Intelligent behavior analysis and operation optimization system based on UWB positioning for hospital

Abstract

The invention relates to the technical field of medical informatization, in particular to a hospital intelligent behavior analysis and operation optimization system based on UWB positioning, which comprises a hierarchical data processing architecture, a dynamic behavior modeling technology, a multi-scene adaptation mechanism and a resource scheduling strategy. The system constructs a high-precision positioning network through the sensing layer, the transmission layer realizes dual-channel data flow design, the analysis layer fuses a time sequence and a spatial topology to generate a dynamic behavior model, and the application layer provides a customized function module. Dynamic behavior modeling is combined with HMM and GNN to capture target behavior characteristics, a multi-scene adaptation mechanism meets differentiation requirements through a rule engine, and a resource scheduling strategy optimizes operation efficiency by utilizing reinforcement learning. The positioning data can be deeply mined, the intelligent management level of a hospital is improved, and efficient technical support is provided for the medical field.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of medical informatization and intelligent management, and specifically relates to a hospital intelligent behavior analysis and operation optimization system based on UWB positioning. BACKGROUND

[0002] With the rapid development of UWB (Ultra-Wideband) positioning technology, its application in the medical field has gradually become a research hotspot. In particular, in the hospital environment, the intelligent behavior analysis and operation optimization system based on UWB positioning can provide important support for the real-time positioning, behavior analysis and resource scheduling of patients, medical staff and medical equipment. However, the existing UWB positioning technology and related systems still have deficiencies in intelligent behavior analysis, multi-scene adaptability and overall operation optimization capability, which limits their wide application in complex medical environments.

[0003] Through retrieval, an indoor positioning method based on UWB is disclosed. This scheme deploys tags and anchor nodes for networking communication, uses a server to uniformly plan anchor node numbers and work processes, and solves the problem of data asynchronization by means of clock synchronization messages, thereby improving the positioning timeliness and accuracy. However, this technical solution mainly focuses on the optimization of positioning algorithms, and lacks depth in the mining and intelligent analysis of positioning data. It has limited support for dynamic monitoring and prediction of behavior patterns of personnel or equipment in the hospital. In addition, this system fails to fully consider the differentiated needs of multiple scenarios in the hospital (such as wards, operating rooms, emergency rooms, etc.), resulting in certain limitations in its applicability and expandability in complex medical environments.

[0004] Through retrieval, a wireless positioning and tracking system based on UWB positioning is also disclosed. This scheme improves the accuracy of the position information of the tracked object by fusing one-dimensional positioning and two-dimensional positioning technologies and combining target floor information. However, this technical solution focuses on improving positioning accuracy and fails to effectively integrate the needs of hospital operation management, such as patient flow path optimization, medical staff workload analysis and medical equipment utilization rate evaluation. In addition, the application of positioning data in this system is limited to single-dimensional tracking, and lacks comprehensive analysis capability for multi-source data, which has limited support for the overall operation efficiency of the hospital.

[0005] The above problems show that the existing UWB positioning-based technical solutions still have limitations in intelligent behavior analysis, multi-scene adaptability, and overall operation optimization capability of the hospital. Therefore, the present application provides a hospital intelligent behavior analysis and operation optimization system based on UWB positioning, which aims to realize dynamic behavior monitoring and prediction of patients, medical staff and medical equipment by deep mining and intelligent analysis of positioning data, combined with the differentiated needs of multiple scenes in the hospital, and to optimize hospital resource scheduling and operation management, so as to meet the urgent needs of modern medical field for efficient and intelligent management system. SUMMARY

[0006] The present application provides a hospital intelligent behavior analysis and operation optimization system based on UWB positioning, which aims to solve the deficiencies of the prior art in intelligent behavior analysis, multi-scene adaptability, and overall operation optimization capability of the hospital. The system solves the technical problems of deep mining and comprehensive analysis of positioning data in complex medical environment by introducing a multi-level data processing architecture and a distributed node cooperation mechanism, combining dynamic behavior modeling and resource scheduling strategy, while meeting the differentiated needs of different scenes in the hospital. The specific implementation scheme of the present application is described in detail below.

[0007] First part: system architecture design The core of the present application is to propose a hierarchical data processing architecture, which consists of a perception layer, a transmission layer, an analysis layer and an application layer. The perception layer includes a plurality of signal acquisition units, each unit is connected with a spatially distributed reference node through an ultra-wideband communication module to form a high-precision positioning network covering the whole hospital. The transmission layer adopts a dual-channel data flow design, the main channel is used for real-time position information transmission, and the auxiliary channel is used for synchronous acquisition of environmental parameters. The analysis layer is deployed with a multi-dimensional data analysis engine, which generates a dynamic behavior model by fusing time series analysis and spatial topological relationship. The application layer provides customized function modules according to the needs of different scenes in the hospital, such as patient flow path planning, medical staff workload assessment and medical equipment utilization rate statistics, etc.

[0008] In the above architecture, the signal acquisition unit of the perception layer is fixed to the ceiling or wall by magnetic attraction type installation method, and the bottom is provided with an angle-adjustable spherical support, which can adjust the signal emission direction according to the actual environment. The main and auxiliary channels of the transmission layer realize data transmission by a mixed mode of optical fiber and wireless, wherein the optical fiber is responsible for long-distance stable transmission, and the wireless link is used for short-distance flexible connection. The data engine of the analysis layer adopts a distributed computing framework, and the sub-modules communicate asynchronously through a message queue, thereby improving the response speed and scalability of the system.

[0009] Second part: dynamic behavior modeling technology To address the shortcomings of existing technologies in dynamic behavior monitoring and prediction, this invention proposes a behavior modeling method based on spatiotemporal feature extraction. This method first acquires the raw location data of the target object through a perception layer and preprocesses the data using a sliding window algorithm to remove noise interference. Then, in the temporal dimension, a Hidden Markov Model (HMM) is used to probabilistically model the motion trajectory of the target object; in the spatial dimension, a Graph Neural Network (GNN) is used to model the interaction relationships between the target object and other objects. Finally, the modeling results from the temporal and spatial dimensions are fused to generate the dynamic behavior pattern of the target object.

[0010] In the above method, the sliding window algorithm dynamically adjusts the window size according to the target object's movement speed to adapt to the data characteristics in different scenarios. The state transition matrix of the Hidden Markov Model is initialized using maximum likelihood estimation and then iteratively optimized using the Expectation-Maximization (EM) algorithm. The input data for the graph neural network is an adjacency matrix, where the elements represent the relative distance and interaction frequency between the target object and other objects. Through this multi-level modeling approach, the system can accurately capture the behavioral characteristics of the target object in complex medical environments.

[0011] Part Three: Multi-Scenario Adaptation Mechanism To address the diverse needs of hospitals across various scenarios, this invention designs a scenario adaptation mechanism based on a rule engine. This mechanism divides different areas of the hospital into several logical partitions by defining a series of scenario adaptation rules, with each partition corresponding to a specific set of functional configurations. For example, in the ward area, the rule engine prioritizes patient safety and privacy protection, restricting entry for unauthorized personnel; in the operating room area, it focuses on optimizing the workflow of medical staff and reducing unnecessary interference; and in the emergency room area, it emphasizes rapid response capabilities to ensure patients receive timely treatment.

[0012] In the above mechanism, the rule engine parses the scene adaptation rules through a condition matching algorithm, and the matching process uses a prefix tree structure to improve efficiency. The boundaries of each logical partition are defined using virtual fence technology, and the shape and range of the virtual fence can be dynamically adjusted according to actual needs. In addition, the rule engine also supports the addition and modification of custom rules. Users can complete the rule configuration through a graphical interface, achieving personalized adaptation without writing code.

[0013] Part Four: Resource Scheduling and Operation Management To improve the overall operational efficiency of hospitals, this invention proposes a resource scheduling strategy based on reinforcement learning. This strategy constructs a multi-agent system comprising patients, medical staff, and medical equipment, and utilizes a deep Q-network (DQN) to solve the resource allocation problem within the system. Specifically, the system first abstracts the resource distribution within the hospital into a state space, with each state corresponding to a resource allocation scheme; then, through the design of a reward function, it guides the agents to find the optimal solution in the state space.

[0014] In the above strategy, the reward function design comprehensively considers multiple factors, including patient waiting time, medical staff workload, and medical equipment utilization. The training process of the deep Q-network employs an experience replay mechanism, storing historical experience data and randomly sampling to avoid data correlation issues during training. Furthermore, the system introduces a global coordinator to monitor the operational status of each agent and intervene in case of conflicts, ensuring the fairness and rationality of resource allocation.

[0015] Part Five: Technical Effects and Beneficial Effects This invention achieves comprehensive coverage and efficient management of complex hospital environments through the organic combination of a hierarchical data processing architecture, dynamic behavior modeling technology, multi-scenario adaptation mechanisms, and resource scheduling strategies. The hierarchical architecture design provides the system with excellent scalability and flexibility, adapting to the needs of hospitals of different sizes. The application of dynamic behavior modeling technology significantly enhances the system's ability to capture the behavioral characteristics of target objects, providing a reliable basis for subsequent analysis and decision-making. The introduction of a multi-scenario adaptation mechanism further enhances the system's applicability in different regions, meeting the diverse management needs of hospitals. Finally, the reinforcement learning-based resource scheduling strategy effectively optimizes the overall operational efficiency of the hospital, reducing resource waste and labor costs.

[0016] In summary, this invention provides an efficient and intelligent management system by deeply mining and intelligently analyzing location data, combined with the differentiated needs of hospitals in various scenarios, thus providing strong technical support for the development of the modern medical field. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the system architecture design of the present invention, showing the data flow relationship between the perception layer, transmission layer, analysis layer and application layer, as well as the distribution of core components within each layer.

[0018] Figure 2 The flowchart for dynamic behavior modeling technology presents the complete processing from the acquisition of raw location data to the fusion of time and spatial dimensions for modeling, and marks the location of key algorithm modules.

[0019] Figure 3 The diagram illustrates the logical partitioning mechanism for multi-scenario adaptation, showing the definition of virtual fences in different areas of the hospital and the matching process of the rule engine for the functional configuration of each logical partition.

[0020] The attached figures are labeled as follows: 1. Perception Layer; 2. Transmission Layer; 3. Analysis Layer; 4. Application Layer; 5. Signal Acquisition Unit; 6. Dynamic Behavior Modeling Module; 7. Rule Engine; 8. Virtual Fence; 9. Resource Scheduling Strategy Module. Detailed Implementation

[0021] This invention provides a hospital-based intelligent behavior analysis and operation optimization system based on UWB positioning, the specific implementation of which is described below. (Combined with...) Figures 1 to 3 The accompanying figures, labeled 1 to 9, provide a detailed description. The perception layer 1 consists of multiple signal acquisition units 5, which are magnetically mounted to the ceiling or walls. Each unit has an adjustable spherical bracket at its base, allowing for adjustment of the signal transmission direction based on the environment. These signal acquisition units 5 connect to spatially distributed reference nodes via an ultra-wideband communication module, forming a high-precision positioning network covering the entire hospital. The transmission layer 2 employs a dual-channel data stream design. The main channel transmits real-time location information, while the auxiliary channel synchronously acquires environmental parameters. Data transmission between the main and auxiliary channels utilizes a hybrid fiber optic and wireless approach, with fiber optics ensuring stable long-distance transmission and wireless links providing flexible short-distance connections. The analysis layer 3 deploys a multi-dimensional data analysis engine that generates dynamic behavior models by fusing time-series analysis and spatial topological relationships. The application layer 4 provides customized functional modules for different hospital scenarios, such as patient flow path planning, healthcare worker workload assessment, and medical equipment utilization statistics.

[0022] Signal acquisition unit 5 interacts with other reference nodes in perception layer 1 via an ultra-wideband communication module, forming a distributed node collaboration mechanism. The position of each signal acquisition unit 5 is precisely arranged to ensure no blind spots in signal coverage and minimize overlapping areas. The spherical bracket of signal acquisition unit 5 is connected to the main body via a threaded structure, allowing manual angle adjustment to adapt to different building structures. The main channel and auxiliary channel in transmission layer 2 are connected to perception layer 1 through independent data interfaces. The main channel receives real-time location data from signal acquisition units 5, while the auxiliary channel collects environmental parameters such as temperature and humidity. The main channel is connected to analysis layer 3 via optical fiber, and the auxiliary channel transmits data to the preprocessing module in analysis layer 3 via a wireless link. The multi-dimensional data analysis engine in analysis layer 3 consists of multiple sub-modules, which communicate asynchronously through message queues, thereby improving system response speed and scalability. The functional modules of application layer 4 interface with the data engine of analysis layer 3 through API interfaces, transforming the analysis results into visualized outputs for user use.

[0023] Dynamic behavior modeling module 6 is located in analysis layer 3, and its core process is as follows: Figure 2 As shown, after the signal acquisition unit 5 acquires the raw position data of the target object, the dynamic behavior modeling module 6 first preprocesses the data using a sliding window algorithm to remove noise interference. The window size of the sliding window algorithm is dynamically adjusted according to the moving speed of the target object to ensure that the data characteristics match the application scenario. Subsequently, in the time dimension, a hidden Markov model is used to probabilistically model the motion trajectory of the target object. The state transition matrix is ​​initialized using the maximum likelihood estimation method and iteratively optimized using the expectation-maximization algorithm. In the spatial dimension, a graph neural network is used to model the interaction relationship between the target object and other objects. The input data is an adjacency matrix, and the elements in the matrix represent the relative distance and interaction frequency between the target object and other objects. The modeling results in the time and spatial dimensions are integrated by the feature fusion module to finally generate the dynamic behavior pattern of the target object.

[0024] Rule Engine 7, located in Application Layer 4, is used to implement a multi-scenario adaptation mechanism. For example... Figure 3As shown, Rule Engine 7 divides different areas of the hospital into several logical partitions by defining a series of scenario adaptation rules. Each partition corresponds to a specific set of functional configurations. The boundaries of the logical partitions are defined by virtual fences 8, the shape and range of which can be dynamically adjusted according to actual needs. Rule Engine 7 parses the scenario adaptation rules using a condition matching algorithm, employing a prefix tree structure to improve efficiency. For example, in the ward area, Rule Engine 7 prioritizes patient safety and privacy protection, restricting unauthorized personnel from entering; in the operating room area, it focuses on optimizing the workflow of medical staff to reduce interference; and in the emergency room area, it emphasizes rapid response capabilities to ensure timely treatment for patients. Rule Engine 7 supports the addition and modification of custom rules, allowing users to configure rules through a graphical interface, achieving personalized adaptation without writing code.

[0025] The resource scheduling strategy module 9, located in application layer 4, implements resource scheduling by constructing a multi-agent system comprising patients, medical staff, and medical equipment. Module 9 utilizes a deep Q-network to solve the resource allocation problem within the system. First, the resource distribution within the hospital is abstracted into a state space, with each state corresponding to a resource allocation scheme. The reward function is designed to comprehensively consider factors such as patient waiting time, medical staff workload, and medical equipment utilization, guiding the agents to find the optimal solution in the state space. The deep Q-network training process employs an experience replay mechanism, storing historical experience data and randomly sampling to avoid data correlation issues during training. Furthermore, module 9 introduces a global coordinator to monitor the operational status of each agent and intervene when conflicts occur, ensuring the fairness and rationality of resource allocation.

[0026] The data flow relationship between the perception layer 1, transmission layer 2, analysis layer 3, and application layer 4 is as follows: Figure 1 As shown, signal acquisition unit 5 sends data to transmission layer 2 via ultra-wideband communication module. Transmission layer 2 then transmits the data to analysis layer 3 via fiber optic and wireless links. The multidimensional data analysis engine in analysis layer 3 processes the received data and sends the results to application layer 4. The dynamic behavior modeling module 6, rule engine 7, and resource scheduling strategy module 9 in application layer 4 implement different functions and are integrated with the hospital management system through API interfaces. The entire system operates efficiently through a distributed computing framework, and data exchange between layers is achieved through standardized protocols, ensuring system stability and reliability.

[0027] To enable those skilled in the art to fully understand and implement this invention, the specific implementation principle of this invention will be further explained below in conjunction with a specific application scenario.

[0028] In a hospital environment, taking the emergency room as an example, the system acquires real-time location information of patients, medical staff, and medical equipment through signal acquisition unit 5 in perception layer 1. Signal acquisition unit 5 establishes a connection with reference nodes via an ultra-wideband communication module, forming a high-precision positioning network. Its spherical support at the bottom can be manually adjusted according to the height and angle of the emergency room ceiling to ensure no blind spots and minimized overlap in signal coverage. The main channel in transmission layer 2 transmits real-time location data of patients and medical staff to analysis layer 3 via optical fiber, while the auxiliary channel simultaneously acquires environmental parameters such as temperature and humidity within the emergency room and sends them wirelessly to the preprocessing module in analysis layer 3. The multi-dimensional data analysis engine in analysis layer 3 processes the received data. First, it uses a sliding window algorithm to remove noise from the raw location data, with the window size dynamically adjusted according to the movement speed of the patient or medical staff. Subsequently, a Hidden Markov Model (HMM) was used to probabilistically model the patient's movement trajectory in the temporal dimension. The state transition matrix was initialized using maximum likelihood estimation and iteratively optimized using the expectation-maximization algorithm. In the spatial dimension, a graph neural network was used to model the interaction relationships between the patient and other objects. The input data was an adjacency matrix, where the elements represented the relative distances and interaction frequencies between the patient and other objects. Finally, the modeling results from the temporal and spatial dimensions were integrated through a feature fusion module to generate the patient's dynamic behavioral pattern.

[0029] Rule Engine 7 defines logical partitions based on the needs of the emergency room scenario. The boundaries of virtual fences 8 are set as areas such as the emergency room entrance, resuscitation area, and waiting area. Rule Engine 7 parses scenario adaptation rules through conditional matching algorithms. For example, in the resuscitation area, priority is given to ensuring smooth workflow for medical staff, restricting unnecessary personnel from entering; in the waiting area, the optimization of patient flow paths is emphasized to avoid congestion. The matching process of Rule Engine 7 uses a prefix tree structure to improve efficiency and allows users to add and modify rules through a graphical interface, completing personalized adaptation without writing code.

[0030] The resource scheduling strategy module 9 implements resource allocation by constructing a multi-agent system involving patients, medical staff, and medical equipment. The system abstracts the resource distribution within the emergency room into a state space, with each state corresponding to a resource allocation scheme. The reward function comprehensively considers factors such as patient waiting time, medical staff workload, and medical equipment utilization, guiding the agents to find the optimal solution in the state space. The training process of the deep Q-network employs an experience replay mechanism, storing historical experience data and randomly sampling to avoid data correlation issues. A global coordinator monitors the operational status of each agent and intervenes when conflicts occur. For example, when multiple patients simultaneously require emergency equipment, the global coordinator allocates equipment according to priority rules, ensuring the fairness and rationality of resource allocation.

[0031] The data flow relationship between the perception layer 1, transmission layer 2, analysis layer 3, and application layer 4 is as follows: Figure 1 As shown, signal acquisition unit 5 sends data to transmission layer 2 via ultra-wideband communication module. Transmission layer 2 then transmits the data to analysis layer 3 via fiber optic and wireless links. The multidimensional data analysis engine in analysis layer 3 processes the received data and sends the results to application layer 4. The dynamic behavior modeling module 6, rule engine 7, and resource scheduling strategy module 9 in application layer 4 implement different functions and are integrated with the hospital management system through API interfaces. The entire system operates efficiently through a distributed computing framework, and data exchange between layers is achieved through standardized protocols, ensuring system stability and reliability.

[0032] In the above scenario, when an emergency patient enters the emergency room, the system captures their location information in real time through the perception layer 1 and transmits the data to the analysis layer 3 through the transmission layer 2. The dynamic behavior modeling module 6 in the analysis layer 3 generates the patient's dynamic behavior pattern and predicts their possible movement path. The rule engine 7 automatically adjusts the boundaries of the virtual fence 8 according to the logical partitioning configuration of the emergency room, ensuring that the patient can quickly enter the resuscitation area. The resource scheduling strategy module 9 prioritizes the allocation of the nearest medical personnel and emergency equipment to the resuscitation area based on the current resource distribution status, reducing patient waiting time. Through these steps, the system achieves comprehensive coverage and efficient management of the complex environment of the emergency room, significantly improving the overall operational efficiency of the hospital.

[0033] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A hospital UWB positioning-based intelligent behavior analysis and operation optimization system, characterized in that It comprises a perception layer (1), a transmission layer (2), an analysis layer (3) and an application layer (4), the perception layer (1) comprises a plurality of signal acquisition units (5), each signal acquisition unit (5) is connected with a spatially distributed reference node through an ultra-wideband communication module; the transmission layer (2) adopts a double-channel data stream design, a main channel is used for real-time position information transmission, and an auxiliary channel is used for synchronous acquisition of environmental parameters; the analysis layer (3) deploys a multi-dimensional data analysis engine, generates a dynamic behavior model by fusing time series analysis and spatial topological relationship; the application layer (4) comprises a dynamic behavior modeling module (6), a rule engine (7) and a resource scheduling strategy module (9). 2.The hospital UWB positioning based intelligent behavior analysis and operation optimization system according to claim 1, characterized in that The dynamic behavior modeling module (6) comprises a sliding window algorithm module, a hidden Markov model module and a graph neural network module, the sliding window algorithm module is used for preprocessing original position data, the hidden Markov model module is used for motion trajectory probability modeling in the time dimension, and the graph neural network module is used for interaction relationship modeling in the space dimension. 3.The hospital UWB positioning-based intelligent behavior analysis and operation optimization system according to claim 1 or 2, characterized in that The rule engine (7) analyzes scene adaptive rules through a conditional matching algorithm, the boundary of a logical partition is defined by a virtual fence (8), the shape and range of the virtual fence (8) are dynamically adjusted according to actual requirements, the rule engine (7) supports addition and modification of custom rules and completes configuration through a graphical interface.

Images