Intelligent ward calling system and method based on edge computing

Abstract

The invention discloses an intelligent ward calling system and method based on edge computing, and belongs to the technical field of intelligent ward calling. According to the invention, a ward calling data set is formed based on patient IDs; when the biological characteristics are abnormal, triggering an emergency call, and when the biological characteristics are in a normal range, judging whether the patient is in an abnormal posture based on the posture change of the patient; when the posture of the patient is abnormal, calling of a corresponding level is triggered, and when the posture of the patient is normal, a wrist micro-motion feature is generated based on the wrist motion; establishing a call content classification model, and when it is detected that the patient has a call demand, determining the call content of the patient; the edge computing gateway receives the call information, and distributes a communication channel for each node according to the communication state and channel occupation condition of each node in the network by adopting a dynamic channel distribution algorithm; calculating position coordinates of the patient, and generating warning information; and the medical care terminal selects an optimal path to receive the calling information of the patient by adopting a dynamic routing algorithm.

Description

Technical Field

[0001] The present invention relates to the field of smart ward call technology, and specifically to a smart ward call system and method based on edge computing. Background Art

[0002] With the rapid development of medical information technology, ward call systems, as an essential component for ensuring patient safety and improving the quality of medical services, play a key role in daily hospital management. Traditional ward call systems rely primarily on wired or simple wireless call buttons. Patients press a button to initiate a call, and medical staff receive call information based on an indicator light or ring tone.

[0003] Traditional ward call systems utilize a centralized data processing model, where patient call information must be transmitted over the network to a central server for processing, which then distributes the information to medical terminals. This model is subject to significant data transmission delays, especially when the hospital network is busy. Information cannot be delivered in a timely manner, resulting in medical staff being unable to quickly respond to patient needs. Relying solely on simple button calls cannot accurately reflect the patient's specific needs and urgency. Furthermore, the existing system lacks an effective anti-false triggering mechanism, making it prone to false calls due to patient misoperation or equipment failure, disrupting normal medical procedures. Traditional call systems are unable to accurately obtain patient location information. After receiving a call, medical staff must spend time searching for the patient's ward. This can further delay treatment, especially in the complex architectural environments of large hospitals. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent ward call system and method based on edge computing to solve the problems raised in the prior art.

[0005] To achieve the above object, the present invention provides the following technical solutions: In a first aspect, the present application provides an edge computing-based smart ward calling method, comprising the following steps: In the wearable terminal, it determines whether the patient has touched the device, collects the patient's biometrics, senses changes in the patient's posture, detects wrist movements, and generates a ward call data set based on the patient's ID. Based on the ward call dataset, when a touch operation is detected, it is determined that the patient has a call request; when the biometric characteristics are abnormal, an emergency call is triggered; when the biometric characteristics are within the normal range, based on the patient's posture changes, whether the patient is in an abnormal posture is determined; when the patient's posture is abnormal, a call of the corresponding level is triggered; when the patient's posture is normal, wrist micro-motion features are generated based on the wrist movements; Establish a call content classification model, perform association annotation, and train the model using historical data. When a patient's call request is detected, the real-time collected data is input into the trained call content classification model, and the model outputs a judgment result to determine the patient's call content. The edge computing gateway receives call information, specifically the ward call data set and call content, and converts it into TCP / IP protocol format. It uses the LoRaWAN protocol to implement low-power wide-area networking and employs a dynamic channel allocation algorithm to allocate communication channels to each node in the network based on the communication status and channel occupancy of each node. The wearable terminal receives call information transmitted by the edge computing gateway. The built-in UWB positioning tag calculates the patient's location coordinates and generates warning information. Medical staff are equipped with medical terminals, which use dynamic routing algorithms to select the optimal path to receive patient call information.

[0006] In combination with the first aspect, in a first implementation of the first aspect of the present application, in the wearable terminal, determining whether the patient has performed a touch operation, collecting the patient's biometrics, sensing the patient's posture changes, detecting wrist movements, and forming a ward call data set based on the patient's ID include: Assign a unique patient ID to each patient, bind this ID to the wearable terminal, and create a corresponding record in the system database, recording the patient's basic information and the relationship between the patient and the wearable terminal; initialize and set the touch matrix, embedded biometric sensors, accelerometers, gyroscopes, and wrist micro-motion detection sensors in the wearable terminal, calibrate sensor parameters, and set the frequency of data collection for each sensor; The touch matrix detects touch signals in real time. When a change in capacitance is detected, it is determined that a touch operation has occurred. The timestamp of the touch is recorded and the touch operation is marked to form touch data containing time information and touch status. The embedded biometric sensor collects the patient's biometrics. Among them, the fingerprint recognition module obtains the patient's fingerprint image data, and the optical heart rate sensor collects the patient's heart rate data. The collected fingerprint image data and heart rate data are recorded together with the collection time to form the patient's biometrics. The accelerometer and gyroscope work together to sense the patient's posture changes in real time and collect the patient's acceleration in three-dimensional space. The system collects acceleration and angular velocity data, and records the collected acceleration components and angular velocity components, as well as the corresponding collection time, to form posture data; the wrist micro-motion detection sensor monitors wrist movements in real time. When it detects a vibration or displacement of the wrist that meets the set threshold, it records the timestamp of the movement, as well as the intensity and duration of the movement, to form wrist micro-motion data; data preprocessing is performed, and the preprocessed touch data, biometrics, posture data, and wrist micro-motion data are integrated using the patient ID as an identifier. The data at the same time point are associated in chronological order to form a ward call data set.

[0007] In combination with the first aspect, in a second implementation of the first aspect of the present application, when the patient's posture is abnormal, triggering a call of a corresponding level, and when the patient's posture is normal, generating wrist micro-movement features based on wrist movements, includes: When the patient's posture is abnormal, a call of the corresponding level is triggered according to the severity; when the patient's posture is normal, the following operations are performed: the characteristics of the wrist movement, including the movement amplitude, frequency, duration and movement pattern, are extracted, and the extracted features are integrated and combined with the acquisition time to generate wrist micro-movement features, which are stored in the field corresponding to the patient ID in the system database.

[0008] In combination with the first aspect, in a third implementation of the first aspect of the present application, establishing a call content classification model, performing association annotation, and training the model using historical data includes: Historical data related to patient calls and corresponding actual call records, including whether the call was triggered and the call type, are extracted from the system database for preprocessing and association annotation. The call content classification model uses a decision tree model, with the input being the labeled historical data and the output being different call content categories. The initial parameters of the model are set, and the labeled historical data is divided into a training set and a test set for model training. The trained model is then optimized.

[0009] In combination with the first aspect, in a fourth implementation of the first aspect of the present application, the edge computing gateway receives call information, specifically a ward call data set and call content, and converts it into a TCP / IP protocol format, including: The edge computing gateway starts the monitoring mode and waits to receive the call information sent by the wearable terminal at the designated communication port. The edge computing gateway starts the data receiving buffer to temporarily store the received data; the edge computing gateway establishes a communication connection with the wearable terminal and confirms the communication protocol and data transmission rules; when the wearable terminal detects the patient's call demand and determines the call content, it packages the ward call data set and call content into a data frame and sends it to the edge computing gateway; the edge computing gateway receives the data frame, stores it in a pre-prepared buffer, and records the timestamp of data reception; the edge computing gateway performs integrity check on the received ward call data set and call content, checks whether the data frame contains all necessary data fields, and when there are missing fields, marks the data as abnormal and attempts to receive it again; For all types of data in the ward call data set, they are converted according to the data format requirements of the TCP / IP protocol; the call content is encoded and converted, and the text-based call content is converted into a character encoding format that complies with the TCP / IP protocol transmission; the converted ward call data set and call content are reorganized according to the data structure of the TCP / IP protocol to form data packets containing header information and data bodies; the converted data packets are further encapsulated, and the header information of the Transmission Control Protocol or User Datagram Protocol is added. According to the actual network transmission requirements and the characteristics of the call information, the appropriate transmission protocol is selected; the edge computing gateway sends the encapsulated data to the hospital internal network through the network interface, and transmits the call information to the target device according to the routing rules of the TCP / IP protocol.

[0010] In combination with the first aspect, in a fifth implementation of the first aspect of the present application, the LoRaWAN protocol is used to implement low-power wide-area networking, and a dynamic channel allocation algorithm is used to allocate a communication channel to each node according to the communication status and channel occupancy of each node in the network, including: LoRaWAN gateways are deployed in the ward area to ensure signal coverage without blind spots. Each LoRaWAN gateway is configured with a unique ID and IP address. The LoRaWAN gateways enable monitoring mode on all available channels, scan and record signal interference in the environment, and perform terminal registration and parameter configuration. The LoRaWAN gateway periodically scans all available channels for RSSI and SNR, recording the idle time ratio, average interference intensity, and packet collision rate of each channel; identifies non-LoRaWAN interference sources through signal feature analysis and marks channels affected by fixed interference; establishes an interference map and records the channel quality distribution in different areas; collects the transmission success rate, ACK response time, and battery power of each terminal and classifies the terminals into different priorities; prioritizes channels with interference below the threshold, calculates the current number of connections and data traffic of each channel, and assigns new terminals to the lightest-loaded channels; reserves dedicated channels for high-priority terminals to ensure priority processing of emergency calls; and sends channel allocation instructions to the terminal via MAC commands to update the terminal's working channel list.

[0011] In combination with the first aspect, in a sixth implementation of the first aspect of the present application, the receiving edge computing gateway transmits the call information, the wearable terminal has a built-in UWB positioning tag, calculates the patient's location coordinates, and generates a warning message, including: Receive call information transmitted by the edge computing gateway; build a UWB positioning network, and the UWB positioning tag built into the wearable terminal interacts with the UWB positioning base station to send and receive UWB signals; after the UWB positioning base station receives the signal sent by the UWB positioning tag, it records the signal arrival time; use the arrival time difference positioning algorithm to calculate the distance between the UWB positioning tag and each base station; based on the distance information between multiple UWB positioning base stations and UWB positioning tags, combined with the known coordinates of the base station in the ward, use the triangulation positioning algorithm to calculate the position coordinates of the wearable terminal in the ward, and combine the position coordinates with the call information to generate a warning message.

[0012] In combination with the first aspect, in a seventh implementation of the first aspect of the present application, the medical staff is equipped with a medical terminal, and the medical terminal uses a dynamic routing algorithm to select an optimal path to receive the patient's call information, including: Each medical staff member is equipped with a medical terminal, which is assigned a unique device identifier and network access credentials. The terminal is registered in the hospital's internal network management system, and a corresponding relationship between the medical staff and the terminal device is established. The medical terminal then accesses the network. The medical terminal starts monitoring mode and waits for patient call information at the designated port. A data receiving buffer is created to temporarily store the received call information. When the edge computing gateway processes the patient's call information and sends it, the medical terminal receives the data frame containing the call information through the network interface and performs a preliminary verification of the data during the receiving process. When the medical terminal receives a transmission request for a call message, it starts a dynamic routing algorithm to select a path and choose the optimal path to receive the patient's call message.

[0013] In combination with the first aspect, in an eighth implementation of the first aspect of the present application, when the medical terminal receives a transmission request for call information, starting a dynamic routing algorithm, performing path selection, and selecting an optimal path to receive the patient's call information includes: Based on the acquired network topology information, all potential paths are sorted out, starting from the edge computing gateway and ending at the medical terminal. For each potential path, the number of network nodes involved, the connection relationship between each node, and whether there is link redundancy or backup links are analyzed. The system takes into account the transmission delay and node processing delay of each link segment and accumulates the total delay value for each path. It queries the current bandwidth usage and total bandwidth capacity of each link in the path to calculate the overall bandwidth utilization of the path. It analyzes the signal strength fluctuations and packet loss rate historical data of each link segment in the path over a period of time, and combines this with the real-time status of the current link to determine the stability of the link. It sets weights for each evaluation indicator, multiplies the evaluation indicator values ​​of each path by the corresponding weights, and then adds them together to obtain a comprehensive score for each path. It sorts the comprehensive scores of all paths and finds the path with the highest comprehensive score as the candidate optimal path. When there are multiple candidate optimal paths with the same score, the optimal path is selected according to the preset priority rules.

[0014] In a second aspect, the present application provides an intelligent ward call system based on edge computing, including: Ward call data set generation module: includes: a data acquisition unit and a data integration unit; wherein the data acquisition unit in the wearable terminal determines whether the patient has touched the device, collects the patient's biometrics, senses the patient's posture changes, and detects wrist movements; the data integration unit generates a ward call data set based on the patient's ID; Wrist micro-motion feature generation module: includes: a call demand determination unit, an abnormal posture determination unit, and a classification operation unit; wherein the call demand determination unit is based on the ward call data set and determines that the patient has a call demand when a touch operation is detected; the abnormal posture determination unit triggers an emergency call when the biometric feature is abnormal, and determines whether the patient is in an abnormal posture based on the patient's posture changes when the biometric feature is within the normal range; the classification operation unit triggers a call of the corresponding level when the patient's posture is abnormal, and generates wrist micro-motion features based on the wrist movements when the patient's posture is normal; The call content confirmation module includes a model building unit and a call content determination unit. The model building unit builds a call content classification model, performs association annotation, and trains the model using historical data. When the call content determination unit detects a patient's call request, it inputs the real-time collected data into the trained call content classification model. The model outputs a judgment result to determine the patient's call content. The communication channel allocation module includes a format conversion unit and a communication channel allocation unit. In the format conversion unit, the edge computing gateway receives call information, specifically the ward call data set and call content, and converts it into the TCP / IP protocol format. The communication channel allocation unit uses the LoRaWAN protocol to implement low-power wide-area networking and adopts a dynamic channel allocation algorithm to allocate a communication channel to each node in the network based on the communication status and channel occupancy of each node. Optimal path selection module: includes: a warning information generation unit and an optimal path selection unit; wherein, the warning information generation unit receives the call information transmitted by the edge computing gateway, and the wearable terminal has a built-in UWB positioning tag to calculate the patient's location coordinates and generate a warning message; in the optimal path selection unit, medical staff are equipped with medical terminals, and the medical terminals use a dynamic routing algorithm to select the optimal path to receive the patient's call information.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention uses edge computing technology to collect and process patient call data in real time at an edge computing gateway close to the data collection end, reducing data transmission delays and achieving rapid response to patient calls. The time from the patient making a call to the medical staff receiving the information is greatly shortened, which can effectively ensure patient safety and is particularly suitable for emergency treatment scenarios.

[0016] 2. The present invention integrates multiple sensors such as a touch matrix, embedded biometric sensors, accelerometers, gyroscopes, and combines them with artificial intelligence algorithms to comprehensively and accurately determine the patient's call needs, urgency, and specific call content. It can not only effectively prevent false triggering, but also intelligently analyze the patient's condition based on multi-dimensional information such as the patient's biometrics and posture changes.

[0017] 3. The present invention adopts UWB precise positioning technology, which can obtain the patient's precise location coordinates in real time. Through the dynamic routing algorithm of the alarm and medical terminal, the call information can be quickly and accurately transmitted to the medical staff, shortening the time for the medical staff to reach the patient. At the same time, the low-power wide-area networking and dynamic channel allocation algorithm based on the LoRaWAN protocol ensure the stability and reliability of data transmission, and realize intelligent and efficient ward management. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a schematic diagram of the steps of an intelligent ward calling method based on edge computing of the present invention; Figure 2 This is a system structure diagram of an intelligent ward call system based on edge computing in the present invention. DETAILED DESCRIPTION

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0020] Example: Figure 1-Figure 2As shown, the present invention provides a technical solution. like Figure 1 As shown in a schematic diagram of a step of an intelligent ward calling method based on edge computing, the present application provides an intelligent ward calling method based on edge computing, comprising the following steps: Step S100: In the wearable terminal, determine whether the patient has performed a touch operation, collect the patient's biometrics, sense the patient's posture changes, detect wrist movements, and form a ward call data set based on the patient's ID; Specifically, each patient is assigned a unique patient ID, which is bound to the wearable terminal. A corresponding record is created in the system database, recording the patient's basic information and the relationship between the patient and the wearable terminal. The touch matrix, embedded biometric sensor, accelerometer, gyroscope, and wrist micro-motion detection sensor in the wearable terminal are initialized and set, the sensor parameters are calibrated, and the frequency of data collection for each sensor is set. The touch matrix detects touch signals in real time. When a change in capacitance is detected, it is determined that a touch operation has occurred. The timestamp of the touch is recorded and the touch operation is marked to form touch data containing time information and touch status. The embedded biometric sensor collects the patient's biometrics. Among them, the fingerprint recognition module obtains the patient's fingerprint image data, and the optical heart rate sensor collects the patient's heart rate data. The collected fingerprint image data and heart rate data are recorded together with the collection time to form the patient's biometrics. The accelerometer and gyroscope work together to sense the patient's posture changes in real time and collect the patient's acceleration in three-dimensional space. The system collects acceleration and angular velocity data, and records the collected acceleration components and angular velocity components, as well as the corresponding collection time, to form posture data; the wrist micro-motion detection sensor monitors wrist movements in real time. When it detects a vibration or displacement of the wrist that meets the set threshold, it records the timestamp of the movement, as well as the intensity and duration of the movement, to form wrist micro-motion data; data preprocessing is performed, and the preprocessed touch data, biometrics, posture data, and wrist micro-motion data are integrated using the patient ID as an identifier. The data at the same time point are associated in chronological order to form a ward call data set.

[0021] Step S200: Based on the ward call data set, when a touch operation is detected, it is determined that the patient has a call request; when the biometric characteristics are abnormal, an emergency call is triggered; when the biometric characteristics are within the normal range, based on the patient's posture changes, whether the patient is in an abnormal posture is determined; when the patient's posture is abnormal, a call of the corresponding level is triggered; when the patient's posture is normal, a wrist micro-motion feature is generated based on the wrist movement; Specifically, when the patient's posture is abnormal, a call of the corresponding level is triggered according to the severity; when the patient's posture is normal, the following operations are performed: the characteristics of the wrist movement are extracted, including the movement amplitude, frequency, duration and movement pattern, the extracted features are integrated, and the collection time is combined to generate wrist micro-movement features, which are stored in the field corresponding to the patient ID in the system database.

[0022] Step S300: Establish a call content classification model, perform association annotation, and train the model using historical data. When a patient's call request is detected, the real-time collected data is input into the trained call content classification model, and the model outputs a judgment result to determine the patient's call content. Specifically, historical data related to patient calls and corresponding actual call records, including whether the call was triggered and the call type, are extracted from the system database for preprocessing and association annotation. The call content classification model uses a decision tree model, with the input being the labeled historical data and the output being different call content categories. The initial parameters of the model are set, and the labeled historical data is divided into a training set and a test set for model training. The trained model is then optimized.

[0023] In one specific example, the experimental scenario was a tertiary hospital inpatient department, comprising five wards and 200 beds. During a one-month data collection period, 10,000 historical records related to patient calls were generated. This historical data was extracted from the hospital's ward management system database. This data included patient biometrics (such as heart rate and blood pressure), posture data (whether a patient had fallen, how long they were still), wrist micro-movement data (such as movement pattern and frequency), and corresponding actual call records. Of these, 3,000 records were triggered calls, while 7,000 were untriggered calls. Of the triggered call records, 800 were emergency calls (e.g., sudden worsening of a patient's condition), 1,500 required medical assistance (e.g., dressing changes, turning patients), and 700 required medication resupply. A labeling team consisting of five experienced medical staff and two data analysts was assembled to annotate the data according to pre-set rules.

[0024] A decision tree model was used, with initial parameters set to a maximum depth of 5 layers, a minimum number of sample splits of 10, and a minimum number of leaf nodes of 5. 8,000 training data points were fed into the decision tree model. The model calculated information gain and selected the optimal feature for splitting at each node. For example, when determining whether a call was an emergency call, biometric data (such as heart rate and blood pressure) with the greatest impact on the classification result was prioritized as the basis for splitting. After 10 training iterations, the model achieved an accuracy of 92% on the training set.

[0025] The trained model was evaluated using a 2,000-item test set, achieving an accuracy of 88%, a recall of 85%, and an F1 score of 86.5%. The decision tree parameters were adjusted, increasing the maximum depth to 6 layers and the minimum number of splits to 8. A data augmentation strategy was added, perturbing some data features (for example, fine-tuning heart rate values ​​within a reasonable range), and expanding the training set to 8,500 items. After retraining the model, the accuracy on the test set increased to 91%, the recall to 88%, and the F1 score to 89.5%.

[0026] Step S400: The edge computing gateway receives call information, specifically the ward call data set and call content, and converts it into TCP / IP protocol format; adopts the LoRaWAN protocol to implement low-power wide-area networking, and adopts a dynamic channel allocation algorithm to allocate a communication channel to each node in the network based on the communication status and channel occupancy of each node; Specifically, the edge computing gateway starts the monitoring mode and waits to receive the call information sent by the wearable terminal at the designated communication port. The edge computing gateway starts the data receiving buffer for temporarily storing the received data; the edge computing gateway establishes a communication connection with the wearable terminal and confirms the communication protocol and data transmission rules; when the wearable terminal detects the patient's call demand and determines the call content, it packages the ward call data set and the call content into a data frame and sends it to the edge computing gateway; the edge computing gateway receives the data frame, stores it in a pre-prepared buffer, and records the timestamp of data reception; the edge computing gateway performs integrity check on the received ward call data set and call content, checks whether the data frame contains all necessary data fields, and when there are missing fields, marks the data as abnormal and attempts to receive it again; For all types of data in the ward call data set, they are converted according to the data format requirements of the TCP / IP protocol; the call content is encoded and converted, and the text-based call content is converted into a character encoding format that complies with the TCP / IP protocol transmission; the converted ward call data set and call content are reorganized according to the data structure of the TCP / IP protocol to form data packets containing header information and data bodies; the converted data packets are further encapsulated, and the header information of the Transmission Control Protocol or User Datagram Protocol is added. According to the actual network transmission requirements and the characteristics of the call information, the appropriate transmission protocol is selected; the edge computing gateway sends the encapsulated data to the hospital internal network through the network interface, and transmits the call information to the target device according to the routing rules of the TCP / IP protocol.

[0027] Furthermore, LoRaWAN gateways were deployed in the ward area to ensure signal coverage without blind spots. Each LoRaWAN gateway was configured with a unique ID and IP address. The LoRaWAN gateways enabled monitoring mode on all available channels, scanning and recording signal interference in the environment, and performing terminal registration and parameter configuration. The LoRaWAN gateway periodically scans all available channels for RSSI and SNR, recording the idle time ratio, average interference intensity, and packet collision rate of each channel; identifies non-LoRaWAN interference sources through signal feature analysis and marks channels affected by fixed interference; establishes an interference map and records the channel quality distribution in different areas; collects the transmission success rate, ACK response time, and battery power of each terminal and classifies the terminals into different priorities; prioritizes channels with interference below the threshold, calculates the current number of connections and data traffic of each channel, and assigns new terminals to the lightest-loaded channels; reserves dedicated channels for high-priority terminals to ensure priority processing of emergency calls; and sends channel allocation instructions to the terminal via MAC commands to update the terminal's working channel list.

[0028] In one specific example, 20 LoRaWAN gateways were deployed in the inpatient area of ​​a hospital. Each gateway was assigned a unique ID (e.g., GW-001 to GW-020) and IP address. Signal strength testing confirmed a 95% overlap in coverage across all gateways, ensuring no signal blind spots. The LoRaWAN gateways activated monitoring mode on all eight available channels (EU863-870 MHz band). Scan results revealed strong WiFi signal interference on channels 3 and 6 during the initial phase of the experiment, with an average interference strength of -75 dBm. The LoRaWAN gateways scanned all eight channels every five minutes for RSSI and SNR. Over the seven days of the experiment, channel status data collected showed that channel 1 had the highest average idle time percentage, reaching 75%, while channel 5 had the lowest average interference strength, at -85 dBm. Signal signature analysis identified 20 non-LoRaWAN interference sources, primarily WiFi devices and some medical equipment within the hospital. An interference map was created, revealing that channel 4 on the third floor of Building B, inpatient building, was the most heavily interfered with, with a packet collision rate as high as 20%.

[0029] Based on each terminal's transmission success rate, ACK response time, and battery level, 600 wearable terminals were divided into three priority levels: emergency call terminals (100), general monitoring terminals (400), and low-frequency data upload terminals (100). During dynamic channel allocation, emergency call terminals were prioritized for channels with interference below -80dBm. Experimental results showed that the data transmission success rate for emergency call terminals increased from an initial 85% to 98%. Using a load balancing strategy, the 50 newly connected terminals were assigned to the lightest-loaded channels 1 and 2, ensuring an even distribution of connections across channels and reducing the proportion of connections in the most heavily loaded channels from 35% to 25%. After reserving dedicated channels for high-priority terminals, in a simulated 100 emergency call scenario, all emergency call messages were transmitted within 2 seconds, a 60% reduction in transmission time compared to when no channels were reserved.

[0030] Step S500: Receive the call information transmitted by the edge computing gateway. The wearable terminal has a built-in UWB positioning tag, calculates the patient's location coordinates, and generates a warning message. Medical staff are equipped with medical terminals, which use dynamic routing algorithms to select the optimal path to receive the patient's call information.

[0031] Specifically, it receives the call information transmitted by the edge computing gateway; builds a UWB positioning network, and the UWB positioning tag built into the wearable terminal interacts with the UWB positioning base station to send and receive UWB signals; after the UWB positioning base station receives the signal sent by the UWB positioning tag, it records the signal arrival time; uses the arrival time difference positioning algorithm to calculate the distance between the UWB positioning tag and each base station; based on the distance information between multiple UWB positioning base stations and the UWB positioning tags, combined with the known coordinates of the base station in the ward, the triangulation positioning algorithm is used to calculate the position coordinates of the wearable terminal in the ward, and the position coordinates are combined with the call information to generate a warning message.

[0032] Furthermore, each medical staff member is equipped with a medical terminal, and a unique device identifier and network access credentials are assigned to the terminal. The terminal is registered in the hospital's internal network management system, and a corresponding relationship between the medical staff and the terminal device is established. The medical terminal then accesses the network. The medical terminal starts monitoring mode and waits to receive patient call information at a designated port. A data receiving buffer is created to temporarily store the received call information. When the edge computing gateway processes the patient's call information and sends it, the medical terminal receives the data frame containing the call information through the network interface and performs a preliminary verification of the data during the receiving process. When the medical terminal receives a transmission request for a call message, it starts a dynamic routing algorithm to select a path and choose the optimal path to receive the patient's call message.

[0033] Furthermore, based on the acquired network topology information, all potential paths are sorted out, starting from the edge computing gateway and ending at the medical terminal itself. For each potential path, the number of network nodes involved, the connection relationship between each node, and whether there is link redundancy or backup link conditions are analyzed. The system takes into account the transmission delay and node processing delay of each link segment and accumulates the total delay value for each path. It queries the current bandwidth usage and total bandwidth capacity of each link in the path to calculate the overall bandwidth utilization of the path. It analyzes the signal strength fluctuations and packet loss rate historical data of each link segment in the path over a period of time, and combines this with the real-time status of the current link to determine the stability of the link. It sets weights for each evaluation indicator, multiplies the evaluation indicator values ​​of each path by the corresponding weights, and then adds them together to obtain a comprehensive score for each path. It sorts the comprehensive scores of all paths and finds the path with the highest comprehensive score as the candidate optimal path. When there are multiple candidate optimal paths with the same score, the optimal path is selected according to the preset priority rules.

[0034] In one specific example, after powering on, the medical terminal automatically connected to the hospital's 5G-WiFi converged network. Network access took an average of 3 seconds, with a 99.5% success rate. Only 15 of 300 terminals failed to connect on the first try; all successfully connected after a second attempt. The medical terminal activated listening mode, waiting for incoming calls on TCP port 9090 and creating a 5MB data receive buffer. During the experiment, the buffer averaged 30% utilization, with no overflows due to excessive data volume.

[0035] Starting with the edge computing gateway and ending at the medical terminal, an average of 12 potential transmission paths were identified each time, with a maximum of 18 and a minimum of 8. An analysis of the number of network nodes involved in each path, connectivity, and link redundancy revealed that 30% of the paths had backup links. Calculation of the total latency for each path showed the shortest latency path was 120ms, the longest was 350ms, and the average latency was 210ms. Overall bandwidth utilization for each path ranged from a low of 20% to a high of 75%, with an average utilization of 45%. Based on historical data on signal strength fluctuations and packet loss rates, the link stability score ranged from a high of 9 out of 10 to a low of 6, with an average score of 7.8. Weights of 0.5, 0.3, and 0.2 were assigned to latency, bandwidth utilization, and link stability, respectively. The calculated overall path scores ranged from a high of 8.9 to a low of 6.2. A path selection test was conducted on 100 call information transmission requests, and 78 of the optimal paths selected had a total latency of less than 150ms, bandwidth utilization exceeding 30%, and a link stability score exceeding 8. During the experiment, network status changes triggered the dynamic routing algorithm to recalculate the path 20 times, with an average recalculation time of 200ms. After switching to the new path, data transmission resumed normal in an average of 350ms, with no information loss due to path switching.

[0036] like Figure 2 As shown in the system structure diagram of an intelligent ward call system based on edge computing, this application provides an intelligent ward call system based on edge computing, including: Ward call data set generation module: includes: a data acquisition unit and a data integration unit; wherein the data acquisition unit in the wearable terminal determines whether the patient has touched the device, collects the patient's biometrics, senses the patient's posture changes, and detects wrist movements; the data integration unit generates a ward call data set based on the patient's ID; Wrist micro-motion feature generation module: includes: a call demand determination unit, an abnormal posture determination unit, and a classification operation unit; wherein the call demand determination unit is based on the ward call data set and determines that the patient has a call demand when a touch operation is detected; the abnormal posture determination unit triggers an emergency call when the biometric feature is abnormal, and determines whether the patient is in an abnormal posture based on the patient's posture changes when the biometric feature is within the normal range; the classification operation unit triggers a call of the corresponding level when the patient's posture is abnormal, and generates wrist micro-motion features based on the wrist movements when the patient's posture is normal; The call content confirmation module includes a model building unit and a call content determination unit. The model building unit builds a call content classification model, performs association annotation, and trains the model using historical data. When the call content determination unit detects a patient's call request, it inputs the real-time collected data into the trained call content classification model. The model outputs a judgment result to determine the patient's call content. The communication channel allocation module includes a format conversion unit and a communication channel allocation unit. In the format conversion unit, the edge computing gateway receives call information, specifically the ward call data set and call content, and converts it into the TCP / IP protocol format. The communication channel allocation unit uses the LoRaWAN protocol to implement low-power wide-area networking and adopts a dynamic channel allocation algorithm to allocate a communication channel to each node in the network based on the communication status and channel occupancy of each node. Optimal path selection module: includes: a warning information generation unit and an optimal path selection unit; wherein, the warning information generation unit receives the call information transmitted by the edge computing gateway, and the wearable terminal has a built-in UWB positioning tag to calculate the patient's location coordinates and generate a warning message; in the optimal path selection unit, medical staff are equipped with medical terminals, and the medical terminals use a dynamic routing algorithm to select the optimal path to receive the patient's call information.

[0037] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A smart ward calling method based on edge computing, characterized in that: The following steps are involved: In the wearable terminal, it determines whether the patient has touched the device, collects the patient's biometrics, senses changes in the patient's posture, detects wrist movements, and generates a ward call data set based on the patient's ID. Based on the ward call dataset, when a touch operation is detected, it is determined that the patient has a call request; when the biometric characteristics are abnormal, an emergency call is triggered; when the biometric characteristics are within the normal range, based on the patient's posture changes, whether the patient is in an abnormal posture is determined; when the patient's posture is abnormal, a call of the corresponding level is triggered; when the patient's posture is normal, wrist micro-motion features are generated based on the wrist movements; Establish a call content classification model, perform association annotation, and train the model using historical data. When a patient's call request is detected, the real-time collected data is input into the trained call content classification model, and the model outputs a judgment result to determine the patient's call content. The edge computing gateway receives the call information, specifically the ward call data set and call content, and converts it into TCP / IP protocol format; The LoRaWAN protocol is used to achieve low-power wide-area networking, and a dynamic channel allocation algorithm is used to allocate communication channels to each node based on the communication status and channel occupancy of each node in the network; The wearable terminal receives call information transmitted by the edge computing gateway. The built-in UWB positioning tag calculates the patient's location coordinates and generates warning information. Medical staff are equipped with medical terminals, which use dynamic routing algorithms to select the optimal path to receive patient call information.

2. The intelligent ward calling method based on edge computing according to claim 1 is characterized in that: The wearable terminal determines whether the patient has performed a touch operation, collects the patient's biometrics, senses the patient's posture changes, detects wrist movements, and forms a ward call data set based on the patient's ID, including: Assign a unique patient ID to each patient, bind this ID to the wearable terminal, and create a corresponding record in the system database, recording the patient's basic information and the relationship between the patient and the wearable terminal; initialize and set the touch matrix, embedded biometric sensors, accelerometers, gyroscopes, and wrist micro-motion detection sensors in the wearable terminal, calibrate sensor parameters, and set the frequency of data collection for each sensor; The touch matrix detects touch signals in real time. When a change in capacitance is detected, it is determined that a touch operation has occurred. The timestamp of the touch is recorded and the touch operation is marked to form touch data containing time information and touch status. The embedded biometric sensor collects the patient's biometrics. Among them, the fingerprint recognition module obtains the patient's fingerprint image data, and the optical heart rate sensor collects the patient's heart rate data. The collected fingerprint image data and heart rate data are recorded together with the collection time to form the patient's biometrics. The accelerometer and gyroscope work together to sense the patient's posture changes in real time and collect the patient's acceleration in three-dimensional space. The system collects acceleration and angular velocity data, and records the collected acceleration components and angular velocity components, as well as the corresponding collection time, to form posture data; the wrist micro-motion detection sensor monitors wrist movements in real time. When it detects a vibration or displacement of the wrist that meets the set threshold, it records the timestamp of the movement, as well as the intensity and duration of the movement, to form wrist micro-motion data; data preprocessing is performed, and the preprocessed touch data, biometrics, posture data, and wrist micro-motion data are integrated using the patient ID as an identifier. The data at the same time point are associated in chronological order to form a ward call data set.

3. The intelligent ward calling method based on edge computing according to claim 1 is characterized in that: When the patient's posture is abnormal, a call of a corresponding level is triggered; when the patient's posture is normal, wrist micro-motion features are generated based on wrist movements, including: When the patient's posture is abnormal, a call of the corresponding level is triggered according to the severity; when the patient's posture is normal, the following operations are performed: the characteristics of the wrist movement, including the movement amplitude, frequency, duration and movement pattern, are extracted, and the extracted features are integrated and combined with the acquisition time to generate wrist micro-movement features, which are stored in the field corresponding to the patient ID in the system database.

4. The intelligent ward calling method based on edge computing according to claim 1 is characterized in that: The process of establishing a call content classification model, performing association annotation, and training the model using historical data includes: Historical data related to patient calls and corresponding actual call records, including whether the call was triggered and the call type, are extracted from the system database for preprocessing and association annotation. The call content classification model uses a decision tree model, with the input being the labeled historical data and the output being different call content categories. The initial parameters of the model are set, and the labeled historical data is divided into a training set and a test set for model training. The trained model is then optimized.

5. The intelligent ward calling method based on edge computing according to claim 1, characterized in that: The edge computing gateway receives call information, specifically the ward call data set and call content, and converts it into TCP / IP protocol format, including: The edge computing gateway starts the monitoring mode and waits to receive the call information sent by the wearable terminal at the designated communication port. The edge computing gateway starts the data receiving buffer to temporarily store the received data; the edge computing gateway establishes a communication connection with the wearable terminal and confirms the communication protocol and data transmission rules; when the wearable terminal detects the patient's call demand and determines the call content, it packages the ward call data set and call content into a data frame and sends it to the edge computing gateway; the edge computing gateway receives the data frame, stores it in a pre-prepared buffer, and records the timestamp of data reception; the edge computing gateway performs integrity check on the received ward call data set and call content, checks whether the data frame contains all necessary data fields, and when there are missing fields, marks the data as abnormal and attempts to receive it again; For all types of data in the ward call data set, they are converted according to the data format requirements of the TCP / IP protocol; the call content is encoded and converted, and the text-based call content is converted into a character encoding format that complies with the TCP / IP protocol transmission; the converted ward call data set and call content are reorganized according to the data structure of the TCP / IP protocol to form data packets containing header information and data bodies; the converted data packets are further encapsulated, and the header information of the Transmission Control Protocol or User Datagram Protocol is added. According to the actual network transmission requirements and the characteristics of the call information, the appropriate transmission protocol is selected; the edge computing gateway sends the encapsulated data to the hospital internal network through the network interface, and transmits the call information to the target device according to the routing rules of the TCP / IP protocol.

6. The intelligent ward calling method based on edge computing according to claim 1 is characterized in that: The LoRaWAN protocol is used to implement low-power wide-area networking, and a dynamic channel allocation algorithm is used to allocate a communication channel to each node based on the communication status and channel occupancy of each node in the network, including: LoRaWAN gateways are deployed in the ward area to ensure signal coverage without blind spots. Each LoRaWAN gateway is configured with a unique ID and IP address. The LoRaWAN gateways enable monitoring mode on all available channels, scan and record signal interference in the environment, and perform terminal registration and parameter configuration. The LoRaWAN gateway periodically scans all available channels for RSSI and SNR, recording the idle time ratio, average interference intensity, and packet collision rate of each channel; identifies non-LoRaWAN interference sources through signal feature analysis and marks channels affected by fixed interference; establishes an interference map and records the channel quality distribution in different areas; collects the transmission success rate, ACK response time, and battery power of each terminal and classifies the terminals into different priorities; prioritizes channels with interference below the threshold, calculates the current number of connections and data traffic of each channel, and assigns new terminals to the lightest-loaded channels; reserves dedicated channels for high-priority terminals to ensure priority processing of emergency calls; and sends channel allocation instructions to the terminal via MAC commands to update the terminal's working channel list.

7. The intelligent ward calling method based on edge computing according to claim 1 is characterized in that: The receiving edge computing gateway transmits the call information, and the wearable terminal has a built-in UWB positioning tag to calculate the patient's location coordinates and generate a warning message, including: Receive call information transmitted by the edge computing gateway; build a UWB positioning network, and the UWB positioning tag built into the wearable terminal interacts with the UWB positioning base station to send and receive UWB signals; after the UWB positioning base station receives the signal sent by the UWB positioning tag, it records the signal arrival time; use the arrival time difference positioning algorithm to calculate the distance between the UWB positioning tag and each base station; based on the distance information between multiple UWB positioning base stations and UWB positioning tags, combined with the known coordinates of the base station in the ward, use the triangulation positioning algorithm to calculate the position coordinates of the wearable terminal in the ward, and combine the position coordinates with the call information to generate a warning message.

8. The intelligent ward calling method based on edge computing according to claim 1, characterized in that: The medical staff is equipped with a medical terminal, which uses a dynamic routing algorithm to select the optimal path to receive the patient's call information, including: Each medical staff member is equipped with a medical terminal, which is assigned a unique device identifier and network access credentials. The terminal is registered in the hospital's internal network management system, and a corresponding relationship between the medical staff and the terminal device is established. The medical terminal then accesses the network. The medical terminal starts monitoring mode and waits for patient call information at the designated port. A data receiving buffer is created to temporarily store the received call information. When the edge computing gateway processes the patient's call information and sends it, the medical terminal receives the data frame containing the call information through the network interface and performs a preliminary verification of the data during the receiving process. When the medical terminal receives a transmission request for a call message, it starts a dynamic routing algorithm to select a path and choose the optimal path to receive the patient's call message.

9. The intelligent ward calling method based on edge computing according to claim 8, characterized in that: When the medical terminal receives a transmission request for call information, it starts a dynamic routing algorithm to select a path and selects the optimal path to receive the patient's call information, including: Based on the acquired network topology information, all potential paths are sorted out, starting from the edge computing gateway and ending at the medical terminal. For each potential path, the number of network nodes involved, the connection relationship between each node, and whether there is link redundancy or backup links are analyzed. The system takes into account the transmission delay and node processing delay of each link segment and accumulates the total delay value for each path. It queries the current bandwidth usage and total bandwidth capacity of each link in the path to calculate the overall bandwidth utilization of the path. It analyzes the signal strength fluctuations and packet loss rate historical data of each link segment in the path over a period of time, and combines this with the real-time status of the current link to determine the stability of the link. It sets weights for each evaluation indicator, multiplies the evaluation indicator values ​​of each path by the corresponding weights, and then adds them together to obtain a comprehensive score for each path. It sorts the comprehensive scores of all paths and finds the path with the highest comprehensive score as the candidate optimal path. When there are multiple candidate optimal paths with the same score, the optimal path is selected according to the preset priority rules.

10. An edge computing-based smart ward call system, using an edge computing-based smart ward call method according to any one of claims 1 to 9, characterized in that: include: Ward call data set generation module: includes: a data acquisition unit and a data integration unit; wherein the data acquisition unit in the wearable terminal determines whether the patient has touched the device, collects the patient's biometrics, senses the patient's posture changes, and detects wrist movements; the data integration unit generates a ward call data set based on the patient's ID; Wrist micro-motion feature generation module: includes: a call demand determination unit, an abnormal posture determination unit, and a classification operation unit; wherein the call demand determination unit is based on the ward call data set and determines that the patient has a call demand when a touch operation is detected; the abnormal posture determination unit triggers an emergency call when the biometric feature is abnormal, and determines whether the patient is in an abnormal posture based on the patient's posture changes when the biometric feature is within the normal range; the classification operation unit triggers a call of the corresponding level when the patient's posture is abnormal, and generates wrist micro-motion features based on the wrist movements when the patient's posture is normal; The call content confirmation module includes a model building unit and a call content determination unit. The model building unit builds a call content classification model, performs association annotation, and trains the model using historical data. When the call content determination unit detects a patient's call request, it inputs the real-time collected data into the trained call content classification model. The model outputs a judgment result to determine the patient's call content. The communication channel allocation module includes a format conversion unit and a communication channel allocation unit. In the format conversion unit, the edge computing gateway receives call information, specifically the ward call data set and call content, and converts it into the TCP / IP protocol format. The communication channel allocation unit uses the LoRaWAN protocol to implement low-power wide-area networking and adopts a dynamic channel allocation algorithm to allocate a communication channel to each node in the network based on the communication status and channel occupancy of each node. Optimal path selection module: includes: a warning information generation unit and an optimal path selection unit; wherein, the warning information generation unit receives the call information transmitted by the edge computing gateway, and the wearable terminal has a built-in UWB positioning tag to calculate the patient's location coordinates and generate a warning message; in the optimal path selection unit, medical staff are equipped with medical terminals, and the medical terminals use a dynamic routing algorithm to select the optimal path to receive the patient's call information.

Images