An epilepsy patient internet knowledge, attitude and practice health management and follow-up system

Abstract

The application relates to the field of internet medical treatment and digital health management, and discloses an internet KAP (knowledge, attitude and practice) health management and follow-up system for epilepsy patients, which comprises a user terminal, a wearable device, a communication gateway and a cloud server. The wearable device calculates a sign energy integral and broadcasts a characteristic inversion signal when the integral exceeds a threshold value; the user terminal activates an abnormality confirmation state machine and a false alarm prevention countdown timer after receiving the signal; the communication gateway detects an uplink bandwidth after confirming the abnormality; when the bandwidth is lower than a transmission lower threshold value, an alarm parameter is serialized and packed into a byte characteristic string and is mapped to a bottom layer control channel for transmission; the cloud server analyzes data and calculates an emergency response priority index, and then executes corresponding hierarchical scheduling instructions. The application avoids the problem that a regular data channel transmission is blocked in a weak network environment, realizes a closed loop of local false alarm prevention interception, network adaptive transmission and cloud dynamic scheduling, and improves the reliability of emergency alarm and the accuracy of rescue.

Description

Technical Field

[0001] This invention relates to the field of internet healthcare and digital health management technology, specifically to an internet-based health management and follow-up system for epilepsy patients. Background Technology

[0002] Epilepsy, as a chronic neurological disease requiring long-term management, necessitates remote monitoring and follow-up outside the hospital to safeguard patient safety. Existing digital medical monitoring systems typically use wearable devices to collect patients' vital signs data, which is then uploaded to a cloud platform for centralized analysis via smart terminals or gateway devices to monitor and alert on sudden abnormal states. This conventional direct-connection architecture, under conditions of good network connectivity and high user cooperation, can accomplish basic health data collection and transmission tasks.

[0003] However, existing systems have significant limitations in practical applications. When wearable devices detect suspected abnormal movements, they often directly trigger an alarm signal to the cloud, lacking a local verification mechanism to prevent false alarms. This simplistic judgment logic can easily misinterpret a patient's everyday vigorous exercise or physical impacts from the device as a disease flare-up. This deficiency not only leads to frequent and invalid alerts being sent to family members or medical staff but also results in the uploading of large amounts of invalid data, increasing unnecessary uplink communication overhead and cloud computing power burden.

[0004] Meanwhile, existing data transmission mechanisms struggle to cope with complex network environments. Epileptic seizures are unpredictable, and when patients are in areas with weak network coverage, relying on regular service channels to transmit full alarm data is highly susceptible to packet delays, timeouts, or loss, resulting in critical alarm information not being delivered in a timely manner. Existing systems lack network condition detection and adaptive degradation mechanisms for weak network environments. Furthermore, existing cloud servers typically only execute fixed notification procedures upon receiving alarms, failing to assess urgency by considering communication latency and spatial distance, and failing to synchronize sudden abnormal alarm data to long-term health baseline models, causing a disconnect between emergency response dispatch and long-term follow-up of chronic diseases. Summary of the Invention

[0005] The technical problem solved by this invention is that existing remote monitoring systems for epilepsy patients are prone to problems such as blocked or lost alarm data transmission in environments with weak network signals. At the same time, existing systems usually lack effective false alarm confirmation mechanisms and network adaptive transmission capabilities on the gateway side, resulting in low reliability of alarms and accuracy of rescue dispatch in emergency situations.

[0006] To address the above problems, the present invention provides the following technical solution:

[0007] This invention provides an internet-based health management and follow-up system for epilepsy patients, comprising user terminals, wearable devices, a communication gateway, and a cloud server that are interconnected.

[0008] The cloud server is used to send management interaction instructions to the user terminal and receive data uploaded by the user terminal and the communication gateway.

[0009] The user terminal is used to receive the management interaction command to generate a prompt event, and collect the touch response features and inertial pose features for the prompt event to calculate the execution confidence value;

[0010] The wearable device is used to collect motion feature data, calculate the integral value of vital signs energy within a time sliding window, and broadcast a feature reversal signal when the integral value of vital signs energy is greater than a preset state threshold.

[0011] The user terminal is also used to activate the local abnormal confirmation state machine after receiving the feature reversal signal, start the anti-false alarm countdown timer, and confirm that the system has entered the abnormal triggering condition when the anti-false alarm countdown timer reaches the end and no interrupt input operation is detected.

[0012] The communication gateway is used to continuously send a series of probe data packets to the cloud server after confirming that the system has entered an abnormal triggering state, and to calculate the uplink bandwidth parameters based on the received confirmation response signal.

[0013] When the uplink bandwidth parameter is greater than or equal to the preset transmission lower limit threshold, a full abnormal alarm message is sent to the cloud server through the regular user plane service channel;

[0014] When the uplink bandwidth parameter is lower than the transmission lower limit threshold, the parameter information is extracted, packaged into a byte feature string, and the byte feature string is mapped to the underlying control channel resources for transmission.

[0015] The cloud server is also used to parse the received full abnormal alarm message or the byte feature string, extract the original trigger timestamp and geographic coordinate parameters, calculate the emergency response priority index in combination with the preset weight coefficient, compare the emergency response priority index with the preset scheduling threshold, and execute the corresponding hierarchical scheduling instruction according to the comparison result.

[0016] Preferably, the user terminal collects touch response features and inertial pose features in response to the prompt event to calculate an execution confidence value, specifically including:

[0017] Record the timestamp of the instruction that generates the prompt event and the timestamp of the wake-up response when the user wake-up operation is detected, and use the difference between the two as the touch response latency data;

[0018] In response to the wake-up operation, a short-time domain observation window is opened, the internal inertial measurement unit is invoked to acquire spatial pose time series data, and pose fluctuation characteristic parameters are extracted.

[0019] The touch response latency data and the pose fluctuation feature parameters are aligned and concatenated to construct a two-dimensional feature vector, which is then input into a pre-deployed confidence verification model to calculate and obtain the execution confidence value.

[0020] If the execution confidence value is greater than the valid confirmation threshold, an interaction log is recorded; if the execution confidence value is less than or equal to the valid confirmation threshold, a secondary confirmation pop-up is generated.

[0021] Preferably, the extraction of pose fluctuation feature parameters specifically includes:

[0022] Discrete triaxial acceleration vectors and triaxial angular velocity vectors are obtained within the short-time domain observation window;

[0023] Calculate the difference between the triaxial acceleration vector and the arithmetic mean acceleration vector at each sampling point, and the difference between the triaxial angular velocity vector and the arithmetic mean angular velocity vector at each sampling point.

[0024] The pose fluctuation characteristic parameters are obtained by performing integral calculations by combining preset acceleration weighting coefficients and angular velocity weighting coefficients.

[0025] Preferably, the integral value of vital sign energy within the calculation time sliding window specifically includes:

[0026] Obtain the triaxial acceleration vector at the k-th sampling moment within the time sliding window;

[0027] The gravity component estimation vector is extracted using a low-pass filtering algorithm;

[0028] The squared L2 of the difference between the triaxial acceleration vector and the gravitational component estimation vector is calculated, multiplied by the time interval between adjacent sampling points, and accumulated within the time sliding window to obtain the integral value of the vital signs energy.

[0029] In a preferred embodiment of the present invention, after the false alarm prevention countdown timer is started, the user terminal further performs the following operations:

[0030] Call the authorized notification interface, vibration interface, sound playback interface, or flashing alert interface to continuously output physical warning signals;

[0031] Record the intervention start timestamp when the false alarm prevention countdown timer is activated;

[0032] If the interrupt input operation for the cancel control is captured before the countdown timer for preventing false alarms reaches its end, the exception confirmation state machine is terminated, the intervention termination timestamp is recorded, and the intervention delay data is calculated and stored in local memory based on the difference between the intervention termination timestamp and the intervention start timestamp.

[0033] Preferably, the step of calculating the uplink bandwidth parameters based on the received acknowledgment signal specifically includes:

[0034] The communication gateway synchronously starts the probe timeout timer, and the probe data packet column consists of a preset number of probe data packets;

[0035] If no acknowledgment signal is received within the set timeout period, the current uplink bandwidth parameter is determined to be zero, the byte feature string is written to the local non-volatile buffer, and an audible and visual or near-field broadcast alarm is triggered.

[0036] If the acknowledgment signal is received, the uplink bandwidth parameter is calculated based on the difference between the first timestamp of the first acknowledgment signal and the second timestamp of the last acknowledgment signal, combined with the number of probe data packets and the fixed byte length of a single probe data packet, and the principle of packet pair dispersion.

[0037] Preferably, the step of extracting parameter information and packaging it into a byte feature string specifically includes:

[0038] The communication gateway invokes the filtering interface to perform pause, rate limiting, or discard processing on the user terminal service data stream forwarded by the gateway.

[0039] Extract the current geographic coordinate parameters, including longitude and latitude values, from the operating system's location service interface;

[0040] The original latitude and longitude values ​​are magnified according to a preset scaling factor and then rounded to obtain the longitude compression value and latitude compression value;

[0041] The device anonymity identifier, abnormal operating condition identifier, trigger timestamp, longitude compression value, latitude compression value, and network degradation identifier are serialized according to bit field layout and packaged to generate the byte feature string.

[0042] In a preferred embodiment of the present invention, the step of calculating the emergency response priority index by combining preset weighting coefficients specifically includes:

[0043] The received byte feature string is deserialized using underlying bitmasks and logical shift operations to extract the original trigger timestamp and the geographic coordinate parameters;

[0044] The geographic coordinate parameters are mapped to a standard Earth reference system, and spatial coordinates are retrieved from a medical point of interest database to calculate the spatial spherical distance from the current alarm location to the target medical rescue institution.

[0045] The communication lag characteristics are calculated based on the receiving system time of the cloud server and the original trigger timestamp, and the estimated rescue time characteristics are calculated based on the spatial spherical distance.

[0046] The emergency response priority index is calculated by multiplying the communication lag feature, the estimated rescue time feature, and the parsed network degradation flag by a preset system normalization weight coefficient and then adding them together.

[0047] Preferably, the step of executing the corresponding hierarchical scheduling instruction based on the comparison result specifically includes:

[0048] When the emergency response priority index is greater than or equal to the high priority dispatch threshold, the voice synthesis system is invoked to initiate a voice call to the preset emergency contact terminal, and an alarm work order containing structured geographic coordinates and abnormal working condition identifiers is pushed to the emergency center dispatch system.

[0049] When the emergency response priority index is less than the high priority scheduling threshold and greater than or equal to the low priority scheduling threshold, an early warning is sent to the family terminal, and the polling cycle parameter for the subsequent status check of the user terminal is modified in the database.

[0050] When the emergency response priority index is less than the low priority scheduling threshold, the parsed feature data is stored in the historical log database and marked as an event to be reviewed.

[0051] Preferably, the cloud server is further used for:

[0052] A long-term operation log is constructed by issuing status verification commands to the user terminal and recording touch response latency data;

[0053] The abnormal alarm events generated by parsing and the intervention delay data uploaded by the user terminal are synchronously written into the long-term operation log to periodically update the state baseline model of the epilepsy patient, and management interaction instructions for the next cycle are generated based on the updated state baseline model.

[0054] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0055] During the anomaly monitoring phase, this invention utilizes wearable devices to calculate the integral value of vital signs energy to initially identify abnormal actions, and triggers the local state machine of the user terminal through near-field communication.

[0056] By utilizing a local countdown mechanism to prevent false alarms, false triggering operations caused by non-abnormal situations can be intercepted before alarm data is uploaded, thereby reducing invalid network transmissions and lowering the computing load on the cloud. During the data transmission phase, this invention incorporates a bandwidth detection mechanism based on packet-pair dispersion principles at the communication gateway side.

[0057] When the network throughput is detected to be below the lower threshold, the communication gateway will suspend or discard unnecessary service data, serialize and compress the core alarm parameters to generate a byte feature string, and map this byte feature string to the underlying control channel for transmission. This degraded transmission mechanism avoids the network congestion problem of conventional data channels and improves the alarm data delivery rate of the system in network-constrained environments.

[0058] During the cloud-based dispatch phase, the server calculates an emergency response priority index based on three parameters: communication lag time, estimated rescue time, and network degradation status, and then implements tiered dispatch measures.

[0059] In addition, the system integrates users' daily operation response records with intervention delay data during the false alarm prevention phase to continuously update the status baseline model, thus achieving an objective combination of emergency rescue intervention and long-term follow-up evaluation.

[0060] This invention provides an internet-based health management and follow-up system for epilepsy patients. It has the following beneficial effects:

[0061] 1. This invention uses wearable devices to calculate vital sign energy integral values ​​to initially identify abnormal actions, and introduces an abnormal confirmation state machine and a false alarm countdown timer on the user terminal side. This mechanism provides users with a local intervention interruption window before alarm data is uploaded, which can effectively intercept false triggers caused by non-disease-related actions, thereby avoiding the push of invalid warnings to family members and medical institutions and reducing unnecessary uplink network communication consumption of the system.

[0062] 2. This invention configures a bandwidth detection and adaptive transmission mechanism at the communication gateway. When the uplink bandwidth is detected to be lower than the transmission lower limit threshold, the gateway extracts the core alarm parameters, packages them into a byte feature string, and maps it to the underlying control channel resources for transmission. This feature enables the system to ensure the delivery of key alarm commands through a degraded channel when facing environments with weak network coverage or congested regular service channels, thereby improving the communication reliability of the system under extreme conditions.

[0063] 3. The cloud server of this invention combines communication lag characteristics, spatial spherical distance, and network degradation indicators to calculate the emergency response priority index to perform hierarchical scheduling, and synchronously updates abnormal alarm events and local intervention delay data to the long-term operation log. This processing method rationally allocates medical rescue resources according to objective calculation indicators, and directly feeds back emergency status data to correct the patient's status baseline model, thus solving the problem of disconnect between emergency medical rescue and long-term chronic disease follow-up. Attached Figure Description

[0064] Figure 1 This is a system architecture topology diagram according to an embodiment of the present invention;

[0065] Figure 2 This is a flowchart of a patient status monitoring and information processing method according to an embodiment of the present invention;

[0066] Figure 3 This is a flowchart of the transient hardware state sampling and confidence verification method according to an embodiment of the present invention;

[0067] Figure 4 This is a flowchart of the edge motion monitoring and false alarm prevention state machine control method according to an embodiment of the present invention;

[0068] Figure 5 This is a flowchart of the link bandwidth detection and asymmetric degradation transmission method according to an embodiment of the present invention;

[0069] Figure 6 This is a flowchart of the cloud-based collaborative parsing and rescue dispatch method according to an embodiment of the present invention;

[0070] Figure 7 This is a sequence diagram of the overall system linkage and interaction in an embodiment of the present invention. Detailed Implementation

[0071] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0072] Example:

[0073] See attached document Figure 1 , Figure 1 This is a system architecture topology diagram according to an embodiment of the present invention.

[0074] This invention provides a patient status monitoring and information processing system based on heterogeneous data fusion, comprising: a user terminal, a wearable device, a cloud server, and a communication gateway.

[0075] User terminals establish local connections with wearable devices through wireless communication protocols.

[0076] The user terminal is equipped with a sensing module to acquire the physical pose characteristics and screen interaction timing characteristics of the terminal device.

[0077] User terminals establish network connections with cloud servers via communication gateways.

[0078] The communication gateway can be a communication proxy module, virtual private network tunnel module, or system-level communication service module deployed within the user terminal, or it can be an independent gateway device that establishes a controlled communication connection with the user terminal. The communication gateway is used to perform transmission scheduling, rate limiting, suspension, or degradation processing on the service data streams forwarded through it.

[0079] Wearable devices are equipped with data acquisition and communication modules to perform motion feature sampling at the edge. A processing module is deployed on the cloud server to aggregate data uploaded from device nodes and perform global processing.

[0080] See attached document Figure 2 , Figure 2 This is a flowchart of a patient status monitoring and information processing method according to an embodiment of the present invention.

[0081] This invention provides a method for patient status monitoring and information processing based on heterogeneous data fusion, comprising the following steps:

[0082] S100: The cloud server sends management interaction commands to the user terminal according to a preset time period.

[0083] The user terminal receives management interaction commands and generates prompt events on the system interface. In response to the wake-up operation for the prompt events, the user terminal triggers the internal hardware state sampling mechanism.

[0084] If in the preset response waiting window No wake-up operation was detected in response to the prompt event. The user terminal marked this management interaction command as an unresponsive event, set the touch response latency data to a preset upper limit value or a null value, and generated a low-confidence interaction record to be stored in the local log storage area.

[0085] The management interaction instructions include at least one of the following: medication reminder instruction, status check instruction, health questionnaire completion instruction, retest reminder instruction, or location confirmation instruction.

[0086] The user terminal collects touch response latency data from the time the prompt event is generated to the time the wake-up operation occurs. Within the time window after entering the foreground running state, it calls the internally configured inertial measurement unit to collect spatial pose timing data.

[0087] The user terminal inputs touch response latency data and spatial pose timing data into the confidence verification model to calculate and obtain the execution confidence value for management interaction commands;

[0088] S200: The wearable device continuously collects edge motion feature data and performs energy integration calculation. When the integrated value is greater than the state threshold, a feature reversal signal is generated and sent to the user terminal.

[0089] The feature reversal signal is a short broadcast payload formed by flipping the abnormal status flag bit in the broadcast data packet of the wearable device from a normal value to an abnormal value. The short broadcast payload includes at least a device anonymity flag, an abnormal status flag bit, an abnormal action type field, a trigger timestamp field, and a verification field.

[0090] The user terminal receives the feature reversal signal and activates the anomaly confirmation state machine locally, controlling the activation of the anti-false alarm countdown timer.

[0091] If the user terminal monitors the screen interface input events during the anti-false alarm countdown timer, and receives an interrupt input operation before the preset time end, the abnormal confirmation state machine process is terminated, the timestamp parameter corresponding to the interrupt input operation is calculated and obtained, and intervention delay data containing the timestamp parameter is generated and stored in the local log storage area.

[0092] If the countdown timer for preventing false alarms reaches the preset end of the duration and no interrupt input operation is detected, the system is confirmed to have entered an abnormal triggering condition.

[0093] S300: After confirming that it has entered the abnormal triggering condition, the communication gateway detects the uplink bandwidth parameters between the user terminal and the cloud server, and performs asymmetric routing scheduling according to the link bandwidth parameters.

[0094] When the link bandwidth parameter is greater than or equal to the transmission lower limit threshold, the communication gateway sends a full abnormal alarm message to the cloud server through the regular user plane service channel. The full abnormal alarm message includes geographic coordinate parameters, trigger time characteristics, abnormal working condition identification parameters, wearable device action characteristic data, user terminal operation log data, and optional short-term multimedia data.

[0095] When the link bandwidth parameter is lower than the transmission lower limit threshold, the communication gateway suspends or limits the multimedia data transmission stream forwarded by the communication gateway, extracts the geographic coordinate parameters, trigger time characteristics and abnormal working condition identification parameters, packages them into a byte feature string, and schedules the low bandwidth short message channel or control plane small data transmission channel to send the byte feature string to the cloud server first.

[0096] S400: The cloud server periodically receives the operation log matrix uploaded by the user terminal, and performs feature discretization processing on the operation log matrix to generate a multi-dimensional observation sequence.

[0097] The operation log matrix includes system interface dwell time parameters, operation function frequency parameters, and intervention time delay data;

[0098] Preferably, the operation log matrix also includes touch response latency data corresponding to management interaction commands, execution confidence values, pose fluctuation characteristic parameters, number of non-response times, number of abnormal trigger times, and intervention latency data generated by the abnormal confirmation state machine. Each row in the operation log matrix corresponds to a statistical period, and each column corresponds to a log feature field.

[0099] The cloud server inputs the multidimensional observation sequence into the hidden Markov model, executes the Viterbi algorithm to decode the multidimensional observation sequence, and calculates and obtains the optimal hidden state estimate representing the current feature stage.

[0100] The hidden states of the Hidden Markov Model (HMM) include at least two of the following: stable response state, slow response state, abnormal interaction state, and high-risk disconnection state. The multidimensional observation sequence is obtained by discretizing the operation log matrix. This discretization includes interval binning or hierarchical encoding of touch response latency, execution confidence value, interface dwell time latency, operation function frequency, and intervention latency data. The initial probability matrix, state transition matrix, and observation emission matrix of the HMM are obtained through statistical training using historical operation log samples.

[0101] Based on the drift parameters of the optimal hidden state estimate relative to the baseline data, the cloud server updates the frequency of management interaction commands sent to the user terminal and the data content scheduling strategy.

[0102] The baseline data consists of the distribution of hidden states, average response latency, average execution confidence, average interface dwell time, and intervention latency of users during the initial stable usage period.

[0103] The drift parameter is used to represent the degree of difference between the current periodic hidden state distribution and the baseline hidden state distribution, or to represent the offset level of the current best hidden state estimate relative to the baseline dominant hidden state.

[0104] When the drift parameter increases, the cloud server increases the frequency of issuing status check management interaction commands and reduces unnecessary multimedia content scheduling; when the drift parameter falls back to the baseline range, the cloud server resumes the normal issuance frequency and normal data content scheduling strategy.

[0105] The following section will provide a detailed explanation of the core processing procedures and data flow mechanisms in the aforementioned systems and methods, taking into account specific technical dimensions and underlying algorithmic logic.

[0106] See attached document Figure 3 , Figure 3This is a flowchart of a transient hardware state sampling and confidence verification method according to an embodiment of the present invention.

[0107] The method provided by this invention includes the following sub-steps regarding the specific processing logic for transient hardware state sampling and confidence verification:

[0108] S101, the cloud server sends management interaction commands in a preset data format to the user terminal through the network communication interface;

[0109] The underlying operating system of the user terminal receives the management interaction command and generates a corresponding prompt event in the system user interface.

[0110] The user terminal invokes the system's built-in clock component to record the precise timestamp of the notification event and defines it as the command issuance timestamp. .

[0111] The user terminal is in a background state, continuously listening for input interruptions from the screen touch sensor.

[0112] When a wake-up operation is detected in response to this notification event, the user terminal records the system timestamp at the time the wake-up operation occurred and defines it as the wake-up response timestamp. .

[0113] The user terminal will wake up the response timestamp. With the command issuance timestamp Perform interpolation calculations to obtain touch response latency data. The specific calculation formula is as follows:

[0114] ;

[0115] If in the preset response waiting window If no wake-up operation is detected in response to the notification event, the user terminal will not execute the wake-up response timestamp. Instead of calculating the difference, this prompt event is recorded as a non-response event, and the touch response delay field is set to a preset upper limit value or a null value.

[0116] The unresponsive event was uploaded to the cloud server as an observation feature in the subsequent operation log matrix.

[0117] S102, in response to the above wake-up operation, the application process of the user terminal is switched from background state to foreground running state;

[0118] At the initial moment of entering the foreground running state, the user terminal opens a set short-term observation window. .

[0119] In this observation window During the duration of the event, the user terminal invokes its internally configured inertial measurement unit.

[0120] The user terminal acquires the three-axis acceleration vector in the discrete time series according to the set sampling frequency. With the three-axis angular velocity vector .

[0121] in This is the sampling sequence number, with a value range of [value range missing]. , For this observation window The total number of sampling points collected internally.

[0122] Get the three-axis acceleration vector With the three-axis angular velocity vector After obtaining the spatial pose temporal data, the user terminal performs feature extraction operations to calculate and obtain pose fluctuation feature parameters. .

[0123] Pose fluctuation characteristic parameters Used to quantitatively characterize the magnitude of physical displacement and rotational changes of a user terminal in space.

[0124] The calculation process is achieved by aggregating the degree to which the data at each sampling point deviates from the mean vector. The specific integral solution formula is as follows:

[0125] ;

[0126] In the above formula, ā represents the observation window. Inside The arithmetic mean acceleration vector of three-axis acceleration vectors;

[0127] For observation window Inside The arithmetic mean angular velocity vector of three-axis angular velocity vectors;

[0128] The preset acceleration weighting coefficient;

[0129] This is the preset angular velocity weighting coefficient.

[0130] For low-pass filtering and noise reduction of the raw data of the inertial measurement unit, those skilled in the art can use standard digital signal processing algorithms. The specific implementation logic is a well-known technology in this field and will not be described in detail here.

[0131] S103, the user terminal will display touch response latency data With pose fluctuation characteristic parameters Align and concatenate the timing sequence and data format to construct a two-dimensional feature vector reflecting the user's physical operation status. Its matrix representation is ;

[0132] The user terminal will construct a two-dimensional feature vector Input into the pre-deployed confidence verification model.

[0133] The confidence verification model contains a lightweight nonlinear mapping function to convert discrete physical feature parameters into probability estimates between zero and one.

[0134] The user terminal performs log-odds calculations using a confidence verification model to obtain the execution confidence value for that management interaction command. The specific mapping operation formula is as follows:

[0135] ;

[0136] In the above formula, The parameter vector is obtained by training the confidence verification model with historical sample data, and is used to measure the weight of the time delay parameter and the pose parameter in the confidence assessment.

[0137] This is the bias constant term of the confidence verification model; Represented by natural constant Exponential operations with base 0.

[0138] When training the confidence verification model, samples in which the user actively responds to the prompt event while in a conscious state and is accompanied by a change in the terminal's posture are marked as valid confirmation samples. Samples of accidental touches in the pocket, accidental touches on the desktop, touches without obvious changes in posture, and unresponsive samples that exceed the preset response waiting window are marked as invalid confirmation samples.

[0139] Before training, the touch response latency data was analyzed. and pose fluctuation characteristic parameters Normalization is performed, followed by training using logistic regression to obtain the parameter vector w and the bias constant term. .

[0140] The effective confirmation threshold is determined by a trade-off between the false positive rate and the false negative rate in the validation samples.

[0141] The user terminal will calculate the obtained execution confidence value. The numerical value is compared with the set valid confirmation threshold.

[0142] If the confidence value is executed If the value exceeds the set valid confirmation threshold, the user terminal determines that the user's execution of the management interaction command is valid and reports the execution record to the cloud server.

[0143] If the confidence value is executed If the value is less than or equal to the valid confirmation threshold, the user terminal determines that the operation is suspected to be an invalid touch and generates a secondary confirmation pop-up on the system interface to block and intervene.

[0144] Through the feature extraction and mathematical mapping process of the underlying hardware data, the system transforms the traditional business logic triggered by a single screen button into a technical verification scheme that relies on the state transition characteristics of the real physical space.

[0145] See attached document Figure 4 , Figure 4 This is a flowchart of an edge motion monitoring and false alarm prevention state machine control method according to an embodiment of the present invention.

[0146] The method provided by this invention includes the following sub-steps in its specific processing logic regarding edge motion monitoring and false alarm prevention state machine control:

[0147] S201, When the system is in the normal operation phase, the wearable device performs high-frequency motion feature sampling on the edge side through its internally configured microelectromechanical system;

[0148] In real-world applications, abnormal situations such as patient falls or sudden limb tremors are often accompanied by drastic changes in spatial acceleration within a short period of time.

[0149] This embodiment aims to filter out interference signals generated by routine, gentle activities by calculating the integral of vital signs at the edge, and to transform abrupt physical shocks into quantifiable numerical characteristics.

[0150] Wearable devices extract triaxial acceleration sequences according to a set edge sampling frequency. .

[0151] To reduce communication power consumption, wearable devices do not directly send the raw sampling sequence to the outside; instead, they perform energy integration calculations within their local microprocessor. The wearable device maintains a sequence of lengths... A time-sliding window is established for each sampling point, and the integral value of vital signs energy within that window is calculated. .

[0152] The specific formula for discrete energy integral calculation is as follows:

[0153] ;

[0154] In the above formula, The first time within the sliding window The three-axis acceleration vector at each sampling time;

[0155] This is the gravity component estimation vector extracted by the wearable device using a low-pass filtering algorithm;

[0156] This represents the time interval between adjacent sampling points.

[0157] For baseline drift calibration and gravity component separation of raw accelerometer data, those skilled in the art can use a combination of standard high-pass and low-pass digital filters. The specific circuit deployment and algorithm implementation are well-known technologies in the field and will not be elaborated here.

[0158] The wearable device will calculate the acquired vital signs energy integral value in real time. State thresholds pre-issued to local registers Perform cyclical comparisons.

[0159] As a preferred implementation, the above-mentioned state threshold Instead of being a fixed constant, the system pre-calibrates and dynamically distributes the data to wearable devices based on the historical behavioral baseline data of the target user in a healthy and stable state, combined with variance statistics.

[0160] Typically, the sliding window length The corresponding time span can be set from 1 to 3 seconds to match the typical physical cycle of a human fall or spasm.

[0161] When the physical signs energy integral value Greater than the set state threshold At that time, the wearable device detected a suspected abnormal movement pattern.

[0162] In a preferred embodiment, the wearable device measures the integral value of vital signs energy. Once the value exceeds the state threshold Eth, at least one auxiliary motion feature is further calculated from the following: peak acceleration, attitude angle change, post-event static duration, or high-frequency tremor energy percentage.

[0163] When the integral value of vital signs energy and at least one auxiliary action feature both meet the corresponding threshold conditions, the wearable device generates a feature inversion signal.

[0164] This reduces the probability of accidental triggering caused by vigorous daily movements such as running, swinging arms, and turning quickly.

[0165] In response to the determination result, the Bluetooth Low Energy communication unit of the wearable device modifies a specific payload identifier bit in its broadcast data packet, generates a feature inversion signal, and broadcasts the feature inversion signal to the outside world through the wireless radio frequency channel.

[0166] The feature reversal signal is a short broadcast payload formed by flipping the abnormal status flag bit in the broadcast data packet of the wearable device from a normal value to an abnormal value. The short broadcast payload includes at least a device anonymity flag, an abnormal status flag bit, an abnormal action type field, a trigger timestamp field, and a verification field.

[0167] S202, The user terminal runs a low-power Bluetooth daemon process in the background of the system. This process continuously scans and parses the broadcast data packets sent by surrounding associated devices. When the user terminal receives and parses the above-mentioned feature reversal signal, the underlying operating system of the user terminal wakes up the upper-layer business application through the inter-process communication mechanism and activates the abnormal confirmation state machine in local memory.

[0168] After the exception confirmation state machine is activated, it triggers a call to the operating system's timer application programming interface, controlling the user terminal to activate a preset duration. False alarm countdown timer.

[0169] It should be noted here that the preset duration The specific value range is usually configured between 10 seconds and 30 seconds. This setting takes into account both the urgency of the system alert response and the physical reaction time required for real users to perform intervention operations.

[0170] At the same time, the user terminal records the precise system time when the false alarm prevention countdown timer starts, defining it as the intervention start timestamp. .

[0171] During the countdown timer for preventing false alarms, the anomaly confirmation state machine calls the user terminal's authorized notification interface, vibration interface, sound playback interface, or flashing alert interface to continuously output physical warning signals including vibration, sound, or flashing bright light to inform the current user that the system is about to trigger the reporting mechanism.

[0172] S203, while controlling the countdown timer for preventing false alarms, the exception confirmation state machine registers authorized input listeners for input events on the screen interface in the main thread, and the user terminal monitors in real time whether specific touch operation feedback occurs on the system interface.

[0173] The authorized input listener is used to listen for cancel input events in the application's exception confirmation interface, system notification operation buttons, floating confirmation controls, or user-authorized auxiliary input channels.

[0174] If the false alarm countdown timer reaches the preset duration... Before the endpoint, the authorized input listener captures an external interrupt input operation targeting the cancel control. The user terminal triggers a call to the state machine reset function, terminating the exception confirmation state machine's workflow and destroying the currently running anti-false alarm countdown timer.

[0175] The user terminal calls the system clock component to obtain the precise system time in which the interrupt input operation occurred, and defines it as the intervention termination timestamp. .

[0176] The user terminal is based on the intervention termination timestamp. With the intervention start time stamp Calculate intervention delay data The specific calculation formula is as follows:

[0177] ;

[0178] The user terminal will calculate the acquired intervention delay data. As a key feature representing a user's conscious state and operational responsiveness, it is stored in the log storage area of ​​local memory for subsequent transmission.

[0179] Conversely, if the false alarm prevention countdown timer continues to run until the preset duration... If the countdown window ends and the user terminal does not detect any valid interrupt input operation through the authorized input listener during the entire countdown window, the exception confirmation state machine executes the state flip instruction.

[0180] The user terminal confirms that the system has officially entered the abnormal triggering state based on the state reversal instruction, thereby cross-verifying the physical action exceeding the limit at the hardware level and the logical non-response at the software level, and completing the local anti-false alarm blocking for false alarm events.

[0181] See attached document Figure 5 , Figure 5 This is a flowchart of a link bandwidth detection and asymmetric degradation transmission method according to an embodiment of the present invention.

[0182] The method provided by this invention includes the following sub-steps regarding the specific processing logic for link bandwidth detection and asymmetric degradation transmission:

[0183] S301, after the above-mentioned false alarm countdown timer finishes running and the system is confirmed to have entered an abnormal triggering state, the communication gateway immediately starts the detection mechanism for the physical network environment.

[0184] Considering that network signal attenuation can easily lead to the failure of sending routine alarm information in remote areas or deep inside buildings, the communication gateway continuously sends a series of data packets consisting of a preset number of probe data packets to the cloud server.

[0185] The communication gateway records the initial timestamp of sending the data packet and waits for the cloud server to return an acknowledgment signal.

[0186] It should be noted that, in order to avoid the system getting stuck in a waiting deadlock due to a complete network outage, the communication gateway starts a probe timeout timer simultaneously while sending data packet sequences.

[0187] If no acknowledgment signal is received within the set timeout period, the communication gateway will directly determine the current uplink bandwidth parameters. If the value is zero, then proceed to execute the subsequent asymmetric routing degradation mechanism;

[0188] when When the number of bytes sent is zero or there are consecutive transmission failures, the communication gateway writes the byte feature string into the local non-volatile buffer, periodically re-probes the link according to the exponential backoff strategy, and triggers the local sound and light, vibration or near-field broadcast alarm of the user terminal or wearable device.

[0189] Once any available uplink is detected to have recovered, the communication gateway will prioritize retransmitting the abnormal alarm data in the buffer.

[0190] To accurately assess the transient throughput capacity of the current network, the calculation process is based on the principle of packet pair dispersion, and the specific bandwidth probing formula is as follows:

[0191] ;

[0192] In the above formula, The total number of probe data packets sent to the communication gateway;

[0193] The fixed byte length for a single probe data packet; For the precise system timestamp of the first probe confirmation response signal received by the communication gateway;

[0194] This is the precise system timestamp when the communication gateway receives the last probe confirmation response signal.

[0195] As a preferred implementation, the number of probe data packets Typically, 5 to 10 probes are configured, and to avoid the probes themselves exacerbating network congestion, the data packet length of a single probe is fixed in bytes. It can be set to a small packet structure of 64 bytes to 128 bytes to ensure detection accuracy while avoiding excessive network detection overhead.

[0196] S302, the communication gateway will calculate the obtained uplink bandwidth parameters. With the pre-configured transmission lower limit threshold Perform numerical comparison;

[0197] Transmission lower limit threshold It is the minimum bandwidth requirement set by the system to ensure that multimedia audio and video streams (such as real-time video streams based on the RTP protocol) can be transmitted without lag. It is usually set according to the base bitrate of the multimedia encoder.

[0198] As a specific engineering implementation example, this transmission lower limit threshold It can be configured between 32kbps and 64kbps.

[0199] When uplink bandwidth parameter Greater than or equal to the lower limit threshold of transmission At that time, the communication gateway sends a full abnormality alarm message to the cloud server through the regular user plane service channel.

[0200] The full-scale anomaly alarm message includes geographic coordinate parameters, trigger time characteristics, abnormal operating condition identification parameters, wearable device action characteristic data, user terminal operation log data, and optional short-term multimedia data.

[0201] The comparison results show the uplink bandwidth parameters. Below the transmission lower limit threshold In this case, it indicates that the current network environment cannot support normal business data reporting.

[0202] The communication gateway immediately triggers an asymmetric routing degradation mechanism in the underlying network protocol stack.

[0203] The communication gateway invokes the network filtering interface to suspend, rate-limit, or drop user terminal service data streams forwarded by the communication gateway, and stops scheduling unnecessary multimedia data packets, thereby releasing congested channel resources.

[0204] The communication gateway does not need to directly modify the network buffers of other unrelated processes in the operating system.

[0205] The communication gateway extracts the current geographic coordinate parameters (including longitude and latitude values) from the operating system's location service interface, extracts the trigger time characteristics when the abnormal trigger condition occurs, and reads the preset abnormal condition identifier parameters.

[0206] The communication gateway serializes the above parameters sequentially according to a compact bit field layout, and packages them into a byte feature string with a length of only tens of bytes.

[0207] In one specific implementation, the byte feature string includes a 1-byte header, a 4-byte device anonymity identifier, a 1-byte abnormal operating condition identifier, a 4-byte or 8-byte trigger timestamp, a 4-byte longitude compression value, a 4-byte latitude compression value, a 1-byte network degradation identifier, and a 2-byte verification field. The longitude and latitude compression values ​​can be obtained by rounding the original longitude and latitude values ​​after amplifying them according to a preset scaling factor.

[0208] This byte feature string abandons the traditional long message structure, removes redundant protocol header fields, and compresses the most critical environmental physical quantities required for alarms with high density.

[0209] S303 After completing the encapsulation of the byte feature string, the communication gateway performs asymmetric routing scheduling for this short data payload. Unlike the default high-bandwidth user plane data bearer channel used by regular business data, the communication gateway maps the byte feature string to the underlying control channel resources for transmission through a specific application programming interface.

[0210] Specifically, when the communication module, user terminal operating system, and network side support control plane small data transmission, the communication gateway embeds the byte feature string into the additional information unit of the network signaling message, or calls the control plane optimized transmission channel of cellular IoT for forwarding.

[0211] In cases where control plane small data transmission is not supported, the communication gateway sends the byte feature string through a short message channel, cellular IoT small data packet channel, low bit rate user plane short message channel, MQTT acknowledgment short message channel, UDP short message channel, LoRa gateway channel, or near-field relay channel.

[0212] These types of lower-level control channels typically have high anti-interference capabilities and transmission priority, which can increase the penetration probability of small data packets under weak network coverage.

[0213] After receiving the control plane message, the signaling gateway of the cloud server parses it and reverse-engineers the geographic coordinate parameters, trigger time characteristics, and abnormal working condition identification parameters. This increases the probability of successfully uploading the patient's abnormal alarm signal and its physical location in network deterioration scenarios, and triggers a global rescue response after successful uploading.

[0214] When the communication gateway detects that the regular user plane channel, the control plane small data transmission channel, and other low-bandwidth short message channels are all unavailable, the communication gateway writes the abnormal alarm data to the local non-volatile buffer and triggers a local audible and visual alarm, a wearable device vibration alarm, or a near-field broadcast alarm.

[0215] Near-field broadcast alarms can be sent to authorized caregiver terminals, relay terminals, or home gateways in the vicinity via Bluetooth Low Energy broadcast, Wi-Fi Direct broadcast, or LAN multicast. Once any communication link is restored, the communication gateway will prioritize retransmitting any abnormal alarm data that was not successfully uploaded from its buffer.

[0216] See attached document Figure 6 , Figure 6 This is a flowchart of a cloud-based collaborative analysis and rescue dispatch method according to an embodiment of the present invention.

[0217] The method provided by this invention specifically includes the following sub-steps regarding the data parsing and rescue dispatch processing logic on the cloud side:

[0218] S401, the cloud server's network access layer continuously monitors the underlying control plane channel and the regular user plane service channel;

[0219] When the cloud server receives an additional information unit sent by the communication gateway through the underlying control channel resources, it triggers the message parsing protocol stack.

[0220] The cloud server performs reverse deserialization processing on the received byte feature string by using the underlying bit mask and logical shift operations, based on the pre-issued compact bit field layout rules.

[0221] Through the above deserialization process, the cloud server extracts the original trigger timestamp when the abnormal trigger condition occurred, the geographic coordinate parameters containing longitude and latitude values, and the abnormal condition identification parameters from the byte feature string of dozens of bytes.

[0222] To address the potential differences in positioning coordinate systems among different device terminals, the cloud server uses a built-in coordinate transformation interface to uniformly map or transform the extracted geographic coordinate parameters to the WGS-84 standard Earth reference system, thereby improving the accuracy of subsequent spatial distance calculations.

[0223] S402, after extracting the core physical quantities, the cloud server calls the computing resource pool to perform an emergency response priority assessment in the spatiotemporal dimension;

[0224] Based on the extracted geographic coordinate parameters, the cloud server performs a spatial coordinate search in the locally maintained medical point of interest database to determine the medical rescue organization closest to the current alarm location.

[0225] The medical point of interest database includes at least the medical rescue organization identifier, the organization's geographical coordinates, the organization's service radius, the communication interface address, and the current availability status field.

[0226] The cloud server filters out target medical rescue organizations based on the spatial spherical distance between the alarm location and the candidate medical rescue organizations, combined with the organization's availability status field.

[0227] The cloud server calculates the emergency response priority index based on the extracted feature parameters.

[0228] The core principle of this assessment calculation is to reduce the time overhead that represents communication lag, the space overhead that represents the difficulty of physical rescue, and the network degradation state that represents the degree of environmental deterioration and fuse them in order to objectively quantify the overall urgency of rescue.

[0229] The specific evaluation calculation formula is as follows:

[0230] ;

[0231] In the above formula, The precise system time at which the alarm signature string was received by the cloud server; This is the original trigger timestamp parsed from the byte feature string; The calculated spherical distance from the current geographic coordinates to the nearest medical aid facility; The preset reference speed for emergency vehicles is set to a specific value that is strictly greater than zero in order to avoid system malfunctions caused by a zero denominator in the calculation. In actual applications, the value is usually taken as 30km / h to 50km / h depending on local road conditions.

[0232] This is a network degradation flag. When alarm data is transmitted through the control plane channel called by the asymmetric degradation mechanism, this value is 1; when transmitted through the regular user plane channel, this value is 0.

[0233] α, β, and γ are all preset system normalization weight coefficients.

[0234] To eliminate the dimensional differences between different physical quantities, in a preferred embodiment, the cloud server separately calculates the communication lag time. and estimated rescue time

[0235] Perform normalization processing to obtain normalized communication lag characteristics. and normalized rescue time characteristics and based on , and network degradation flag Calculate the emergency response priority index. Normalization can be achieved using a linear normalization function, a piecewise normalization function, or a lookup table mapping function.

[0236] It is understandable that the specific values ​​of the above normalized weight coefficients can be predetermined based on the analytic hierarchy process.

[0237] Typically, network degradation in remote areas means a very high risk of users going offline; therefore, the weight of network degradation factors is significant. The value should be significantly greater than and This ensures that alarms in weak network environments are responded to first.

[0238] The spherical distance between two geographical coordinates The calculation of the semi-sine can be solved by those skilled in the art using the standard semi-sine formula. The specific mathematical derivation and code implementation are well-known technologies in this field and will not be elaborated here.

[0239] S403, the cloud server will calculate the emergency response priority index. Compare the values ​​with the scheduling thresholds configured in the database;

[0240] The scheduling threshold mapping matches different levels of intervention strategies.

[0241] As an example of engineering implementation, the system can be configured with a high-priority scheduling threshold of 80 and a low-priority scheduling threshold of 60.

[0242] When the emergency response priority index When the alarm value is greater than or equal to the set high-priority scheduling threshold, it indicates that the alarm user may be in a network coverage blind spot, far from the rescue point, or that the alarm data has experienced a long transmission delay in the network, posing a high threat to their life.

[0243] The cloud server calls the emergency communication application programming interface to initiate a voice call to the preset emergency contact terminal through the automatic speech synthesis system, and simultaneously uses the secure hypertext transfer protocol to push an alarm work order containing structured geographic coordinates and abnormal working condition indicators to the emergency center dispatch system.

[0244] The alarm work order includes user anonymity identifier, device anonymity identifier, abnormal working condition identifier, trigger timestamp, alarm reception time, geographic coordinate parameters, target medical rescue institution identifier, emergency response priority index, network degradation identifier, callback contact information, and event verification fields.

[0245] When the emergency response priority index When the priority is less than the high-priority scheduling threshold but greater than or equal to the low-priority scheduling threshold, the cloud server calls the message push service to send an early warning to the preset family member's terminal in the form of a short message service or mobile application push notification. It also establishes a high-frequency tracking file for the user terminal in the cloud database and instructs the system to shorten the polling cycle of subsequent status verification in order to continuously track the vital signs of the person who triggered the alarm.

[0246] Conversely, if the emergency response priority index If the data is less than the set low-priority scheduling threshold, the cloud server will only store the parsed feature data as a regular health event in the local historical log database and update the target user's status baseline model accordingly, without triggering direct external physical intervention.

[0247] Meanwhile, the cloud server marks the event as a low-priority event awaiting review and can issue a status verification management interaction command to the user terminal in the next management interaction cycle to confirm whether the low-priority event needs to be escalated.

[0248] It should be noted that the periodic state phase identification process in S400 is similar to the attached... Figure 6 The cloud-based collaborative analysis and rescue dispatch process shown can be executed in parallel.

[0249] The S400 is used to update management interaction policies based on long-term user operation logs. Figure 6 The process shown is used to perform immediate alarm analysis and rescue dispatch after the wearable device triggers an abnormal action and is confirmed by the user terminal.

[0250] Cloud servers can attach Figure 6 The abnormal alarm events generated by the process shown are synchronously written into the operation log matrix used by S400, so that long-term status assessment can absorb information such as the number of abnormal triggers, intervention delays and unresponsive events, thereby forming a closed loop of periodic management and real-time alarms.

[0251] Thus, this embodiment forms a closed-loop information processing flow between the user terminal, wearable device, communication gateway and cloud server.

[0252] The user terminal is responsible for receiving management interaction commands, collecting touch response latency and spatial pose timing data, and judging the validity of the interaction behavior through a confidence verification model;

[0253] Wearable devices are responsible for collecting motion features at the edge and generating feature inversion signals when suspected abnormal movements occur;

[0254] After receiving the feature reversal signal, the user terminal starts the anomaly confirmation state machine and completes the local false alarm verification through the anti-false alarm countdown timer and interrupt input operation.

[0255] The communication gateway selects full abnormal alarm message transmission, short byte feature string transmission, control plane small data transmission, or local buffer retransmission based on the uplink bandwidth status.

[0256] The cloud server performs hidden status identification of long-term operation logs and rescue dispatch for immediate abnormal alarms.

[0257] Through the above process, even in scenarios such as accidental user touch, false action alarms, or weak network conditions, the system can still maintain continuity between status recognition, alarm confirmation, data degradation transmission, and cloud scheduling.

[0258] See attached document Figure 7 , Figure 7This is a system overall linkage and interaction timing diagram according to an embodiment of the present invention. To further illustrate the technical solution and actual technical effects of the present invention, the multi-terminal linkage process of the above-mentioned core steps is described below in conjunction with specific application scenarios.

[0259] The target user is defined as a patient with a specific disease who wears a smartwatch (corresponding to the aforementioned user terminal). The patient's activity area is the basement of a building in a remote mountainous area. This space is a typical physically shielded area and a weak network coverage area, where the signal attenuation of regular business data channels is significant, making it difficult to support continuous audio and video data stream transmission.

[0260] When the patient fell in the basement, the microelectromechanical system inside the smartwatch he was wearing on his wrist continuously performed high-frequency sampling.

[0261] The smartwatch calculates the vital signs energy integral value in real time within a sliding time window on the edge side.

[0262] Because the fall caused a sudden change in spatial acceleration, the energy integral value within that time window exceeded the state threshold set by the system based on the patient's daily baseline of gentle activity.

[0263] Based on this, the smartwatch determines that it has detected a suspected abnormal behavior pattern, and then modifies a specific load flag in its low-power Bluetooth broadcast data packet to continuously broadcast a feature reversal signal to the home smart gateway (corresponding to the aforementioned communication gateway) deployed indoors.

[0264] After the background daemon of the home smart gateway scans and parses the feature reversal signal, it activates the abnormal confirmation state machine locally and simultaneously starts a countdown timer with a preset duration of 20 seconds to prevent false alarms.

[0265] During the countdown, the home smart gateway and smartwatch synchronously output physical warning signals through a vibration motor and speaker.

[0266] During the entire 20-second countdown window, the countdown timer reached its end because the system did not capture any interrupt input to the cancel control. The state machine inside the home smart gateway executed a state transition instruction, confirming that the system had entered an abnormal trigger condition.

[0267] Once the system enters an abnormal triggering state, the home smart gateway immediately sends a data packet column containing 8 probe data packets to the cloud server to detect the actual network throughput capacity of the basement.

[0268] Due to the shielding effect of the physical environment, probe data packets may be delayed or lost during transmission.

[0269] The home smart gateway calculates the current uplink bandwidth parameter as 12kbps based on the arrival time difference of the received valid response signal. This value is lower than the 64kbps transmission lower limit threshold set by the system to ensure basic services.

[0270] Based on the above comparison results, the home smart gateway triggers an asymmetric routing degradation mechanism in the underlying network protocol stack.

[0271] The home smart gateway calls the filtering interface to intercept the environmental monitoring multimedia video streams that the internal system attempts to upload, and clears the network layer transmission buffer.

[0272] The home smart gateway reads the statically preset geographic coordinate parameters from the operating system, combines them with the current trigger timestamp and the abnormal identifier parameters representing the fall condition, serializes them according to the bit field layout, and packages them to generate a byte feature string with a length of 32 bytes.

[0273] Subsequently, the home smart gateway invokes the control plane channel of the cellular IoT to send the byte signature string as an additional data segment of the network underlying signaling to the cloud server.

[0274] This asymmetric routing scheduling utilizes the high priority of control channel resources to achieve data penetration and transmission in weak network environments.

[0275] After the network access layer of the cloud server captures the underlying signaling message, it uses the underlying bitmask operation to reverse deserialize the geographic coordinate parameters and anomaly identification parameters.

[0276] The cloud server maps the extracted coordinates to the WGS-84 reference system and retrieves the nearest emergency center (assuming a spherical distance of 25 kilometers) from the local medical point of interest database.

[0277] The cloud server substitutes the distance parameter, time cost, and current network degradation flag (with a value of 1) into the aforementioned emergency response priority evaluation formula.

[0278] After normalized weight calculation, the resulting emergency response priority index is 92, which is greater than the set high priority scheduling threshold of 80.

[0279] In response to the assessment results, the cloud server invokes the emergency communication application programming interface to initiate a voice call to the patient's pre-set emergency contact terminal through the automatic speech synthesis system.

[0280] Simultaneously, the cloud server uses the secure hypertext transfer protocol to push alarm work orders containing structured geographic coordinates and abnormal operating condition identifiers to the emergency center dispatch system.

[0281] Through the aforementioned edge-side action monitoring, gateway-side network detection and degraded transmission, and cloud-side priority scheduling and evaluation, the system effectively reported abnormal signals in a weak network environment, ensuring the rescue response mechanism in emergency situations.

[0282] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A health management and follow-up system for epilepsy patients using the internet, characterized in that: This includes interconnected user terminals, wearable devices, communication gateways, and cloud servers; The cloud server is used to send management interaction instructions to the user terminal and receive data uploaded by the user terminal and the communication gateway. The user terminal is used to receive the management interaction command to generate a prompt event, and collect the touch response features and inertial pose features for the prompt event to calculate the execution confidence value; The wearable device is used to collect motion feature data, calculate the integral value of vital signs energy within a time sliding window, and broadcast a feature reversal signal when the integral value of vital signs energy is greater than a preset state threshold. The user terminal is also used to activate the local abnormal confirmation state machine after receiving the feature reversal signal, start the anti-false alarm countdown timer, and confirm that the system has entered the abnormal triggering condition when the anti-false alarm countdown timer reaches the end and no interrupt input operation is detected. The communication gateway is used to continuously send a series of probe data packets to the cloud server after confirming that the system has entered an abnormal triggering state, and to calculate the uplink bandwidth parameters based on the received confirmation response signal. When the uplink bandwidth parameter is greater than or equal to the preset transmission lower limit threshold, a full abnormal alarm message is sent to the cloud server through the regular user plane service channel; When the uplink bandwidth parameter is lower than the transmission lower limit threshold, the parameter information is extracted, packaged into a byte feature string, and the byte feature string is mapped to the underlying control channel resources for transmission. The cloud server is also used to parse the received full abnormal alarm message or the byte feature string, extract the original trigger timestamp and geographic coordinate parameters, calculate the emergency response priority index in combination with the preset weight coefficient, compare the emergency response priority index with the preset scheduling threshold, and execute the corresponding hierarchical scheduling instruction according to the comparison result.

2. The Internet-based knowledge, information, and behavior health management and follow-up system for epilepsy patients according to claim 1, characterized in that, The user terminal collects touch response features and inertial pose features in response to the prompt event to calculate the execution confidence value, specifically including: Record the timestamp of the instruction that generates the prompt event and the timestamp of the wake-up response when the user wake-up operation is detected, and use the difference between the two as the touch response latency data; In response to the wake-up operation, a short-time domain observation window is opened, the internal inertial measurement unit is invoked to acquire spatial pose time series data, and pose fluctuation characteristic parameters are extracted. The touch response latency data and the pose fluctuation feature parameters are aligned and concatenated to construct a two-dimensional feature vector, which is then input into a pre-deployed confidence verification model to calculate and obtain the execution confidence value. If the execution confidence value is greater than the valid confirmation threshold, record the interaction log; If the execution confidence value is less than or equal to the valid confirmation threshold, a secondary confirmation pop-up window is generated.

3. The Internet-based knowledge, information, and behavior health management and follow-up system for epilepsy patients according to claim 2, characterized in that, The extraction of pose fluctuation feature parameters specifically includes: Discrete triaxial acceleration vectors and triaxial angular velocity vectors are obtained within the short-time domain observation window; Calculate the difference between the triaxial acceleration vector and the arithmetic mean acceleration vector at each sampling point, and the difference between the triaxial angular velocity vector and the arithmetic mean angular velocity vector at each sampling point. The pose fluctuation characteristic parameters are obtained by performing integral calculations by combining preset acceleration weighting coefficients and angular velocity weighting coefficients.

4. The Internet-based knowledge, information, and behavior health management and follow-up system for epilepsy patients according to claim 1, characterized in that, The integral value of vital sign energy within the calculation time sliding window specifically includes: Obtain the triaxial acceleration vector at the k-th sampling moment within the time sliding window; extract the gravity component estimation vector using a low-pass filtering algorithm; The squared L2 of the difference between the triaxial acceleration vector and the gravitational component estimation vector is calculated, multiplied by the time interval between adjacent sampling points, and accumulated within the time sliding window to obtain the integral value of the vital signs energy.

5. The Internet-based knowledge, information, and behavior health management and follow-up system for epilepsy patients according to claim 1, characterized in that, After the countdown timer for preventing false alarms is started, the following is also included: The user terminal calls the authorized notification interface, vibration interface, sound playback interface or flashing prompt interface to continuously output physical warning signals; Record the intervention start timestamp when the false alarm prevention countdown timer is activated; If the interrupt input operation for the cancel control is captured before the countdown timer for preventing false alarms reaches its end, the exception confirmation state machine is terminated, the intervention termination timestamp is recorded, and the intervention delay data is calculated and stored in local memory based on the difference between the intervention termination timestamp and the intervention start timestamp.

6. The Internet-based knowledge, information, and behavior health management and follow-up system for epilepsy patients according to claim 1, characterized in that, The calculation of uplink bandwidth parameters based on the received acknowledgment signal specifically includes: The communication gateway synchronously starts the probe timeout timer, and the probe data packet column consists of a preset number of probe data packets; If no acknowledgment signal is received within the set timeout period, the current uplink bandwidth parameter is determined to be zero, the byte feature string is written to the local non-volatile buffer, and an audible and visual or near-field broadcast alarm is triggered. If the acknowledgment signal is received, the uplink bandwidth parameter is calculated based on the difference between the first timestamp of the first acknowledgment signal and the second timestamp of the last acknowledgment signal, combined with the number of probe data packets and the fixed byte length of a single probe data packet, and the principle of packet pair dispersion.

7. The Internet-based knowledge, information, and behavior health management and follow-up system for epilepsy patients according to claim 1, characterized in that, The extraction of parameter information and packaging it into a byte feature string specifically includes: The communication gateway invokes the filtering interface to perform pause, rate limiting, or discard processing on the user terminal service data stream forwarded by the gateway. Extract the current geographic coordinate parameters, including longitude and latitude values, from the operating system's location service interface; The original latitude and longitude values ​​are amplified according to a preset scaling factor and then rounded to obtain the longitude compression value and latitude compression value; the device anonymity identifier, abnormal working condition identifier, trigger timestamp, the longitude compression value, the latitude compression value, and the network degradation identifier are serialized according to the bit field layout and packaged to generate the byte feature string.

8. The Internet-based knowledge, information, and behavior health management and follow-up system for epilepsy patients according to claim 1, characterized in that, The calculation of the emergency response priority index by combining preset weighting coefficients specifically includes: The received byte feature string is deserialized using underlying bitmasks and logical shift operations to extract the original trigger timestamp and the geographic coordinate parameters; The geographic coordinate parameters are mapped to a standard Earth reference system, and spatial coordinates are retrieved from the medical point of interest database to calculate the spatial spherical distance from the current alarm location to the target medical rescue institution. The communication lag characteristics are calculated based on the receiving system time of the cloud server and the original trigger timestamp, and the estimated rescue time characteristics are calculated based on the spatial spherical distance. The emergency response priority index is calculated by multiplying the communication lag feature, the estimated rescue time feature, and the parsed network degradation flag by a preset system normalization weight coefficient and then adding them together.

9. The Internet-based knowledge, information, and behavior health management and follow-up system for epilepsy patients according to claim 1, characterized in that, The step of executing corresponding hierarchical scheduling instructions based on the comparison results specifically includes: When the emergency response priority index is greater than or equal to the high priority dispatch threshold, the voice synthesis system is invoked to initiate a voice call to the preset emergency contact terminal, and an alarm work order containing structured geographic coordinates and abnormal working condition identifiers is pushed to the emergency center dispatch system. When the emergency response priority index is less than the high priority scheduling threshold and greater than or equal to the low priority scheduling threshold, an early warning is sent to the family terminal, and the polling cycle parameter for the subsequent status verification of the user terminal is modified in the database. When the emergency response priority index is less than the low priority scheduling threshold, the parsed feature data is stored in the historical log database and marked as an event to be reviewed.

10. The Internet-based knowledge, information, and behavior health management and follow-up system for epilepsy patients according to claim 1, characterized in that, The cloud server is also used for: A long-term operation log is constructed by issuing status verification commands to the user terminal and recording touch response latency data; The abnormal alarm events generated by parsing and the intervention delay data uploaded by the user terminal are synchronously written into the long-term operation log to periodically update the state baseline model of the epilepsy patient, and management interaction instructions for the next cycle are generated based on the updated state baseline model.

Images