Intelligent wearable health monitoring system

Abstract

The invention provides an intelligent wearable health monitoring system, and relates to the technical field of sports biomechanics, and the method comprises the steps: collecting original sensing data streams of a triaxial accelerometer and a gyroscope in real time, and converting the original sensing data streams to a human skeleton coordinate system through a rotation matrix; calculating the Euler angle of the limb joint based on the converted data, and extracting gait phase parameters in combination with the curvature features of the motion trail; calculating a real-time stride frequency according to the gait phase parameter; when a plurality of continuous stride frequency sampling values exceed a preset medium-high intensity motion threshold value, generating a medium-high intensity motion state identifier; synchronously executing according to the medium-high intensity motion state identification, constructing a blood glucose dynamic evaluation window based on a continuous sampling sequence of a blood glucose sensor, calculating a time sequence gradient of blood glucose reading in the window, and generating a dynamic compensation coefficient according to the time sequence gradient. According to the invention, dynamic and accurate health monitoring and risk early warning in a motion scene are realized.

Description

Technical Field

[0001] This invention relates to the field of sports biomechanics technology, and in particular to a smart wearable health monitoring system. Background Technology

[0002] With increasing health awareness and growing demand for chronic disease management, smart wearable devices are playing an increasingly important role in the daily health monitoring of diabetic patients. Diabetic patients need to maintain stable blood sugar levels through appropriate exercise, but blood sugar levels fluctuate easily due to metabolic intensity during exercise. Simultaneously, physiological indicators such as heart rate also change with exercise intensity. If monitoring is not timely or accurate, it may lead to risks such as hypoglycemia and abnormal heart rate. Therefore, higher requirements are placed on dynamic health monitoring during exercise.

[0003] However, existing smart wearable health monitoring systems still have the following key technical shortcomings in practical applications for diabetic patients during exercise: Blood glucose monitoring accuracy is easily affected by exercise. During strenuous exercise such as running or skipping rope, the patient's arm movement can cause unstable contact between the blood glucose sensor and the skin, increasing the error in blood glucose testing. This error may lead to misjudgment, such as misreading a normal blood glucose value as a low blood glucose value, thus triggering unnecessary alarms, interfering with the patient's exercise plan, or causing panic.

[0004] Motion posture recognition suffers from latency. For sudden high-intensity exercise, existing motion posture recognition chips typically require more than 3 seconds of accumulated motion data to determine the intensity of the exercise. This results in abnormal heart rates at the beginning of exercise not being captured in time, posing a risk of missed detection and failing to meet the real-time risk warning requirements for sudden high-intensity exercise. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a smart wearable health monitoring system that enables dynamic and accurate health monitoring and risk warning in sports scenarios.

[0006] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: In a first aspect, a health monitoring method for smart wearable devices, the method comprising: Step 1: Real-time acquisition of raw sensor data streams from the three-axis accelerometer and gyroscope; conversion of the raw sensor data streams to the human skeletal coordinate system using a rotation matrix; calculation of Euler angles of limb joints based on the converted data; extraction of gait phase parameters by combining the curvature features of the motion trajectory; calculation of real-time gait frequency based on gait phase parameters; generation of a medium-to-high intensity exercise status identifier when multiple consecutive gait frequency sampling values ​​exceed a preset medium-to-high intensity exercise threshold. Step 2: Synchronously execute according to the medium-to-high intensity exercise state indicator, construct a dynamic blood glucose assessment window based on the continuous sampling sequence of the blood glucose sensor, calculate the temporal gradient of blood glucose readings within the window, and generate dynamic compensation coefficients based on the temporal gradient. Step 3: The dynamic compensation coefficient is weighted and fused with the preset baseline blood glucose compensation value to output the adaptive blood glucose compensation amount. The adaptive blood glucose compensation amount is superimposed on the current original blood glucose sampling value to generate the corrected blood glucose value. The first-order difference value of adjacent heart rate sampling points is calculated in real time. When the first-order difference value exceeds the preset heart rate change threshold, a heart rate acceleration event identifier is generated. Step 4: Make a combined judgment based on the corrected blood glucose value and the heart rate surge event indicator. When the corrected blood glucose value is lower than the hypoglycemia warning threshold and the heart rate surge event indicator is present, a combined warning is triggered. Step 5: Based on the composite warning, output the energy replenishment command through the tactile and voice feedback device, and push the warning information containing exercise time data, corrected blood glucose sequence and heart rate spike marker to the remote management platform.

[0007] Furthermore, the raw sensor data streams from the three-axis accelerometer and gyroscope are acquired in real time, and then transformed into the human skeletal coordinate system using a rotation matrix, including: Step 1.1: Receive the raw sensor data stream from the triaxial accelerometer and gyroscope uploaded in real time by the smart wearable device, perform timestamp alignment processing on the raw sensor data stream, and generate a synchronous sensor data sequence. Step 1.2: Load the pre-stored personalized rotation matrix parameters, perform coordinate transformation operations based on the synchronous sensing data sequence, and generate three-dimensional motion vectors of the hip and knee joints in the skeletal coordinate system. Step 1.3: Based on the three-dimensional motion vectors of the hip and knee joints, calculate the Euler angles of flexion and extension in the sagittal plane and the Euler angles of adduction and abduction in the coronal plane for the lower limb joints.

[0008] Furthermore, Euler angles of limb joints are calculated based on the transformed data, and gait phase parameters are extracted by combining the curvature features of the motion trajectory; real-time gait frequency is calculated based on the gait phase parameters; when multiple consecutive gait frequency samples exceed a preset medium-to-high intensity exercise threshold, a medium-to-high intensity exercise state identifier is generated, including: Step 1.4: Calculate the instantaneous curvature characteristics of the lower limb movement trajectory based on the Euler angles of flexion and extension of the hip and knee joints in the sagittal plane and the Euler angles of adduction and abduction in the coronal plane. Step 1.5: Extract key phase points of the gait cycle based on the extreme points of curvature characteristics, and generate phase segmentation parameters for the gait support phase and swing phase; Step 1.6: Calculate the single-step gait cycle based on the time interval of adjacent phase segmentation parameters, and generate the real-time step frequency based on the number of gait cycles per unit time. Step 1.7: When multiple consecutive step frequency sampling values ​​exceed the preset medium-high intensity exercise threshold, a medium-high intensity exercise status identifier is generated.

[0009] Furthermore, based on the synchronous execution of moderate-to-high intensity exercise status indicators, and based on the continuous sampling sequence of the blood glucose sensor, a dynamic blood glucose assessment window is constructed. The temporal gradient of blood glucose readings within the window is calculated, and dynamic compensation coefficients are generated based on the temporal gradient, including: Step 2.1: Based on the generation of medium-to-high intensity exercise state identifiers, start the real-time streaming data reception of the blood glucose sensor and output a continuous blood glucose sampling sequence with timestamp alignment. Step 2.2: Based on the continuous blood glucose sampling sequence, starting from the moment the motion state identifier is generated, the time span of the evaluation window is dynamically expanded according to the duration of the identifier to generate a dynamic blood glucose evaluation window dataset. Step 2.3: Perform time-series linear regression analysis on the dynamic blood glucose assessment window dataset and extract the regression slope as the time-series gradient value of blood glucose changes; Step 2.4: Input the time-series gradient values ​​into the preset gradient-compensation coefficient mapping table and output the dynamic compensation coefficients.

[0010] Furthermore, the dynamic compensation coefficient is weighted and fused with the preset baseline blood glucose compensation value to output an adaptive blood glucose compensation amount; the adaptive blood glucose compensation amount is added to the current raw blood glucose sampling value to generate a corrected blood glucose value; the first-order difference value between adjacent heart rate sampling points is calculated in real time; when the first-order difference value exceeds the preset heart rate change threshold, a heart rate acceleration event identifier is generated, including: Step 3.1: Call the dynamic compensation coefficient and perform a weighted fusion calculation with the pre-stored baseline blood glucose compensation value through a preset weight allocation strategy to generate an adaptive blood glucose compensation amount; Step 3.2: Obtain the current raw sample value uploaded by the blood glucose sensor in real time, and calculate the adaptive blood glucose compensation amount by positively superimposing it with the raw sample value to generate the exercise state corrected blood glucose value; Step 3.3, in parallel with the blood glucose processing flow, performs time-series difference calculation on the real-time data stream of the heart rate sensor, extracts continuous heart rate readings at fixed sampling intervals, and calculates the absolute value of the difference between two adjacent heart rate readings as the first-order difference value. Step 3.4: Compare the first-order difference value with the preset heart rate change threshold in real time. When multiple consecutive first-order difference values ​​exceed the threshold, generate a heart rate acceleration event identifier.

[0011] Furthermore, a combined judgment is made based on the corrected blood glucose value and the heart rate surge event indicator. When the corrected blood glucose value is lower than the hypoglycemia warning threshold and the heart rate surge event indicator is present, a combined warning is triggered, including: Step 4.1: Obtain the exercise-corrected blood glucose value, compare this value with the preset low blood glucose warning threshold, and output a judgment status indicator that the blood glucose is lower than the threshold. Step 4.2: Read the heart rate acceleration event identifier and its timestamp, verify whether the identifier is within the preset validity period based on the current system time, and output the identifier activation status verification result; Step 4.3: Receive the judgment status identifier and the identifier activation status verification result, and perform logical AND operation to generate a composite early warning trigger command.

[0012] Furthermore, based on the composite warning, an energy replenishment command is output through tactile and voice feedback devices, and a warning message containing exercise time data, corrected blood glucose sequence, and heart rate spike markers is pushed to the remote management platform, including: Step 5.1: Based on the composite warning, extract the exercise period data, correct the blood glucose sequence and heart rate acceleration markers; Step 5.2: Generate dual-channel instructions based on the extracted warning data, and generate local feedback instruction frames with tactile vibration pattern encoding and voice text encoding; encapsulate a remote warning information package containing exercise period data, corrected blood glucose sequence and heart rate acceleration markers; Step 5.3: Push local feedback command frames to the smart wearable device through a low-latency message queue to trigger the device to perform tactile vibration and voice broadcast; Step 5.4: Transmit the remote early warning information packet to the remote management platform through the secure API interface, and attach a composite early warning trigger timestamp.

[0013] Secondly, a smart wearable health monitoring system includes: The acquisition module is used to collect raw sensor data streams from the triaxial accelerometer and gyroscope in real time, and convert the raw sensor data streams to the human skeletal coordinate system through a rotation matrix; calculate the Euler angles of the limb joints based on the converted data, and extract gait phase parameters by combining the curvature features of the motion trajectory; calculate the real-time step frequency based on the gait phase parameters; when multiple consecutive step frequency sampling values ​​exceed the preset medium-high intensity exercise threshold, a medium-high intensity exercise state identifier is generated. The calculation module is used to synchronously execute according to the medium-to-high intensity exercise status indicator. Based on the continuous sampling sequence of the blood glucose sensor, it constructs a dynamic blood glucose assessment window, calculates the temporal gradient of blood glucose readings within the window, and generates a dynamic compensation coefficient based on the temporal gradient. The dynamic compensation coefficient is weighted and fused with a preset baseline blood glucose compensation value to output an adaptive blood glucose compensation amount. The adaptive blood glucose compensation amount is superimposed on the current original blood glucose sampling value to generate a corrected blood glucose value. The first-order difference value of adjacent heart rate sampling points is calculated in real time. When the first-order difference value exceeds the preset heart rate change rate threshold, a heart rate acceleration event indicator is generated. The processing module is used to make a combined judgment based on the corrected blood glucose value and the heart rate surge event flag. When the corrected blood glucose value is lower than the hypoglycemia warning threshold and the heart rate surge event flag is present, a combined warning is triggered. Based on the combined warning, the module outputs an energy replenishment command through tactile and voice feedback devices and pushes a warning message containing exercise time data, corrected blood glucose sequence and heart rate surge flag to the remote management platform.

[0014] Thirdly, a computing device includes: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0015] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0016] The above-described solution of the present invention has at least the following beneficial effects: By fusing multi-source sensor data and optimizing dynamic algorithms, the system achieves accurate monitoring and efficient early warning of the health status of diabetic patients during exercise. It quickly identifies medium- to high-intensity exercise through skeletal coordinate system transformation and gait feature extraction, solving the problem of delayed exercise status determination. With the help of a dynamic blood glucose assessment window and a weighted compensation mechanism, it eliminates blood glucose measurement errors caused by exercise interference, improving data accuracy. By correcting the combined judgment of blood glucose value and sudden heart rate rise events, it reduces misjudgment of a single indicator and improves the reliability of early warning. At the same time, by combining tactile and voice dual-modal local feedback and standardized data remote push, it ensures that patients respond to risks in a timely manner. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a health monitoring method for smart wearable devices provided in an embodiment of the present invention.

[0018] Figure 2 This is a schematic diagram of a smart wearable health monitoring system provided in an embodiment of the present invention. Detailed Implementation

[0019] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0020] like Figure 1As shown, an embodiment of the present invention proposes a health monitoring method for smart wearables, the method comprising the following steps: Step 1: Real-time acquisition of raw sensor data streams from the three-axis accelerometer and gyroscope; conversion of the raw sensor data streams to the human skeletal coordinate system using a rotation matrix; calculation of Euler angles of limb joints based on the converted data; extraction of gait phase parameters by combining the curvature features of the motion trajectory; calculation of real-time gait frequency based on gait phase parameters; generation of a medium-to-high intensity exercise status identifier when multiple consecutive gait frequency sampling values ​​exceed a preset medium-to-high intensity exercise threshold. Step 2: Synchronously execute according to the medium-to-high intensity exercise state indicator, construct a dynamic blood glucose assessment window based on the continuous sampling sequence of the blood glucose sensor, calculate the temporal gradient of blood glucose readings within the window, and generate dynamic compensation coefficients based on the temporal gradient. Step 3: The dynamic compensation coefficient is weighted and fused with the preset baseline blood glucose compensation value to output the adaptive blood glucose compensation amount. The adaptive blood glucose compensation amount is superimposed on the current original blood glucose sampling value to generate the corrected blood glucose value. The first-order difference value of adjacent heart rate sampling points is calculated in real time. When the first-order difference value exceeds the preset heart rate change threshold, a heart rate acceleration event identifier is generated. Step 4: Make a combined judgment based on the corrected blood glucose value and the heart rate surge event indicator. When the corrected blood glucose value is lower than the hypoglycemia warning threshold and the heart rate surge event indicator is present, a combined warning is triggered. Step 5: Based on the composite warning, output the energy replenishment command through the tactile and voice feedback device, and push the warning information containing exercise time data, corrected blood glucose sequence and heart rate spike marker to the remote management platform.

[0021] In this embodiment of the invention, a comprehensive and reliable health guarantee is provided for diabetic patients during exercise through multi-step collaborative optimization: First, it quickly identifies moderate to high-intensity exercise through skeletal coordinate system transformation and gait feature analysis, solving the problem of delayed exercise status determination; then, it effectively corrects blood glucose measurement errors caused by exercise interference by combining a dynamic blood glucose assessment window and a weighted compensation mechanism, improving data accuracy; by correcting the combined judgment of blood glucose value and sudden heart rate rise events, it reduces misjudgment of a single indicator and improves the accuracy of early warning; finally, it ensures timely response to risks by using tactile and voice dual-modal feedback and remote data push, comprehensively improving the timeliness and accuracy of health monitoring during exercise and ensuring patient safety during exercise.

[0022] In a preferred embodiment of the present invention, step 1 above may include: Step 1.1: Receive the raw sensor data stream from the triaxial accelerometer and gyroscope uploaded in real time by the smart wearable device, perform timestamp alignment processing on the raw sensor data stream, and generate a synchronous sensor data sequence. Step 1.2: Load the pre-stored personalized rotation matrix parameters, perform coordinate transformation operations based on the synchronous sensing data sequence, and generate three-dimensional motion vectors of the hip and knee joints in the skeletal coordinate system. Step 1.3: Based on the three-dimensional motion vectors of the hip and knee joints, calculate the Euler angles of flexion and extension in the sagittal plane and the Euler angles of adduction and abduction in the coronal plane for the lower limb joints.

[0023] In this embodiment of the invention, by aligning the original data from the triaxial accelerometer and gyroscope with timestamps, the time synchronization of multi-source sensor data is ensured, effectively eliminating errors caused by data acquisition delays. Loading personalized rotation matrix parameters for coordinate transformation adapts to the differences in skeletal structure among individuals, improving the calculation accuracy of the three-dimensional motion vectors of the hip and knee joints, making the data more closely match the actual movement characteristics of the human body. Furthermore, the targeted calculation of Euler angles of the lower limb joints in the sagittal and coronal planes accurately captures key movement states such as joint flexion, extension, adduction, and abduction, ensuring the accuracy and specificity of movement state recognition from the source.

[0024] In this embodiment of the invention, the specific steps include: Step 1.1: The triaxial accelerometer and gyroscope on the smart wearable device continuously collect raw data on the user's limb acceleration and angular velocity during movement, and upload it in real time in the form of timestamped data packets. After receiving the two data streams, the timestamp field of each data packet is extracted, and time-series calibration is performed on data packets with timestamp discrepancies. If there is a time misalignment between the sampled value of the acceleration data at time t1 and the sampled value of the gyroscope data at time t2 (t1≠t2), the intermediate time data is supplemented or the sampling interval is adjusted to align the two data streams under the same time reference. After the above processing, a synchronous sensing data sequence is generated to ensure that the acceleration and angular velocity data correspond one-to-one at each time point.

[0025] In a specific computational example, assume that the triaxial accelerometer uploads data at 10ms, 12ms, and 15ms, and the gyroscope uploads data at 11ms, 13ms, and 14ms. Using a timestamp alignment algorithm, the missing values ​​of the acceleration data at 11ms, 13ms, and 14ms are linearly interpolated (e.g., the 11ms acceleration value is fitted from the 10ms and 12ms data). The missing values ​​of the gyroscope data at 10ms and 15ms are processed in the same way. Finally, a synchronized sensor data sequence with timestamps of 10ms, 11ms, 12ms, 13ms, 14ms, and 15ms is generated, achieving time synchronization of the two data streams.

[0026] Step 1.2: After the synchronized sensor data sequence is generated, personalized rotation matrix parameters are loaded from the local storage unit. These parameters are pre-calculated and stored during the user's initial use through static posture calibration to adapt to the differences in skeletal anatomy (such as limb length and joint position) among different users. Subsequently, the synchronized acceleration vector and angular velocity vector are used as inputs and substituted into the personalized rotation matrix (e.g., a 3×3 orthogonal matrix) to perform coordinate transformation operations. Through matrix multiplication, the motion data in the sensor coordinate system (e.g., the device's default wearing coordinate system) is mapped to a coordinate system based on the human skeleton (e.g., a coordinate system defined along the limb's long axis, anterior-posterior axis, and internal-external axis, with the hip joint as the origin). After the above transformation, the three-dimensional motion vectors of the hip and knee joints in the skeletal coordinate system are output (e.g., the (x) of the hip joint... h y h , z h ) vector, knee joint (x) k y k , z k (Vector), which directly reflects the movement displacement and rotation trend of joints in the anatomical dimensions of the human body.

[0027] In a specific calculation example, assume the user-stored personalized rotation matrix is ​​M = [[0.99, 0.05, 0.02], [-0.04, 0.98, 0.17], [0.03, -0.16, 0.99]], and the acceleration vector at a certain moment in the synchronous sensing data is [1.8, -0.6, 1.1] (sensor coordinate system); substituting into matrix M, the calculation is as follows: x h =1.8×0.99+(-0.6)×(-0.04)+1.1×0.03, y h =1.8×0.05+(-0.6)×0.98+1.1×(-0.16),z h =1.8×0.02+(-0.6)×0.17+1.1×0.99, the three-dimensional motion vector of the hip joint is calculated as [1.82, -0.51, 1.05] (skeletal coordinate system); similarly, the three-dimensional motion vector of the knee joint is derived to complete the anatomical coordinate system mapping of the joint motion data.

[0028] Step 1.3: Based on the generated three-dimensional motion vectors of the hip and knee joints, the projection rules for the sagittal plane (anterior-posterior motion plane) and the coronal plane (lateral motion plane) are first defined. The sagittal plane is the plane formed by the limb's major axis (e.g., x-axis) and the anterior-posterior axis (e.g., y-axis) of the three-dimensional vector, while the coronal plane is the plane formed by the limb's major axis (x-axis) and the medial-lateral axis (e.g., z-axis). Subsequently, through vector projection and angle calculation, the three-dimensional motion vector is decomposed into the sagittal and coronal planes. Sagittal plane: Calculates the rotation angle of the lower limb joints around the internal and external axes (z-axis) to obtain the flexion angle (the degree of joint flexion, such as the positive angle when the knee joint is flexed) and the extension angle (the degree of joint extension, such as the negative angle or zero when the knee joint is extended). Coronal plane: Calculate the rotation angle of the lower limb joints around the anteroposterior axis (y-axis) to obtain the adduction angle (the limb moves towards the midline of the body, and the angle is negative) and the abduction angle (the limb moves away from the midline of the body, and the angle is positive).

[0029] In a specific calculation example, assume the three-dimensional motion vector of the knee joint is [1.2, 0.7, -0.2] (skeletal coordinate system, x is the major axis of the limb, y is the anterior-posterior axis, and z is the lateral axis): Sagittal plane analysis: The plane formed by the x-axis (major axis) and y-axis (anteroposterior axis) is extracted, and the vector projection is [1.2, 0.7]. The flexion angle is calculated by arctangent operation and is approximately 30.26° (representing a knee flexion of 30 degrees). Coronal plane analysis: The plane formed by the x-axis (major axis) and z-axis (internal and external axes) is extracted, and the vector projection is [1.2, -0.2]. The adduction / abduction angle is calculated by arctangent operation and is approximately -9.46° (the negative sign indicates that the knee joint adduction is 9 degrees).

[0030] By continuously calculating the sagittal flexion / extension angle and coronal adduction / abduction angle of the hip and knee joints, the dynamic changes in the movement posture of the lower limb joints are captured.

[0031] In a preferred embodiment of the present invention, step 1 above may include: Step 1.4: Calculate the instantaneous curvature characteristics of the lower limb movement trajectory based on the Euler angles of flexion and extension of the hip and knee joints in the sagittal plane and the Euler angles of adduction and abduction in the coronal plane. Step 1.5: Extract key phase points of the gait cycle based on the extreme points of curvature characteristics, and generate phase segmentation parameters for the gait support phase and swing phase; Step 1.6: Calculate the single-step gait cycle based on the time interval of adjacent phase segmentation parameters, and generate the real-time step frequency based on the number of gait cycles per unit time. Step 1.7: When multiple consecutive step frequency sampling values ​​exceed the preset medium-high intensity exercise threshold, a medium-high intensity exercise status identifier is generated.

[0032] In this embodiment of the invention, by calculating the instantaneous curvature features of the lower limb movement trajectory based on the Euler angles of the joints, subtle dynamic changes in gait are accurately captured, providing a more discriminative feature basis for gait cycle analysis. Furthermore, by extracting key phase points through curvature extreme points, the precise segmentation of the gait support phase and swing phase is achieved, ensuring the reliability of the phase parameters. On this basis, the step frequency is calculated based on the phase interval, and the movement intensity is judged by continuous sampling values, effectively avoiding misjudgments caused by fluctuations in a single sampling value, and improving the accuracy and stability of medium-to-high intensity movement state recognition.

[0033] In this embodiment of the invention, the specific steps include: Step 1.4 is based on the acquired Euler angles for sagittal flexion / extension and coronal adduction / abduction of the hip and knee joints. First, a geometric model of the lower limb movement trajectory is constructed: the hip joint is used as the proximal reference node for limb movement, and the knee joint as the intermediate linkage node. Based on the temporal changes of the joint Euler angles (such as the dynamic curve of the sagittal flexion angle over time and the real-time fluctuation of the coronal adduction angle), the spatial posture of the lower limbs (such as the thigh and calf) is derived through inverse kinematics, and then the three-dimensional movement trajectory curve of the lower limb distals (such as the ankle joint) is fitted. Based on this trajectory, instantaneous curvature characteristics are calculated: the trajectory curve is discretized (the coordinates of trajectory points are extracted at fixed time intervals), and the curvature of the curve is derived through the rate of change of position of adjacent trajectory points and the second-order rate of change (the larger the curvature value, the more severe the trajectory curvature). This captures the differences in trajectory morphology between the limb swing and support phases in the gait cycle (e.g., the curvature of the trajectory during the swing phase is usually significantly higher than that during the support phase due to the large swing of the limb).

[0034] In a specific computational example, it is assumed that the sagittal flexion angle of the knee joint changes as a periodic curve over time, and the coronal adduction angle of the hip joint is θadduct(t) = 8°cos(πt / 300). The trajectory of the ankle joint in the sagittal plane is derived as a "parabolic" curve. A polynomial curve is fitted to the trajectory points collected every 20ms, and the curvature at each time point is calculated through discrete differential operations. If the first-order rate of change of the trajectory point at a certain moment is (Δx / Δt, Δy / Δt), and the second-order rate of change is (Δ... 2 x / Δt 2 Δ 2 y / Δt 2 The instantaneous curvature at that moment is calculated using the curvature formula (discretized derivation). The bending characteristics of the trajectory are accurately quantified.

[0035] Step 1.5: Perform extreme value detection on the generated instantaneous curvature time-series data (such as the k(t) curve). Use a sliding window method (such as a 50ms window) or a difference algorithm to identify local maxima and minima of the curvature curve. These extreme points correspond to key phase nodes of the gait cycle: for example, curvature maxima often occur at the heel strike (start of the support phase) or toe lift (end of the support phase, start of the swing phase) (because the curvature of the trajectory reaches its peak at this point), while curvature minima mostly correspond to the middle of the swing phase (when the trajectory is relatively straight). Through preset phase mapping rules, extract key phase points of the gait cycle (such as the start time ts of the support phase, the start time tw of the swing phase, etc.), and based on these time markers, generate phase segmentation parameters for the gait support and swing phases to achieve accurate division of gait stages.

[0036] In a specific computational example, assume that the instantaneous curvature curve k(t) reaches its first maximum value (k=0.025mm) at t=120ms. -1 (corresponding to heel contact with the ground, the support phase begins), a minimum value (k=0.008mm) appears at t=350ms. -1 (corresponding to the middle of the oscillation phase), a second maximum value (k=0.022mm) appears at t=580ms. -1 (This corresponds to the toes leaving the ground, the end of the support phase, and the beginning of the oscillation phase). By marking these time points, support phase segmentation parameters [120ms, 580ms] and oscillation phase segmentation parameters [580ms, 1000ms] are generated (assuming the next support phase begins at 1000ms), providing clear stage boundaries for gait periodicity analysis.

[0037] Step 1.6: Extract the time interval between adjacent phase segmentation parameters. This interval is the single-step gait cycle, representing the time it takes for the user to complete each step. Then, using a fixed time window (e.g., 5 seconds) as the statistical unit, count the number of gait cycles within the window (e.g., 10 gait cycles detected within the window), and divide this by the window duration (5 seconds) to obtain the real-time cadence (e.g., 10 ÷ 5 = 2 steps / second, equivalent to 120 steps / minute). By continuously sliding the time window (e.g., sliding for 1 second each time) to update the calculation, dynamic monitoring of the cadence is achieved, reflecting changes in the user's movement rhythm in real time.

[0038] In a specific calculation example, if the start times of three consecutive support phases are ts1=1000ms, ts2=1500ms, and ts3=2000ms, then the single-step gait cycle Tgait1=1500-1000=500ms and Tgait2=2000-1500=500ms (assuming a stable stride length). Using a 1-minute statistical window, if 120 gait cycles are detected within 60 seconds, the real-time cadence is 120÷60=2 steps / second=120 steps / minute, quantifying the user's current movement speed characteristics.

[0039] Step 1.7: Pre-store the medium-to-high intensity exercise threshold (e.g., cadence ≥ 120 steps / minute, which can be calibrated through clinical trials or kinematic standards). For the real-time cadence sampling values ​​generated in Step 1.6 (e.g., calculating cadence every second to form a sampling sequence f(1), f(2), ..., f(n)), perform continuous threshold verification. If N consecutive sampling values ​​(e.g., N=5, to avoid misjudgment caused by instantaneous fluctuations) all exceed the preset threshold (e.g., 120 steps / minute), the user is determined to have entered a medium-to-high intensity exercise state, and a corresponding status flag (e.g., Boolean flag FlagMotion=1) is generated.

[0040] In a specific calculation example, assume the step frequency sampling sequence is [123, 131, 127, 130, 126] (unit: steps / minute), the preset threshold is 120, and the number of consecutive judgments N=5. Since all 5 sample values ​​are greater than 120, a medium-to-high intensity exercise state identifier is generated; if a sample value is 118, the consecutive count is reset (the number of consecutive times exceeding the threshold is counted again) to avoid false triggering due to short-term fluctuations and to ensure the robustness and accuracy of the exercise state determination.

[0041] In a preferred embodiment of the present invention, step 2 above may include: Step 2.1: Based on the generation of medium-to-high intensity exercise state identifiers, start the real-time streaming data reception of the blood glucose sensor and output a continuous blood glucose sampling sequence with timestamp alignment. Step 2.2: Based on the continuous blood glucose sampling sequence, starting from the moment the motion state identifier is generated, the time span of the evaluation window is dynamically expanded according to the duration of the identifier to generate a dynamic blood glucose evaluation window dataset. Step 2.3: Perform time-series linear regression analysis on the dynamic blood glucose assessment window dataset and extract the regression slope as the time-series gradient value of blood glucose changes; Step 2.4: Input the time-series gradient values ​​into the preset gradient-compensation coefficient mapping table and output the dynamic compensation coefficients.

[0042] In this embodiment of the invention, by synchronously receiving real-time data from the blood glucose sensor and aligning the timestamps when starting medium-to-high intensity exercise, the temporal consistency between blood glucose sampling and exercise is ensured, avoiding interference from time-series deviations in the analysis. Using the start time of exercise as a benchmark, the evaluation window is dynamically expanded according to the duration of exercise, allowing the analysis range of blood glucose data to adapt to the exercise process and comprehensively capture the dynamic changes in blood glucose during exercise. By extracting the gradient of blood glucose changes through time-series linear regression, the rate of rise and fall of blood glucose over time is accurately quantified, providing an objective basis for the generation of compensation coefficients. Combined with a preset gradient-compensation coefficient mapping table, dynamic compensation coefficients are output, allowing the compensation amount to be flexibly adjusted according to the trend of blood glucose changes, rather than using a fixed value, thus improving the targeting and accuracy of blood glucose compensation in exercise scenarios.

[0043] In this embodiment of the invention, the specific steps include: Step 2.1: Upon generating a medium-to-high intensity exercise status identifier, the smart wearable device immediately triggers the blood glucose sensor to enter real-time sampling mode, continuously receiving the uploaded blood glucose data packets with timestamps. The device extracts the timestamp field of each data packet and compares it with the time reference of the synchronized triaxial accelerometer and gyroscope data (such as the unified timeline of the device's built-in system), performing a synchronization calibration operation.

[0044] Specifically, if a discrepancy is detected between the blood glucose sampling timestamp and the exercise data timeline (for example, the exercise data is synchronized with a 1ms precision, while the initial timestamp of the blood glucose data has a ±5ms offset), the device uses a linear interpolation algorithm to fit the blood glucose value at the intermediate moment based on the values ​​of adjacent valid blood glucose sampling points, thus completing the missing data in the time series. Through the above operations, a continuous blood glucose sampling sequence with strictly aligned timestamps is finally generated, ensuring that the blood glucose data and exercise status data remain consistent in the time dimension.

[0045] Specific Implementation: Assume that the moderate-to-high intensity exercise status indicator is generated at system time t=2000ms. Subsequently, the blood glucose sensor successively uploads data packets with timestamps of t=2003ms (corresponding to a blood glucose value of 6.1 mmol / L) and t=2008ms (corresponding to a blood glucose value of 6.0 mmol / L). After detecting the timestamp deviation, the device uses forward prediction fitting to obtain blood glucose values ​​of 6.08 mmol / L at t=2001ms and 6.04 mmol / L at t=2002ms for the interval from t=2000ms to t=2003ms; for the interval from t=2003ms to t=2008ms, the data is supplemented in the same way. Finally, a continuous blood glucose sequence with timestamps covering t=2000ms to t=2008ms and time intervals of 1ms is generated, achieving time synchronization with the exercise data.

[0046] Step 2.2: Based on the output continuous blood glucose sampling sequence, the smart wearable device sets the generation time of the medium-to-high intensity exercise status identifier as the starting point of the dynamic blood glucose assessment window, and dynamically adjusts the time span of the assessment window according to the duration of the exercise status identifier.

[0047] Specifically, an initial duration of the assessment window is preset (e.g., T0 = 90 seconds), and a window expansion rule is defined: whenever the duration of moderate-to-high intensity exercise increases by ΔT = 60 seconds, the duration of the assessment window is simultaneously expanded by ΔText = 30 seconds (this expansion logic is based on the metabolic patterns of exercise and the time lag characteristics of blood glucose changes, and has been determined through previous experimental calibration). The device monitors the duration of moderate-to-high intensity exercise in real time, and extracts blood glucose data for the corresponding time period according to the above rules to form a dynamic blood glucose assessment window dataset. Through this dynamic adjustment method, the assessment window can adapt to the dynamic characteristics of blood glucose changes during exercise (for example, blood glucose fluctuates rapidly in the early stages of exercise, and a short window can quickly capture the trend of change; during the stable period of exercise, a long window can better integrate the overall change pattern).

[0048] Specific Implementation: If a moderate-to-high intensity exercise state is generated at t=0 seconds, and the initial window duration is set to T0=90 seconds, then the dataset for the first evaluation window corresponds to the blood glucose sequence from t=0 to t=90 seconds. When the exercise continues to t=150 seconds, because the duration of the exercise state has increased by 60 seconds compared to the initial moment, a window expansion is triggered. At this time, the window duration expands to 90+30=120 seconds, and the new dataset corresponds to the blood glucose data from t=0 to t=120 seconds. If the exercise continues further to t=210 seconds (the duration increases by another 60 seconds), the window further expands to 150 seconds, and the dataset covers the blood glucose data from t=0 to t=150 seconds. Through dynamic expansion, the evaluation window always covers the blood glucose change process from the start of exercise to the current moment, and the window duration adaptively adjusts with the duration of exercise.

[0049] Step 2.3, in the data preparation stage, the device extracts two core pieces of information for each blood glucose sampling point from the window dataset: first, the time difference between the sampling point and the window start point (i.e., the interval from the moment the motion status identifier is generated to the moment the sampling point is collected, in seconds), which is used as the independent variable of the model (denoted as t); second, the blood glucose value corresponding to the sampling point (in mmol / L), which is used as the dependent variable of the model (denoted as v). Through this extraction process, a set of sample data pairs containing the (t, v) correspondence is formed (e.g., (t1, v1), (t2, v2)...(t...v1)...v2 ... n v n (), which provides basic samples for model building.

[0050] Furthermore, based on the aforementioned sample data pairs, the basic form of a linear regression model is constructed. This model aims to characterize the linear trend of blood glucose levels changing over time, assuming a linear relationship between the dependent variable v and the independent variable t: v = a × t + b (where a is the regression slope, reflecting the rate of change of blood glucose over time; b is the intercept, reflecting the baseline blood glucose level at the beginning of the window). The core of model construction lies in solving for the values ​​of a and b through sample data, so that this linear relationship can best reflect the actual trend of blood glucose changes.

[0051] In this process, the least squares method is used as the core algorithm for model fitting. Specifically, this algorithm calculates the sum of squared deviations between the actual blood glucose value v and the model's predicted value (a×t+b), and finds the values ​​of a and b that minimize this sum of squares, thereby determining the best-fitting straight line. The essence of this process is to use mathematical optimization to make the model as close as possible to the actual data distribution, thus accurately capturing the linear change characteristics of blood glucose over time.

[0052] Specific Implementation: A dynamic blood glucose assessment window covers a time period from t=0 to t=120 seconds, containing six sampling points: blood glucose level of 7.2 mmol / L at t=20 seconds, 7.0 mmol / L at t=40 seconds, 6.8 mmol / L at t=60 seconds, 6.6 mmol / L at t=80 seconds, 6.4 mmol / L at t=100 seconds, and 6.2 mmol / L at t=120 seconds. After fitting the data using the least squares method, the regression slope k = -0.017 mmol / L / second (i.e., blood glucose decreases by 1.02 mmol / L per minute) is obtained. This value represents the temporal gradient of blood glucose changes within the current window.

[0053] Step 2.4: The smart wearable device pre-stores a gradient-compensation coefficient mapping table, which is generated through clinical trial calibration. For scenarios such as rapid blood glucose decrease, slow decrease, and abnormal increase, the table tests the degree of influence of motion interference on blood glucose measurement results and establishes the correspondence between the time-series gradient value of blood glucose change and the compensation coefficient (the compensation coefficient is used to correct the blood glucose measurement deviation caused by motion interference; the positive coefficient compensates for the measurement deviation of blood glucose decrease, and the negative coefficient corrects the measurement deviation of false increase in blood glucose).

[0054] In practice, the device inputs the time-series gradient values ​​of blood glucose changes into the mapping table, and outputs the corresponding dynamic compensation coefficient through key-value matching. This coefficient serves as a time-varying parameter, providing data support for the calculation of adaptive blood glucose compensation in step 3.

[0055] Specific Implementation: A preset gradient-compensation coefficient mapping table specifies that when the temporal gradient value of blood glucose change is k = -0.017 mmol / L / second (corresponding to a decrease of 1.02 mmol / L per minute), the corresponding compensation coefficient is 0.25; when the gradient value is k = -0.02 mmol / L / second, the corresponding compensation coefficient is 0.3. If the calculated temporal gradient value of blood glucose change is k = -0.017 mmol / L / second, the device outputs a dynamic compensation coefficient of 0.25 by looking up the table.

[0056] In a preferred embodiment of the present invention, step 3 above may include: Step 3.1: Call the dynamic compensation coefficient and perform a weighted fusion calculation with the pre-stored baseline blood glucose compensation value through a preset weight allocation strategy to generate an adaptive blood glucose compensation amount; Step 3.2: Obtain the current raw sample value uploaded by the blood glucose sensor in real time, and calculate the adaptive blood glucose compensation amount by positively superimposing it with the raw sample value to generate the exercise state corrected blood glucose value; Step 3.3, in parallel with the blood glucose processing flow, performs time-series difference calculation on the real-time data stream of the heart rate sensor, extracts continuous heart rate readings at fixed sampling intervals, and calculates the absolute value of the difference between two adjacent heart rate readings as the first-order difference value. Step 3.4: Compare the first-order difference value with the preset heart rate change threshold in real time. When multiple consecutive first-order difference values ​​exceed the threshold, generate a heart rate acceleration event identifier.

[0057] In this embodiment of the invention, an adaptive compensation amount is generated by weighted fusion of the dynamic compensation coefficient and the baseline blood glucose compensation value. This takes into account the real-time blood glucose fluctuation characteristics under exercise conditions, preserves the stability of the baseline compensation, and improves the adaptability of blood glucose compensation. The compensation amount is then superimposed on the original blood glucose value to obtain a corrected blood glucose value, which can accurately eliminate measurement deviations caused by exercise interference and ensure the authenticity of blood glucose data. Simultaneously, by calculating the first-order difference value of heart rate in real time and monitoring whether it exceeds a threshold to generate a heart rate acceleration event identifier, abnormal heart rate change trends can be keenly captured. The above steps work synergistically to ensure the accuracy of blood glucose data in exercise scenarios and achieve timely capture of abnormal heart rate changes.

[0058] In this embodiment of the invention, the specific steps include: Step 3.1 is based on the dynamic compensation coefficient and the device's pre-stored baseline blood glucose compensation value. The smart wearable device first retrieves the above two parameters from the storage unit. The baseline blood glucose compensation value is a preset benchmark correction amount for the exercise scenario (such as obtained through statistical analysis of blood glucose measurement deviations from a large number of exercisers, used to cover the common interference of exercise on blood glucose sensors), while the dynamic compensation coefficient reflects the real-time trend of blood glucose changes during the current exercise.

[0059] Specifically, the preset weighting strategy is dynamically adjusted based on the duration of exercise and the intensity of blood glucose changes. For example, in the early stages of exercise (e.g., the first 5 minutes), the dynamic compensation coefficient has a lower weight (e.g., 30%), while the baseline blood glucose compensation value has a higher weight (e.g., 70%), to avoid over-correction before the blood glucose trend has stabilized. When exercise lasts for more than 10 minutes and the blood glucose gradient is significant, the weight of the dynamic compensation coefficient increases (e.g., 60%), while the weight of the baseline value decreases (e.g., 40%), making the compensation amount more closely match real-time changes. Through this weighted fusion calculation, the final adaptive blood glucose compensation amount retains the stability of the baseline correction while incorporating real-time dynamic adjustments, achieving precise adaptation to blood glucose measurement deviations during exercise.

[0060] In a specific embodiment, if the current dynamic compensation coefficient is 0.25, the baseline blood glucose compensation value is 0.3, and the duration of exercise has reached 15 minutes (corresponding to a dynamic coefficient weight of 60%), then the adaptive blood glucose compensation amount is 0.25×60%+0.3×40%=0.27, which combines real-time trends and baseline corrections.

[0061] Step 3.2: After generating the adaptive blood glucose compensation, the smart wearable device receives the current raw sample value uploaded by the blood glucose sensor in real time (i.e., the uncorrected instantaneous blood glucose reading, which may be affected by motion vibration, blood flow changes, etc.). During this process, the device performs a positive superposition operation between the adaptive blood glucose compensation and the raw sample value. This is because, during exercise, the blood glucose sensor often results in lower readings due to limb movement (e.g., unstable sensor contact with the skin affects signal acquisition). The positive superposition compensation can effectively offset such systematic biases, making the result closer to the true blood glucose level.

[0062] By superimposing the values, a corrected blood glucose value under exercise conditions is generated, which better reflects the user's actual blood glucose status compared to the original value. In a specific embodiment, if the current original blood glucose sampling value is 5.2 mmol / L and the adaptive blood glucose compensation is 0.27 mmol / L, the superimposed corrected blood glucose value is 5.47 mmol / L, which corrects for measurement bias caused by exercise interference.

[0063] Step 3.3 is executed in parallel with the blood glucose processing flow described above to ensure the real-time and synchronous nature of the heart rate data. The smart wearable device continuously receives real-time data streams from the heart rate sensor (such as pulse signals collected by a photoelectric sensor) and extracts continuous heart rate readings (unit: beats / minute) at fixed sampling intervals (such as once every 1 second) to form a heart rate time series (such as h1, h2, h3, ..., h...). n Where h1 is the heart rate in the first second, h2 is the heart rate in the second second, and so on.

[0064] Based on this, the device performs temporal difference calculation on the sequence: taking two adjacent heart rate readings (such as h2 and h1, h3 and h2), and calculating the absolute value of the difference between them, which is the first-order difference value (such as |h2-h1|, |h3-h2|). The magnitude of the first-order difference value directly reflects the instantaneous rate of change of heart rate. The larger the difference, the more drastic the fluctuation of heart rate per unit time. This feature can effectively capture sudden increases in heart rate that may occur during exercise (such as a sudden increase in heart rate caused by sympathetic nerve excitation due to hypoglycemia).

[0065] In a specific embodiment, if the continuous heart rate readings are 80, 85, 92, and 100 (unit: beats / minute), the corresponding first-order difference values ​​are |85-80|=5, |92-85|=7, and |100-92|=8, respectively. These values ​​quantify the magnitude of heart rate change over time.

[0066] Step 3.4: The smart wearable device stores a pre-defined heart rate variability threshold (e.g., 5 beats / minute, calibrated using clinical data to distinguish between normal heart rate fluctuations and abnormal spikes). The device compares the generated first-order difference value with this threshold in real time, and introduces a continuous judgment mechanism: only when multiple consecutive (e.g., 3 or more) first-order difference values ​​exceed the preset threshold is it determined to be a heart rate acceleration event.

[0067] The purpose of this mechanism is to avoid misjudgments caused by momentary interference (such as accidental limb tremors leading to abnormal heart rate readings) and to ensure the reliability of event identifiers. When the condition of continuous exceeding the threshold is met, the device generates a heart rate acceleration event identifier (such as a Boolean identifier FlagHR=1).

[0068] In a specific embodiment, if the preset heart rate change threshold is 5 beats / minute, and the consecutive first-order difference values ​​are 6, 7, and 8 (all exceeding 5), the device determines it as a heart rate acceleration event and generates a corresponding identifier; if one of the difference values ​​is 4 (below the threshold), the consecutive count is reset, and the number of consecutive times exceeding the threshold needs to be counted again, so as to ensure the rigor and accuracy of the heart rate acceleration event judgment.

[0069] In a preferred embodiment of the present invention, step 4 above may include: Step 4.1: Obtain the exercise-corrected blood glucose value, compare this value with the preset low blood glucose warning threshold, and output a judgment status indicator that the blood glucose is lower than the threshold. Step 4.2: Read the heart rate acceleration event identifier and its timestamp, verify whether the identifier is within the preset validity period based on the current system time, and output the identifier activation status verification result; Step 4.3: Receive the judgment status identifier and the identifier activation status verification result, and perform logical AND operation to generate a composite early warning trigger command.

[0070] In this embodiment of the invention, the accuracy of blood glucose risk assessment is ensured by first accurately determining whether the blood glucose level is below the low blood glucose warning threshold based on the corrected blood glucose value; then, the timeliness of the heart rate acceleration event identifier is verified to ensure the correlation between heart rate abnormality and the current state; finally, the combined judgment of the two is achieved by logical AND operation, which avoids the situation where a single indicator (such as simple blood glucose fluctuation or occasional heart rate acceleration) falsely triggers the warning, thereby improving the accuracy and reliability of the warning.

[0071] In this embodiment of the invention, the specific steps include: Step 4.1 establishes a modified blood glucose value based on the exercise state output. This modified value has eliminated exercise interference through a dynamic compensation mechanism and can accurately reflect the user's current blood glucose level. The smart wearable device first retrieves this modified blood glucose value from the data cache unit and simultaneously loads a pre-stored hypoglycemia warning threshold. This threshold is set based on clinical diabetes management standards (e.g., 3.9 mmol / L, which can be adjusted during initial configuration according to individual user circumstances to accommodate differences in hypoglycemia tolerance among different users).

[0072] Specifically, the device compares the corrected blood glucose level with a hypoglycemia warning threshold: if the corrected blood glucose level is less than the preset threshold, it is determined to be a hypoglycemic state, and a status indicator indicating that the blood glucose level is below the threshold is output (e.g., a "1" indicates that the hypoglycemia risk condition is met); if the corrected blood glucose level is greater than or equal to the preset threshold, it is determined to be a normal blood glucose state, and a corresponding status indicator is output (e.g., a "0" indicates that the hypoglycemia risk condition is not met). Through this operation, a preliminary screening of blood glucose risk is achieved.

[0073] In a specific embodiment, if the preset low blood sugar warning threshold is 3.9 mmol / L, and the current exercise state corrects the blood sugar value to 3.5 mmol / L, since 3.5 < 3.9, the device outputs a judgment status flag "1" indicating that the blood sugar is below the threshold; if the corrected blood sugar value is 4.2 mmol / L, since 4.2 ≥ 3.9, the judgment status flag "0" is output, thereby clarifying whether the blood sugar is in the risk range.

[0074] Step 4.2: Based on the generated heart rate acceleration event identifier and its associated timestamp (which records the exact moment the identifier was generated, accurate to the millisecond level), the smart wearable device first reads the heart rate acceleration event identifier (e.g., "FlagHR=1") and its corresponding timestamp (e.g., thr=16:23:10.500) from the event recording unit, and simultaneously obtains the current system time (e.g., tnow=16:23:12.300).

[0075] Based on this, the device verifies whether the identifier is within a preset validity period. This validity period is set according to physiological response patterns (e.g., 5 minutes; heart rate acceleration caused by hypoglycemia is usually time-related to low blood sugar. Exceeding this duration may be due to other factors and does not need to be included in the composite judgment). The specific verification method is as follows: calculate the difference between the current system time and the identifier's timestamp (e.g., tnow-thr=1.8 seconds). If the difference is less than the preset validity period (5 minutes), the identifier is determined to be still active, and the identifier activation status verification result is output as "valid"; if the difference is greater than or equal to the preset validity period, the identifier is determined to be invalid, and the verification result is output as "invalid". Through timeliness verification, the correlation between the heart rate acceleration event and the current blood sugar status is ensured, avoiding interference from historical events in the judgment.

[0076] In a specific embodiment, if the preset validity period is 5 minutes, the timestamp of the heart rate acceleration event identifier is 16:20:00, the current system time is 16:23:30, and the difference between the two is 3 minutes and 30 seconds (less than 5 minutes), then the output identifier activation status verification result is "valid"; if the current system time is 16:26:00, and the difference is 6 minutes (greater than 5 minutes), then the output verification result is "invalid", thereby ensuring the time matching between the heart rate event and the current state.

[0077] Step 4.3: The smart wearable device first receives these two parameters: the judgment status identifier (indicating whether blood glucose is below the warning threshold) and the identifier activation status verification result (indicating whether the heart rate acceleration event is valid and related to the current time).

[0078] Based on this, the device performs a logical AND operation on two parameters: a composite warning trigger command (e.g., command signal "Trigger=1") is generated only when the judgment status flag is "1" (blood glucose is below the threshold) and the flag activation status verification result is "valid" (heart rate acceleration event is within the validity period); if either of these conditions is not met (e.g., blood glucose is normal or the heart rate event has expired), no trigger command is generated (e.g., command signal "Trigger=0"). Through the logical AND operation, the device achieves a coordinated judgment of blood glucose risk and heart rate abnormality, avoiding false warnings caused by fluctuations in a single indicator (e.g., transient hypoglycemia without heart rate abnormality, or isolated heart rate acceleration with normal blood glucose), thus ensuring the rigor of warning triggering.

[0079] In a specific embodiment, if the output judgment status identifier is "1" (blood sugar is below the threshold) and the output identifier activation status verification result is "valid" (heart rate event is within the validity period), then the logical AND operation result is "true", and a compound warning trigger command is generated; if the judgment status identifier is "0" (blood sugar is normal), even if the heart rate event is valid or has expired, even if blood sugar is below the threshold, the logical AND operation result is "false", and no trigger command is generated, thereby achieving accurate identification of hypoglycemia risk.

[0080] In a preferred embodiment of the present invention, step 5 above may include: Step 5.1: Based on the composite warning, extract the exercise period data, correct the blood glucose sequence and heart rate acceleration markers; Step 5.2: Generate dual-channel instructions based on the extracted warning data, and generate local feedback instruction frames with tactile vibration pattern encoding and voice text encoding; encapsulate a remote warning information package containing exercise period data, corrected blood glucose sequence and heart rate acceleration markers; Step 5.3: Push local feedback command frames to the smart wearable device through a low-latency message queue to trigger the device to perform tactile vibration and voice broadcast; Step 5.4: Transmit the remote early warning information packet to the remote management platform through the secure API interface, and attach a composite early warning trigger timestamp.

[0081] In this embodiment of the invention, by first extracting exercise time data, correcting blood glucose sequences and heart rate acceleration markers, the completeness and relevance of the warning information are ensured. Then, a local feedback command frame containing tactile vibration and voice text is generated and a remote warning information package is encapsulated, taking into account both the user's local real-time perception and the synchronous awareness needs of the remote management. By using a low-latency message queue to push local commands, bimodal feedback can be quickly triggered, ensuring that the patient receives the energy replenishment command as soon as possible. Information is transmitted to the remote platform through a secure API interface with an attached timestamp, ensuring the security of data transmission.

[0082] In this embodiment of the invention, the specific steps include: Step 5.1: Based on the generated composite warning trigger command, this command indicates that the risk response mechanism needs to be activated. The smart wearable device first responds to the composite warning and extracts three types of core information from the local data storage unit: First, exercise time data, covering the start time, duration, and real-time cadence sequence of medium-to-high intensity exercise (derived from the exercise state analysis results in Step 1); second, corrected blood glucose sequence, i.e., blood glucose time-series data after dynamic compensation during exercise (derived from the calculation results in Step 3.2); and third, heart rate acceleration markers, including the generation time, duration, and corresponding heart rate change curve of the heart rate acceleration event (derived from the event record in Step 3.4).

[0083] Specifically, the extraction process is achieved through a pre-defined data index: exercise time period data is bound to timestamps indicating moderate to high intensity exercise status; blood glucose sequences are aligned with the exercise time period along the timeline; and heart rate acceleration markers are associated with compound warning trigger commands via event IDs. This operation ensures that the extracted information fully covers the exercise background, blood glucose change trends, and abnormal heart rate characteristics of the risk event.

[0084] In a specific embodiment, if the composite warning is triggered at 16:23:15, the extracted exercise period data is the step frequency record from 16:18:00 (exercise start) to 16:23:15, the corrected blood glucose sequence is the correction value every 10 seconds during this period (such as 3.8, 3.6, 3.5 mmol / L), and the heart rate acceleration is marked as the event generated at 16:23:10 and the corresponding record of the heart rate suddenly rising from 80 to 100 beats / minute, so as to completely restore the physiological and exercise state at the time of the risk.

[0085] Step 5.2: Based on the extracted early warning data, simultaneously generate local feedback instructions for users and remote early warning information for management, forming a dual-channel response mechanism.

[0086] For local feedback command frames: The device first generates a tactile vibration pattern code based on the risk level (compound warnings correspond to high risk). For example, it uses a pattern of "3 rapid vibrations (0.5 seconds each, 0.2 seconds interval) + 1 long vibration (2 seconds)" to convey the urgency level through the combination of vibration frequency and duration. At the same time, it generates a voice text code to convert the warning information into natural language text (such as "Hypoglycemia risk detected, heart rate increased, please replenish sugar immediately") and converts it into an audio encoding format that the device can recognize (such as a pulse code modulation signal adapted to the speaker). Subsequently, the tactile vibration pattern code and the voice text code are encapsulated into a unified local feedback command frame. The frame structure includes a command type identifier ("local feedback"), priority ("urgent"), and data check bits to ensure error-free command transmission.

[0087] For remote early warning information packets: the device organizes the extracted exercise time data, corrected blood glucose sequences, and heart rate acceleration markers into a preset format (such as JSON format). The exercise data is presented as time-step frequency key-value pairs, the blood glucose sequences are arranged in array form, and the heart rate markers include event timestamps and heart rate curve segments. Simultaneously, the information packet is encrypted (e.g., using the AES algorithm) and a unique device identifier and user ID are added, encapsulating it into a remote early warning information packet. The packet structure includes data type, encryption identifier, and timestamp to ensure the security and traceability of information transmission.

[0088] In a specific embodiment, the vibration code of the local feedback command frame is "0x03050202" (corresponding to 3 rapid vibrations + 1 long vibration), and the voice code is an audio signal of "Detected risk of hypoglycemia..."; the remote information packet is encrypted and contains content such as "exercise period: 16:18-16:23, step frequency 120-130 steps / minute; blood glucose sequence: [3.8, 3.6, 3.5]; heart rate event: 16:23:10, from 80 to 100", so as to achieve accurate generation of local and remote information.

[0089] Step 5.3, based on the generated local feedback instruction frame as input, aims to deliver warning information to the user in real time through the smart wearable device. The smart wearable device calls its internal low-latency message queue, which adopts a first-in-first-out mechanism and assigns the highest priority to emergency instructions to avoid being blocked by other low-priority messages (such as regular data synchronization), ensuring that the latency from instruction generation to push is controlled within 100 milliseconds.

[0090] Specifically, after receiving a local feedback instruction frame, the message queue immediately pushes it to the wearable device via the device's internal bus (such as the I2C bus). Specifically, the tactile vibration pattern is encoded and sent to the vibration motor drive unit, triggering the motor to operate according to the encoded rules (such as vibrating at a preset frequency and duration); the voice text is encoded and sent to the audio processing unit, where it is converted from digital to analog and then played as a voice prompt through a speaker. This process enables users to instantly perceive risks and prompts them to quickly take energy replenishment measures.

[0091] In a specific embodiment, after receiving the instruction frame, the low-latency message queue completes the push to the vibration motor and speaker within 50 milliseconds: the vibration motor starts in the mode of "3 rapid vibrations + 1 long vibration", and at the same time the speaker broadcasts "low blood sugar risk detected, heart rate increased, please replenish sugar immediately", so that the user can obtain warning information through both tactile and auditory perception.

[0092] Step 5.4: The early warning information is simultaneously pushed to a remote management platform (such as a hospital monitoring system or a family member terminal) to achieve multi-party collaborative response. The smart wearable device calls a preset secure API interface, which integrates identity authentication (such as device certificate verification), data encryption (such as TLS transport layer encryption), and access control mechanisms to ensure that information is only received by authorized platforms.

[0093] Specifically, the device uploads the generated remote early warning information packet via a secure API interface, attaching a composite early warning trigger timestamp (e.g., 16:23:15.300) to the packet header. This timestamp, accurate to the millisecond, is used by the remote platform to synchronize the event occurrence time. Upon receiving the information packet, the remote management platform automatically decrypts and parses the data, displaying early warning information (such as user ID, risk type, and physiological data curves) on the monitoring interface, and can trigger secondary notifications (e.g., sending SMS messages to family members). This process enables the remote management party to be aware of risk events in real time, providing data support for remote intervention.

[0094] In a specific embodiment, after authentication, the secure API interface transmits the encrypted remote early warning information packet to the hospital management platform. The timestamp attached to the information packet is 16:23:15.300. After receiving the packet, the platform immediately displays "User XXX triggered a low blood sugar + sudden increase in heart rate early warning at 16:23:15" on the monitoring screen and simultaneously displays the exercise, blood sugar, and heart rate data curves. At the same time, it pushes reminders to the on-duty doctor's terminal to achieve remote real-time monitoring.

[0095] like Figure 2 As shown, embodiments of the present invention also provide a smart wearable health monitoring system, comprising: The acquisition module is used to collect raw sensor data streams from the triaxial accelerometer and gyroscope in real time, and convert the raw sensor data streams to the human skeletal coordinate system through a rotation matrix; calculate the Euler angles of the limb joints based on the converted data, and extract gait phase parameters by combining the curvature features of the motion trajectory; calculate the real-time step frequency based on the gait phase parameters; when multiple consecutive step frequency sampling values ​​exceed the preset medium-high intensity exercise threshold, a medium-high intensity exercise state identifier is generated. The calculation module is used to synchronously execute according to the medium-to-high intensity exercise status indicator. Based on the continuous sampling sequence of the blood glucose sensor, it constructs a dynamic blood glucose assessment window, calculates the temporal gradient of blood glucose readings within the window, and generates a dynamic compensation coefficient based on the temporal gradient. The dynamic compensation coefficient is weighted and fused with a preset baseline blood glucose compensation value to output an adaptive blood glucose compensation amount. The adaptive blood glucose compensation amount is superimposed on the current original blood glucose sampling value to generate a corrected blood glucose value. The first-order difference value of adjacent heart rate sampling points is calculated in real time. When the first-order difference value exceeds the preset heart rate change rate threshold, a heart rate acceleration event indicator is generated. The processing module is used to make a combined judgment based on the corrected blood glucose value and the heart rate surge event flag. When the corrected blood glucose value is lower than the hypoglycemia warning threshold and the heart rate surge event flag is present, a combined warning is triggered. Based on the combined warning, the module outputs an energy replenishment command through tactile and voice feedback devices and pushes a warning message containing exercise time data, corrected blood glucose sequence and heart rate surge flag to the remote management platform.

[0096] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A health monitoring method for smart wearable devices, characterized in that, The method includes: Step 1: Real-time acquisition of raw sensor data streams from the three-axis accelerometer and gyroscope; conversion of the raw sensor data streams to the human skeletal coordinate system using a rotation matrix; calculation of Euler angles of limb joints based on the converted data; extraction of gait phase parameters by combining the curvature features of the motion trajectory; calculation of real-time gait frequency based on gait phase parameters; generation of a medium-to-high intensity exercise status identifier when multiple consecutive gait frequency sampling values ​​exceed a preset medium-to-high intensity exercise threshold. Step 2: Synchronously execute according to the medium-to-high intensity exercise state indicator, construct a dynamic blood glucose assessment window based on the continuous sampling sequence of the blood glucose sensor, calculate the temporal gradient of blood glucose readings within the window, and generate dynamic compensation coefficients based on the temporal gradient. Step 3: The dynamic compensation coefficient is weighted and fused with the preset baseline blood glucose compensation value to output the adaptive blood glucose compensation amount. The adaptive blood glucose compensation amount is superimposed on the current original blood glucose sampling value to generate the corrected blood glucose value. The first-order difference value of adjacent heart rate sampling points is calculated in real time. When the first-order difference value exceeds the preset heart rate change threshold, a heart rate acceleration event identifier is generated. Step 4: Make a combined judgment based on the corrected blood glucose value and the heart rate surge event indicator. When the corrected blood glucose value is lower than the hypoglycemia warning threshold and the heart rate surge event indicator is present, a combined warning is triggered. Step 5: Based on the composite warning, output the energy replenishment command through the tactile and voice feedback device, and push the warning information containing exercise time data, corrected blood glucose sequence and heart rate spike marker to the remote management platform.

2. The health monitoring method for smart wearable devices according to claim 1, characterized in that, Real-time acquisition of raw sensor data streams from triaxial accelerometers and gyroscopes, followed by transformation of the raw sensor data streams to the human skeletal coordinate system using a rotation matrix, including: Step 1.1: Receive the raw sensor data stream from the three-axis accelerometer and gyroscope uploaded in real time by the smart wearable device, perform timestamp alignment processing on the raw sensor data stream, and generate a synchronous sensor data sequence. Step 1.2: Load the pre-stored personalized rotation matrix parameters, perform coordinate transformation operations based on the synchronous sensing data sequence, and generate three-dimensional motion vectors of the hip and knee joints in the skeletal coordinate system. Step 1.3: Based on the three-dimensional motion vectors of the hip and knee joints, calculate the Euler angles of flexion and extension in the sagittal plane and the Euler angles of adduction and abduction in the coronal plane for the lower limb joints.

3. The health monitoring method for smart wearable devices according to claim 2, characterized in that, Euler angles of limb joints are calculated based on the transformed data, and gait phase parameters are extracted by combining the curvature features of the motion trajectory. Calculate the real-time step frequency based on gait phase parameters; When multiple consecutive step frequency sampling values ​​exceed a preset medium-to-high intensity exercise threshold, a medium-to-high intensity exercise status identifier is generated, including: Step 1.4: Calculate the instantaneous curvature characteristics of the lower limb movement trajectory based on the Euler angles of flexion and extension of the hip and knee joints in the sagittal plane and the Euler angles of adduction and abduction in the coronal plane. Step 1.5: Extract key phase points of the gait cycle based on the extreme points of curvature characteristics, and generate phase segmentation parameters for the gait support phase and swing phase; Step 1.6: Calculate the single-step gait cycle based on the time interval of adjacent phase segmentation parameters, and generate the real-time step frequency based on the number of gait cycles per unit time. Step 1.7: When multiple consecutive step frequency sampling values ​​exceed the preset medium-high intensity exercise threshold, a medium-high intensity exercise status identifier is generated.

4. The health monitoring method for smart wearable devices according to claim 3, characterized in that, Synchronized with the medium-to-high intensity exercise status indicator, a dynamic blood glucose assessment window is constructed based on the continuous sampling sequence of the blood glucose sensor. The temporal gradient of blood glucose readings within the window is calculated, and dynamic compensation coefficients are generated based on the temporal gradient, including: Step 2.1: Based on the generation of medium-to-high intensity exercise state identifiers, start the real-time streaming data reception of the blood glucose sensor and output a continuous blood glucose sampling sequence with timestamp alignment. Step 2.2: Based on the continuous blood glucose sampling sequence, starting from the moment the motion state identifier is generated, the time span of the evaluation window is dynamically expanded according to the duration of the identifier to generate a dynamic blood glucose evaluation window dataset. Step 2.3: Perform time-series linear regression analysis on the dynamic blood glucose assessment window dataset and extract the regression slope as the time-series gradient value of blood glucose changes; Step 2.4: Input the time-series gradient values ​​into the preset gradient-compensation coefficient mapping table and output the dynamic compensation coefficients.

5. A health monitoring method for smart wearable devices according to claim 4, characterized in that, The dynamic compensation coefficient is weighted and fused with a preset baseline blood glucose compensation value to output an adaptive blood glucose compensation amount. This adaptive blood glucose compensation amount is then added to the current raw blood glucose sample value to generate a corrected blood glucose value. The first-order difference between adjacent heart rate sampling points is calculated in real time. When the first-order difference exceeds a preset heart rate change threshold, a heart rate acceleration event identifier is generated, including: Step 3.1: Call the dynamic compensation coefficient and perform a weighted fusion calculation with the pre-stored baseline blood glucose compensation value through a preset weight allocation strategy to generate an adaptive blood glucose compensation amount; Step 3.2: Obtain the current raw sample value uploaded by the blood glucose sensor in real time, and calculate the adaptive blood glucose compensation amount by positively superimposing it with the raw sample value to generate the exercise state corrected blood glucose value; Step 3.3, in parallel with the blood glucose processing flow, performs time-series difference calculation on the real-time data stream of the heart rate sensor, extracts continuous heart rate readings at fixed sampling intervals, and calculates the absolute value of the difference between two adjacent heart rate readings as the first-order difference value. Step 3.4: Compare the first-order difference value with the preset heart rate change threshold in real time. When multiple consecutive first-order difference values ​​exceed the threshold, generate a heart rate acceleration event identifier.

6. The health monitoring method for smart wearable devices according to claim 5, characterized in that, A combined alert is triggered based on a modified blood glucose level and a sudden increase in heart rate event flag. The alert includes the following: When the modified blood glucose level falls below the hypoglycemia warning threshold and a sudden increase in heart rate event flag is present, a combined alert is activated. Step 4.1: Obtain the exercise-corrected blood glucose value, compare this value with the preset low blood glucose warning threshold, and output a judgment status indicator that the blood glucose is lower than the threshold. Step 4.2: Read the heart rate acceleration event identifier and its timestamp, verify whether the identifier is within the preset validity period based on the current system time, and output the identifier activation status verification result; Step 4.3: Receive the judgment status identifier and the identifier activation status verification result, and perform logical AND operation to generate a composite early warning trigger command.

7. A health monitoring method for smart wearable devices according to claim 6, characterized in that, Based on the composite warning, an energy replenishment command is output via tactile and voice feedback devices, and a warning message containing exercise time data, corrected blood glucose sequence, and heart rate spike markers is pushed to the remote management platform, including: Step 5.1: Based on the composite warning, extract the exercise period data, correct the blood glucose sequence and heart rate acceleration markers; Step 5.2: Generate dual-channel instructions based on the extracted warning data, and generate local feedback instruction frames with tactile vibration pattern encoding and voice text encoding; encapsulate a remote warning information package containing exercise period data, corrected blood glucose sequence and heart rate acceleration markers; Step 5.3: Push local feedback command frames to the smart wearable device through a low-latency message queue to trigger the device to perform tactile vibration and voice broadcast; Step 5.4: Transmit the remote early warning information packet to the remote management platform through the secure API interface, and attach a composite early warning trigger timestamp.

8. A smart wearable health monitoring system, wherein the system implements the method as described in any one of claims 1 to 7, characterized in that, include: The acquisition module is used to collect raw sensor data streams from the triaxial accelerometer and gyroscope in real time, and convert the raw sensor data streams to the human skeleton coordinate system through a rotation matrix; Euler angles of limb joints are calculated based on the transformed data, and gait phase parameters are extracted by combining the curvature features of the motion trajectory; real-time gait frequency is calculated based on the gait phase parameters. When multiple consecutive step frequency sampling values ​​exceed the preset medium-to-high intensity exercise threshold, a medium-to-high intensity exercise status identifier is generated; The calculation module is used to synchronously execute according to the medium-to-high intensity exercise status indicator. Based on the continuous sampling sequence of the blood glucose sensor, it constructs a dynamic blood glucose assessment window, calculates the temporal gradient of blood glucose readings within the window, and generates a dynamic compensation coefficient based on the temporal gradient. The dynamic compensation coefficient is weighted and fused with a preset baseline blood glucose compensation value to output an adaptive blood glucose compensation amount. The adaptive blood glucose compensation amount is superimposed on the current original blood glucose sampling value to generate a corrected blood glucose value. The first-order difference value of adjacent heart rate sampling points is calculated in real time. When the first-order difference value exceeds the preset heart rate change rate threshold, a heart rate acceleration event indicator is generated. The processing module is used to make a combined judgment based on the corrected blood glucose value and the heart rate surge event flag. When the corrected blood glucose value is lower than the hypoglycemia warning threshold and the heart rate surge event flag is present, a combined warning is triggered. Based on the composite warning, the system outputs an energy replenishment command through tactile and voice feedback devices, and pushes warning information including exercise time data, corrected blood glucose sequence, and heart rate spike markers to the remote management platform.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images