficiency of the system.

[0029] Furthermore, based on the contextualized location information and locations of the K target patients, the K target patients are aggregated into similar groups to obtain M aggregated target patient sets, including:

[0030] According to the preset aggregation weight, the contextualized location information of the K target patients and the location of the K target patients are weighted and aggregated to obtain M initial aggregated target patient sets; the overall contour coefficient analysis within the set is performed on the M initial aggregated target patient sets to obtain M overall contour coefficients; it is determined whether the M overall contour coefficients meet the preset requirements. If so, the M initial aggregated target patient sets are used as the M aggregated target patient sets.

[0031] Preferably, firstly, based on preset aggregation weights, the contextualized location information and position information of K target patients are weighted and aggregated. That is, the features in the contextualized location information and position information of each target patient are multiplied by their corresponding aggregation weights one by one to obtain a weighted feature vector for each target patient. The preset aggregation weights are set according to the actual needs of the medical scenario. For example, in a waiting area, location and workload may have higher weights, while in an examination area, time window and equipment status may be emphasized. Then, these weighted feature vectors are input into a clustering algorithm, such as K-means, DBSCAN, or a hierarchical aggregation algorithm. Taking K-means as an example, the number M of cluster sets is determined using the elbow method, and M weighted feature vectors of target patients are randomly selected from the weighted feature space as initial cluster centers. The Euclidean distance between each weighted feature vector and each cluster center is then calculated, and the target patient is assigned to the nearest cluster center. Afterward, the mean value of features within each cluster set is calculated as a new cluster center. This process is repeated until the change in cluster centers is below a set threshold or the maximum number of iterations is reached, outputting M initial aggregated target patient sets. After initial aggregation, a silhouette coefficient analysis is performed on each initial aggregated target patient set. The silhouette coefficient is a metric for clustering quality, primarily used to assess the density of samples within the same aggregated set and the separation between different sets. It is calculated by taking the average distance between each target patient and other target patients within its own set, and the average distance to target patients in its nearest neighbor set, and then calculating the difference between these two distances. Then, the overall silhouette coefficients of the M initial aggregated target patient sets are averaged to obtain M overall silhouette coefficients, which are compared with a preset aggregation quality threshold. If all M overall silhouette coefficients meet the aggregation quality threshold, the aggregation quality is considered good, and these M initial aggregated target patient sets are directly used as the final M aggregated target patient sets, providing a stable grouping basis for subsequent strategy generation and targeted information delivery. If one or more overall silhouette coefficients fail to meet the preset requirements, the system will fine-tune the preset aggregation weights, for example, by appropriately increasing the weight of geographical location, decreasing the weight of the noise threshold, or adjusting the time window parameters. After adjustment, the same type of aggregation calculation and silhouette coefficient analysis will be re-executed until the aggregation results meet the expected quality requirements. Through this cyclical mechanism of "aggregation-evaluation-fine-tuning-re-aggregation", high-precision and highly adaptable patient aggregation can be achieved in different in-hospital scenarios, effectively improving the targeting of subsequent information delivery and the reliability of strategy analysis.

[0032] Furthermore, an in-set global profile coefficient analysis is performed on each of the M initial aggregated target patient sets to obtain M global profile coefficients, including:

[0033] Extract the first initial aggregated target patient set from the M initial aggregated target patient sets; calculate the average distance between any initial aggregated target patient in the first initial aggregated target patient set and other initial aggregated target patients in the first initial aggregated target patient set to obtain the average distance set within the first set; calculate the average distance between any initial aggregated target patient in the first initial aggregated target patient set and other initial aggregated target patient sets in the M-1 initial aggregated target patient sets, and extract the minimum value to obtain the minimum average distance set between the first sets; analyze based on the average distance set within the first set and the minimum average distance set between the first sets to obtain the first target patient contour coefficient set; calculate the mean of the first target patient contour coefficient set to obtain the first overall contour coefficient, and add the first overall contour coefficient to the M overall contour coefficients.

[0034] Optionally, to calculate the overall profile coefficient of each aggregated set, firstly, an arbitrary set is extracted from the M initial aggregated target patient sets as the analysis object, and this set is used as the first initial aggregated target patient set for the current analysis. Then, for any target patient within the first initial aggregated target patient set, the weighted feature vector of that target patient is extracted, and the distance between its feature vector and the feature vectors of other target patients in the same set is calculated using Euclidean distance. This average distance reflects the degree of aggregation of the target patient in the current set; the smaller the average distance, the closer the features of the target patient are to those of other patients in the set, and the better the aggregation effect. By summing all the average distances, the set of average distances within the first set can be obtained. Next, for the same target patient, the average distance between it and all patients in each of the remaining M-1 initial aggregated target patient sets is calculated, and the minimum value is extracted as the minimum average distance between sets. This minimum average distance between sets represents the distance between the target patient and the most similar external aggregated set; the larger the minimum average distance between sets, the higher the distinguishability between different aggregated sets. By summing all the minimum average distances between sets, the set of minimum average distances between sets can be obtained. Then, the obtained sets of average distances within the first set and the sets of minimum average distances between the first sets are analyzed. The difference between the minimum average distance between sets and the average distance within a set is calculated, and this difference is divided by the largest of these two values ​​to obtain the profile coefficient for each target patient. A profile coefficient close to 1 indicates that the target patient has a good aggregation effect in this set and is clearly distinguishable from other sets; a profile coefficient close to 0 indicates that the target patient is located in the aggregation boundary region; a profile coefficient less than 0 may indicate that the target patient has been incorrectly aggregated into the current set. The system summarizes the profile coefficients of all target patients to form the first target patient profile coefficient set and calculates its mean to obtain the overall profile coefficient of this aggregation set. Finally, this overall profile coefficient is recorded and added to M overall profile coefficient sets, providing a basis for subsequent self-checking and optimization of aggregation quality, ensuring the scientific nature and stability of patient clustering.

[0035] An interaction relationship graph structure is constructed for the K information delivery tasks by combining the M aggregated target patient sets, and the information delivery strategy is analyzed based on the construction results. Based on the cloud-edge collaborative positioning network, the K information delivery tasks are delivered in a targeted manner according to the information delivery strategy.

[0036] In one embodiment, after generating K information delivery tasks and M aggregated target patient sets, the system maps and matches the M aggregated target patient sets with the corresponding K information delivery tasks, establishing a relationship between patient groups and task content. Each information delivery task is considered an interaction node and bound to its corresponding aggregated target patient set, forming a basic node set. Then, based on these binding relationships, an interaction relationship graph structure is constructed, and the optimal information delivery strategy is generated by parsing the interaction relationship graph. Finally, based on a cloud-edge collaborative positioning network, the system executes targeted information delivery on different terminals according to the parsed information delivery strategy. This ensures accurate information delivery by person, by domain, and on time in complex, multi-terminal, and multi-patient environments within the hospital, improving the quality of information interaction and the patient experience.

[0037] Furthermore, by constructing an interaction graph structure for the K information delivery tasks based on the M aggregated target patient sets, and parsing the information delivery strategy according to the construction results, the cloud-edge collaborative positioning network performs targeted information delivery for the K information delivery tasks according to the information delivery strategy, including:

[0038] Based on the M aggregated target patient sets, the K information delivery tasks are mapped and aggregated to obtain M mapped aggregated information delivery task sets. A first mapped aggregated information delivery task set is extracted from the M mapped aggregated information delivery task sets. Each mapped aggregated information delivery task in the first mapped aggregated information delivery task set is treated as an interaction node, and the interaction node is bound to the corresponding target patient to obtain a first bound interaction node set. Multi-dimensional interaction edges are constructed on the first bound interaction node set to obtain a first interaction association graph structure, and this first interaction association graph structure is added to the M first interaction association graph structures. Based on the M first interaction association graph structures, information delivery strategy parsing is performed to obtain the information delivery strategy, and the cloud-edge collaborative positioning network is used to deliver targeted information to the K information delivery tasks according to the information delivery strategy.

[0039] Preferably, after generating M aggregated target patient sets and K information delivery tasks, each information delivery task is first matched with its corresponding target patient. Then, these information delivery tasks are re-aggregated and categorized according to the aggregated target patient set to which the target patient belongs, forming M mapping aggregated information delivery task sets, making the task distribution more closely match the actual situation within the hospital. Subsequently, any one of these M mapping aggregated information delivery task sets is extracted as a first mapping aggregated information delivery task set. In this first mapping aggregated information delivery task set, each mapping aggregated information delivery task is treated as an independent interaction node. The system binds these interaction nodes one by one with their corresponding target patients, forming a first bound interaction node set. This binding relationship ensures that the correspondence between tasks and patients can be clearly expressed, laying the foundation for the subsequent construction of the relationship graph structure. Subsequently, based on the multi-dimensional interaction edge construction rules, multi-dimensional interaction edges are constructed for the first set of bound interaction nodes. When any two bound interaction nodes meet a set threshold on one or more edge rules, a weighted edge is established between these two nodes, thus forming a first interaction association graph structure that can truly reflect the multi-dimensional relationship between the task and the patient. The system repeats the above process for M sets of mapping aggregated information delivery tasks, ultimately obtaining M interaction association graph structures. Then, the system enters the information delivery strategy parsing stage. In this stage, the system identifies regular nodes and potentially risky nodes by performing proximity recognition analysis on the M interaction association graph structures. The information delivery strategy parser then parses these nodes, calculates and outputs the optimal information delivery strategy. This information delivery strategy not only determines the broadcast order and priority of the tasks but also optimizes the delivery paths of different terminals, achieving optimal allocation of resources and timeliness. Finally, the system relies on the cloud-edge collaborative positioning network to execute strategic targeted information delivery. In this process, the central positioning cloud node is responsible for the overall task distribution and path scheduling, while the edge nodes are responsible for specific broadcasts and feedback collection within their respective areas. Once information is delivered to target devices such as bedside screens, corridor speakers, and mobile terminals, the system monitors the delivery effectiveness and patient interaction in real time. Based on the feedback, it dynamically optimizes strategies, forming a closed-loop process of strategy-execution-feedback-re-optimization. This mechanism enables accurate, efficient, and low-interference information broadcasting and delivery within complex hospital environments.

[0040] Furthermore, a multi-dimensional interaction connection is constructed on the first set of bound interaction nodes to obtain a first interaction association graph structure, including:

[0041] Obtain multi-dimensional interaction connection construction rules, wherein the multi-dimensional interaction connection construction rules include: the target patients of two bound interaction nodes have overlapping or adjacent fences, the intersection of time windows reaches a first set threshold, the waiting load reaches a second set threshold, and the noise feature similarity reaches a third set threshold; when any two first bound interaction nodes in the first bound interaction node set satisfy any one of the multi-dimensional interaction connection construction rules, an interaction connection is established to obtain the first interaction association graph structure.

[0042] Optionally, to construct an interactive association graph structure that accurately reflects the multidimensional relationships between different tasks and patients, predefined multidimensional interactive edge construction rules are first loaded. These rules are preset and dynamically managed by the strategy configuration unit in the central positioning cloud node, and are used to determine whether there is an association between different bound interactive nodes. Specifically, the strategy configuration unit first detects the spatial relationship of the target patients, that is, it determines whether the target patients corresponding to two bound interactive nodes are in overlapping or adjacent geofences, which is calculated using the hospital area model to calculate the spatial distance or boundary intersection between patients. If the geofences of the two patients overlap, or the spatial distance between them is lower than the preset geographical proximity threshold, then the two nodes are considered to have a spatial association. Secondly, each bound interactive node carries task execution time parameters, such as waiting time, examination appointment time, treatment cycle, etc. The strategy configuration unit calculates the overlap ratio of the time windows of the two nodes. When the overlap length of the time windows reaches or exceeds the first set threshold, such as 50%, it is determined that there is a valid association between the two in the time dimension. Third, the waiting load parameters are compared. The waiting load reflects the density of patients or the intensity of people flow in a certain area. The strategy configuration unit collects real-time personnel statistics and queuing information from various departments and waiting areas to calculate the load index of the area corresponding to the two nodes. If the load values ​​of both exceed the second set threshold, it indicates that the environment in which the two patients are located is relatively crowded, and there may be information interference or competition for broadcast priority. Fourth, the strategy configuration unit calculates noise characteristic parameters, such as noise intensity and frequency distribution, through audio data collected by environmental sensors or terminals, and then uses cosine similarity to calculate the noise characteristic similarity of the noise characteristic parameters. When the noise characteristic similarity is greater than the third set threshold, it indicates that the two nodes are in an area with similar acoustic environment, which is suitable for unified delivery strategy or synchronous silent control. When any two nodes in the first set of bound interaction nodes are detected to meet any of the above rule conditions, the edge establishment mechanism will be triggered. A weighted interaction edge will be generated for the two nodes in the graph structure. The weight of the edge is automatically allocated according to the type and number of rules met. For example, the weight of an edge that meets both spatial proximity and time window overlap will be higher than that of an edge that only meets a single condition. Finally, by traversing all node pairs in the first set of bound interaction nodes, the multi-dimensional interaction edge construction of the entire graph is gradually completed, and the first interaction association graph structure is finally obtained. This first interaction association graph structure comprehensively describes the interaction relationships between different task nodes in multiple dimensions such as space, time, load and environment, providing a structured foundation for subsequent information delivery strategy analysis, enabling the system to achieve multi-source association analysis and strategy-based delivery optimization in complex hospital environments.

[0043] Furthermore, based on the M first interaction association graph structures, information delivery strategy parsing is performed to obtain the information delivery strategy, including:

[0044] Traverse each bound interaction node in the M first interaction association graph structures to identify proximity, determine the M risk-bound interaction node sets and the M regular binding interaction node sets; call the information delivery strategy parser to parse the M first interaction association graph structures, the M risk-bound interaction node sets and the M regular binding interaction node sets to obtain the information delivery strategy.

[0045] Optionally, after obtaining M first interaction association graph structures, each interaction association graph structure is traversed to extract all bound interaction nodes. Then, a proximity index is calculated based on the weighted edge information between nodes, and nodes are divided into different categories according to a set proximity threshold, forming M sets of risk-bound interaction nodes and M sets of regular interaction nodes. The risk-bound interaction node sets typically represent high-priority task scenarios, such as emergency patient information push, voice control tasks in high-noise areas, or broadcast suppression requirements in densely populated areas. The regular interaction node sets correspond to tasks with relatively lower priority and can be executed when resources are sufficient or there are no conflicts. Afterwards, an information delivery strategy parser is invoked to uniformly parse the obtained M first interaction association graph structures, M sets of risk-bound interaction nodes, and M sets of regular interaction nodes. This information delivery strategy parser is built based on a deep neural network model to iteratively learn the sample association graphs, sample node sets, and sample information delivery strategies through forward propagation, loss calculation, backpropagation, and parameter optimization. When the system inputs the structural features of M first-level interaction graph structures, M sets of risk-bound interaction nodes, and M sets of regular interaction nodes—such as the number of nodes, edge weights, proximity distribution, and task priority parameters—as input vectors to the information delivery strategy parser, the parser analyzes the dependencies and potential conflict patterns between tasks through a deep feature extraction layer and generates the optimal information delivery strategy at the output layer. Finally, the system applies the generated information delivery strategy to the cloud-edge collaborative positioning network for task scheduling, thereby achieving high-precision, intelligent information delivery control in a multi-terminal, multi-task environment within the hospital.

[0046] Furthermore, by traversing each bound interaction node in the M first interaction association graph structures to perform proximity identification, M sets of risky bound interaction nodes and M sets of normal bound interaction nodes are determined, including:

[0047] The interaction edges of each bound interaction node in the M first interaction association graph structures are traversed and weighted proximity identification is performed to obtain M sets of weighted proximity scores for bound interaction nodes; the weighted proximity scores of bound interaction nodes in the M sets of weighted proximity scores that are greater than or equal to a preset proximity threshold are added to the corresponding M sets of risky bound interaction nodes; the bound interaction nodes in the M first interaction association graph structures other than the M sets of risky bound interaction nodes are added to the M sets of regular bound interaction nodes.

[0048] Optionally, to further refine the strength of interaction relationships between nodes and achieve intelligent classification of risk nodes and regular nodes, the system first traverses all bound interaction nodes in each interaction graph structure. For each bound interaction node, the weight of each interaction edge is extracted. These weights comprehensively reflect the closeness of the relationship between nodes and are mainly composed of multiple parameters such as spatial proximity, time window overlap, waiting load similarity, and noise feature similarity. Subsequently, these parameters are superimposed according to different weight ratios using weighted summation to calculate the weighted proximity between the node and each adjacent node. Then, these proximity values ​​are averaged to obtain the weighted proximity index of the bound interaction node. This weighted proximity index can intuitively reflect the clustering degree and interaction influence of the node in the entire graph structure. The larger the value, the closer its relationship with surrounding nodes, the more frequent the interaction, and the higher the information dissemination potential. After the weighted proximity of all bound interaction nodes is calculated, the system stores the results as M sets of weighted proximity of bound interaction nodes, with each set corresponding to an interaction graph structure. Next, nodes are classified according to a preset proximity threshold. For bound interaction nodes with a weighted proximity value greater than or equal to the preset proximity threshold, they are judged to have high relevance and propagation risk in the task network, and are added to the corresponding M risk-bound interaction node sets. These nodes typically represent high-priority or sensitive tasks, such as emergency information broadcasts, high-load area alerts, or tasks that interact with multiple task nodes. Simultaneously, for the remaining nodes in each interaction graph structure that are not classified into the risk node set, they are automatically assigned to the regular bound interaction node set. These nodes typically correspond to ordinary information tasks with lower priority, flexible execution timing, or low likelihood of causing interference. Through this process, the system achieves automatic hierarchical classification of risk nodes and regular nodes based on traversal analysis and weighted calculation, laying a structured foundation for subsequent information delivery strategy analysis. This ensures timely delivery of key information while reducing redundant propagation and cross-interference, improving the overall accuracy and reliability of information delivery.

[0049] In summary, the embodiments of this application have at least the following technical effects:

[0050] First, the multi-source indoor positioning devices in the target area are uniformly connected to the edge nodes to obtain an edge node set. This edge node set is then connected to a central positioning cloud node to construct a cloud-edge collaborative positioning network. Next, the central positioning cloud node is interacted with to retrieve delivery requests according to preset delivery indicators, resulting in K information delivery tasks, where K is a positive integer. Then, K target patients for the K information delivery tasks are identified, and contextualized positioning is performed on these K patients using the cloud-edge collaborative positioning network to obtain their locations and contextualized positioning information. Then, based on the contextualized positioning information and locations of the K target patients, they are aggregated into M aggregated target patient sets, where M is a positive integer less than or equal to K. Finally, an interaction graph structure is constructed for the K information delivery tasks using the M aggregated target patient sets, and the information delivery strategy is analyzed based on the construction results. The cloud-edge collaborative positioning network then delivers targeted information to the K information delivery tasks according to the information delivery strategy. It solves the technical problems of delivery conflicts and low efficiency caused by rule conflicts and lack of adaptive mechanisms in the existing in-hospital information delivery. It achieves the technical effect of realizing accurate, dynamic and compliant targeted information delivery in medical scenarios based on cloud-edge collaborative architecture, reducing misbroadcast rate and environmental interference, and improving patient access efficiency and medical service quality.

[0051] Example 2, based on the same inventive concept as the contextualized orientation information delivery method based on the real-time location of the patient in the foregoing examples, such as... Figure 2 As shown, this application provides a contextualized orientation information delivery system based on the real-time location of the patient. The system includes:

[0052] Communication connection module 11: Unifies edge node access for the multi-source indoor positioning device set in the target area, obtains an edge node set, and establishes communication connections between the edge node set and the central positioning cloud node to construct a cloud-edge collaborative positioning network; Delivery request retrieval module 12: Interacts with the central positioning cloud node to retrieve delivery requests according to preset delivery indicators, obtaining K information delivery tasks, where K is a positive integer; Contextualized positioning module 13: Obtains K target patients for the K information delivery tasks, and performs contextualized positioning of the K target patients in conjunction with the cloud-edge collaborative positioning network to obtain K target... The system includes: patient location and contextualized location information for K target patients; a homogeneous aggregation module 14: aggregates the K target patients based on their contextualized location information and locations to obtain M aggregated target patient sets, where M is a positive integer less than or equal to K; and a targeted information delivery module 15: constructs an interaction graph structure for the K information delivery tasks based on the M aggregated target patient sets, parses the information delivery strategy based on the construction results, and performs targeted information delivery for the K information delivery tasks based on the cloud-edge collaborative positioning network according to the information delivery strategy.

[0053] Furthermore, the delivery demand retrieval module 12 is used to perform the following method:

[0054] Preset delivery metrics include target patients, geofence, waiting load, time window, and terminal type.

[0055] Furthermore, the contextualized positioning module 13 is used to perform the following method:

[0056] The edge node set of the cloud-edge collaborative positioning network is used to collect and preprocess the terminal signals of the K target patients to obtain K initial positioning information; the K initial positioning information is corrected by combining the scene fingerprint database to determine the location of the K target patients; the location of the K target patients is matched and retrieved with the target area scene model of the central positioning cloud node to obtain the contextualized positioning information of the K target patients.

[0057] Furthermore, the similar aggregation module 14 is used to perform the following method:

[0058] According to the preset aggregation weight, the contextualized location information of the K target patients and the location of the K target patients are weighted and aggregated to obtain M initial aggregated target patient sets; the overall contour coefficient analysis within the set is performed on the M initial aggregated target patient sets to obtain M overall contour coefficients; it is determined whether the M overall contour coefficients meet the preset requirements. If so, the M initial aggregated target patient sets are used as the M aggregated target patient sets.

[0059] Furthermore, the similar aggregation module 14 is used to perform the following method:

[0060] Extract the first initial aggregated target patient set from the M initial aggregated target patient sets; calculate the average distance between any initial aggregated target patient in the first initial aggregated target patient set and other initial aggregated target patients in the first initial aggregated target patient set to obtain the average distance set within the first set; calculate the average distance between any initial aggregated target patient in the first initial aggregated target patient set and other initial aggregated target patient sets in the M-1 initial aggregated target patient sets, and extract the minimum value to obtain the minimum average distance set between the first sets; analyze based on the average distance set within the first set and the minimum average distance set between the first sets to obtain the first target patient contour coefficient set; calculate the mean of the first target patient contour coefficient set to obtain the first overall contour coefficient, and add the first overall contour coefficient to the M overall contour coefficients.

[0061] Furthermore, the targeted information delivery module 15 is used to perform the following method:

[0062] Based on the M aggregated target patient sets, the K information delivery tasks are mapped and aggregated to obtain M mapped aggregated information delivery task sets. A first mapped aggregated information delivery task set is extracted from the M mapped aggregated information delivery task sets. Each mapped aggregated information delivery task in the first mapped aggregated information delivery task set is treated as an interaction node, and the interaction node is bound to the corresponding target patient to obtain a first bound interaction node set. Multi-dimensional interaction edges are constructed on the first bound interaction node set to obtain a first interaction association graph structure, and this first interaction association graph structure is added to the M first interaction association graph structures. Based on the M first interaction association graph structures, information delivery strategy parsing is performed to obtain the information delivery strategy, and the cloud-edge collaborative positioning network is used to deliver targeted information to the K information delivery tasks according to the information delivery strategy.

[0063] Furthermore, the targeted information delivery module 15 is used to perform the following method:

[0064] Obtain multi-dimensional interaction connection construction rules, wherein the multi-dimensional interaction connection construction rules include: the target patients of two bound interaction nodes have overlapping or adjacent fences, the intersection of time windows reaches a first set threshold, the waiting load reaches a second set threshold, and the noise feature similarity reaches a third set threshold; when any two first bound interaction nodes in the first bound interaction node set satisfy any one of the multi-dimensional interaction connection construction rules, an interaction connection is established to obtain the first interaction association graph structure.

[0065] Furthermore, the targeted information delivery module 15 is used to perform the following method:

[0066] Traverse each bound interaction node in the M first interaction association graph structures to identify proximity, determine the M risk-bound interaction node sets and the M regular binding interaction node sets; call the information delivery strategy parser to parse the M first interaction association graph structures, the M risk-bound interaction node sets and the M regular binding interaction node sets to obtain the information delivery strategy.

[0067] Furthermore, the targeted information delivery module 15 is used to perform the following method:

[0068] The interaction edges of each bound interaction node in the M first interaction association graph structures are traversed and weighted proximity identification is performed to obtain M sets of weighted proximity scores for bound interaction nodes; the weighted proximity scores of bound interaction nodes in the M sets of weighted proximity scores that are greater than or equal to a preset proximity threshold are added to the corresponding M sets of risky bound interaction nodes; the bound interaction nodes in the M first interaction association graph structures other than the M sets of risky bound interaction nodes are added to the M sets of regular bound interaction nodes.

[0069] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for contextualized targeted information delivery based on the real-time location of a patient, characterized in that, The method includes: A unified edge node access is performed on the multi-source indoor positioning device set in the target area to obtain an edge node set. The edge node set is then connected to the central positioning cloud node to construct a cloud-edge collaborative positioning network. The central positioning cloud node is interacted with to retrieve delivery requests according to preset delivery indicators and obtain K information delivery tasks, where K is a positive integer; K target patients for K information delivery tasks are obtained, and the K target patients are contextualized and located using the cloud-edge collaborative positioning network to obtain the location of the K target patients and the contextualized positioning information of the K target patients. Based on the contextualized location information and location of the K target patients, the K target patients are aggregated into similar categories to obtain M aggregated target patient sets, where M is a positive integer less than or equal to K; The interaction relationship graph structure of the K information delivery tasks is constructed by combining the M aggregated target patient sets, and the information delivery strategy is analyzed according to the construction results. Based on the cloud-edge collaborative positioning network, the K information delivery tasks are delivered in a targeted manner according to the information delivery strategy. Specifically, an interaction relationship graph structure is constructed for the K information delivery tasks by combining the M aggregated target patient sets, and the information delivery strategy is parsed based on the construction results. Then, based on the cloud-edge collaborative positioning network, targeted information delivery is performed on the K information delivery tasks according to the information delivery strategy, including: Based on the M aggregated target patient sets, the K information delivery tasks are mapped and aggregated to obtain M mapped and aggregated information delivery task sets, and the first mapped and aggregated information delivery task set is extracted from the M mapped and aggregated information delivery task sets. Each mapping aggregation information delivery task in the first mapping aggregation information delivery task set is taken as an interaction node, and the interaction node is bound to the corresponding target patient to obtain the first bound interaction node set; Multi-dimensional interaction connection is constructed on the first set of bound interaction nodes to obtain the first interaction association graph structure, and the first interaction association graph structure is added to M first interaction association graph structures; Based on the M first interactive association graph structures, the information delivery strategy is parsed to obtain the information delivery strategy, and the cloud-edge collaborative positioning network is used to deliver targeted information to the K information delivery tasks according to the information delivery strategy. Specifically, based on the M first interaction association graph structures, information delivery strategy parsing is performed to obtain the information delivery strategy, including: Traverse each bound interaction node in the M first interaction association graph structure to identify proximity, and determine M sets of risky bound interaction nodes and M sets of normal bound interaction nodes. The risky bound interaction node set represents task scenarios with high priority that need to be processed first; the normal bound interaction node set corresponds to tasks with relatively low priority that can be executed when resources are sufficient or there are no conflicts. The information delivery strategy parser is invoked to parse the M first interaction association graph structures, the M risk-bound interaction node sets, and the M regular binding interaction node sets to obtain the information delivery strategy.

2. The contextualized targeted information delivery method based on the real-time location of the patient as described in claim 1, characterized in that, Preset delivery metrics include target patients, geofence, waiting load, time window, and terminal type.

3. The contextualized targeted information delivery method based on the real-time location of the patient as described in claim 1, characterized in that, K target patients for K information delivery tasks are acquired, and the K target patients are contextually located using the cloud-edge collaborative positioning network to obtain the locations of the K target patients and their contextualized location information, including: The edge node set of the cloud-edge collaborative positioning network is used to collect and preprocess the terminal signals of the K target patients to obtain K initial positioning information. The K initial positioning information are corrected by combining the scene fingerprint database to determine the locations of K target patients; The locations of the K target patients are matched and retrieved with the target area scene model of the central positioning cloud node to obtain the contextualized positioning information of the K target patients.

4. The contextualized targeted information delivery method based on the real-time location of the patient as described in claim 1, characterized in that, Based on the contextualized location information and locations of the K target patients, the K target patients are aggregated into similar categories to obtain M aggregated target patient sets, including: According to the preset aggregation weight, the contextualized location information of the K target patients and the location of the K target patients are weighted and aggregated in the same category to obtain M initial aggregated target patient sets; Perform in-set global contour coefficient analysis on the M initial aggregated target patient sets respectively to obtain M global contour coefficients; Determine whether the M overall contour coefficients meet the preset requirements. If so, then use the M initial aggregated target patient sets as the M aggregated target patient sets.

5. The contextualized targeted information delivery method based on the real-time location of the patient as described in claim 4, characterized in that, For each of the M initial aggregated target patient sets, an in-set global profile coefficient analysis is performed to obtain M global profile coefficients, including: Extract the first initial aggregated target patient set from the M initial aggregated target patient sets; Calculate the average distance between any one initial aggregated target patient in the first initial aggregated target patient set and the other initial aggregated target patients in the first initial aggregated target patient set to obtain the average distance set within the first set; Calculate the average distance between any initial aggregated target patient in the first initial aggregated target patient set and the other initial aggregated target patient sets in the M-1 initial aggregated target patient sets, and extract the minimum value to obtain the set of minimum average distances between the first sets; Based on the analysis of the average distance set within the first set and the minimum average distance set between the first sets, the first target patient contour coefficient set is obtained; Calculate the mean of the first target patient contour coefficient set to obtain the first overall contour coefficient, and add the first overall contour coefficient to the M overall contour coefficients.

6. The contextualized targeted information delivery method based on the real-time location of the patient as described in claim 1, characterized in that, Multi-dimensional interaction edges are constructed on the first set of bound interaction nodes to obtain the first interaction association graph structure, including: Obtain multi-dimensional interaction connection construction rules, wherein the multi-dimensional interaction connection construction rules include: the target patients of two bound interaction nodes have overlapping or adjacent fences, the intersection of time windows reaches a first set threshold, the waiting load reaches a second set threshold, and the noise feature similarity reaches a third set threshold. An interaction edge is established when any two first-bound interaction nodes in the first-bound interaction node set satisfy any one of the multi-dimensional interaction edge construction rules, thereby obtaining the first interaction association graph structure.

7. The contextualized targeted information delivery method based on the real-time location of the patient as described in claim 1, characterized in that, Traverse each bound interaction node in the M first interaction association graph structures to perform proximity identification, and determine M sets of risky bound interaction nodes and M sets of normal bound interaction nodes, including: By traversing the interaction edges of each bound interaction node in the M first interaction association graph structures, weighted proximity identification is performed to obtain the weighted proximity set of the M bound interaction nodes; Add the weighted proximity scores of the M bound interaction nodes that are greater than or equal to a preset proximity threshold to the corresponding M risk bound interaction node sets. Add the binding interaction nodes in the M first interaction association graph structures, excluding the M risk binding interaction node sets, to the M regular binding interaction node sets respectively.

8. A contextualized, directional information delivery system based on the real-time location of a patient, characterized in that: For implementing the contextualized orientation information delivery method based on the real-time location of a patient as described in any one of claims 1-7, the system comprises: Communication connection module: Unified access of edge nodes to the set of multi-source indoor positioning devices in the target area to obtain the set of edge nodes, and connect the set of edge nodes to the central positioning cloud node to build a cloud-edge collaborative positioning network; Delivery Request Retrieval Module: Interacts with the central positioning cloud node, retrieves delivery requests according to preset delivery indicators, and obtains K information delivery tasks, where K is a positive integer; Contextualized positioning module: acquires K target patients for K information delivery tasks, and performs contextualized positioning of the K target patients in conjunction with the cloud-edge collaborative positioning network to obtain the location of the K target patients and the contextualized positioning information of the K target patients; Similar aggregation module: Based on the contextualized location information and location of the K target patients, the K target patients are aggregated into similar categories to obtain M aggregated target patient sets, where M is a positive integer less than or equal to K; Targeted information delivery module: Combines the M aggregated target patient sets to construct an interaction relationship graph structure for the K information delivery tasks, and parses the information delivery strategy based on the construction results. Based on the cloud-edge collaborative positioning network, it performs targeted information delivery for the K information delivery tasks according to the information delivery strategy.

Images