High-performance bracelet based on physical information detection

Abstract

The application discloses a high-performance bracelet based on physical information detection, which comprises a physiological signal detection module, a core control module, a display module and a circuit module; the physiological signal detection module collects real-time rotation angle of a user's arm, pulse signals, blood oxygen signals, blood vessel contraction pressure and external pressure changes, and after being processed by a preprocessing module, the core control module judges the user's state. The bracelet can accurately measure the blood pressure and posture of the user, realizes the key attention to the user's state based on the blood pressure and the total cost of the best pulse signal, and can realize the accurate detection of the abnormal state of the user through the synchronous training based on the convolution network parameters and the cost threshold.

Description

Technical Field

[0001] This invention relates to the field of wristband design technology, and in particular to a high-performance wristband based on vital sign information detection. Background Technology

[0002] The use of smart bracelets to monitor users' health status is necessary. Bracelets can overcome the shortcomings of traditional health monitoring methods (such as regular physical examinations and home medical devices) by being static and lagging behind, achieving continuous, 24 / 7 monitoring without being noticed. For drivers, health monitoring based on bracelets is even more crucial for safety. A significant number of road traffic accidents are related to driver health problems (such as arrhythmia, abnormal blood pressure) or fatigue. By monitoring indicators such as pulse, blood oxygen, and blood pressure in real time, bracelets can provide early warnings of asymptomatic health crises that may occur while driving, such as transient myocardial ischemia and acute hypoxia, thus becoming a proactive protective tool for driving safety.

[0003] Currently, existing wristbands primarily employ two methods for measuring user blood pressure. While cuff-based solutions achieve medical-grade accuracy, they require complex structures integrating miniature air pumps, multi-layered airbags, and high-precision differential pressure sensors, resulting in complex equipment that impacts wearing comfort and daily activities. Cuffless solutions based on PPG photoelectric or PPG+ECG dual-mode methods offer greater convenience, but due to the delicate nature of wrist blood vessels and weak blood flow signals, they are easily affected by factors such as wearing tightness, skin color, and activity level, making accurate blood pressure measurement a core challenge for current technology. Regarding user status assessment, existing wristbands lack sufficient intelligence and real-time performance: most rely on a single indicator for judgment, lacking multi-parameter cross-validation; they also lack continuous, all-time tracking; and their algorithm models suffer from weak recognition capabilities for complex user states due to insufficient training data. Summary of the Invention

[0004] To achieve accurate and real-time monitoring of users' health status, this invention provides a high-performance wristband based on vital sign information detection, which can more conveniently realize accurate measurement and real-time monitoring of users' blood pressure; at the same time, the invented wristband is based on multiple physiological parameters and adopts a two-step discrimination method, which makes the identification of abnormal user conditions more accurate.

[0005] This invention proposes a high-performance wristband based on vital sign information detection, comprising a physiological signal detection module, a core control module, a display module, and a power module;

[0006] The physiological signal detection module consists of a triaxial accelerometer, a photoelectric sensor, a pressure sensor, and a preprocessing module. The preprocessing module is connected to the triaxial accelerometer, photoelectric sensor, and pressure sensor in a star topology. The triaxial accelerometer and the preprocessing module are installed inside the wristband below the display module, while the photoelectric sensor and pressure sensor are distributed and installed inside the watchband.

[0007] The system includes a triaxial accelerometer to acquire the user's real-time arm rotation angle, a photoelectric sensor to acquire pulse and blood oxygenation signals, and a pressure sensor to acquire the user's vasoconstriction pressure and external pressure changes. In the preprocessing module, the user's systolic blood pressure is obtained through preprocessing. The optimal total cost of the pulse signal is calculated based on the reference pulse signal to obtain the user's diastolic blood pressure and the pulse signal segment corresponding to the diastolic pressure time. The total cost of the blood oxygenation signal segment for each cycle is calculated and arranged by cycle to obtain the blood oxygenation cost time series. The user's posture sequence is obtained through preprocessing.

[0008] The core control module includes a main controller and a coprocessor. The main controller is responsible for power management, display management, and user interaction, while the coprocessor is responsible for physiological feature fusion and user status recognition. The core control module is installed below the display module and is connected to the preprocessing module in the physiological signal monitoring module. The core control module initially judges the user status based on the total cost values ​​of systolic blood pressure, diastolic blood pressure, and the optimal pulse signal. If the user status is judged to be a key monitoring status, a trained convolutional network is used to determine whether the user is in an abnormal state.

[0009] The display module has a dynamic brightness adjustment function, which reduces the power consumption of the screen when the ambient light intensity is lower than a preset threshold, and automatically turns off the screen when there is no operation; the display module is installed on the outside of the wristband and connected to the core control module;

[0010] The power module supports stable voltage output.

[0011] Furthermore, the triaxial accelerometer collects the real-time rotation angle of the user's arm, specifically:

[0012] The triaxial accelerometer is attached to the user's arm. An X-axis is established along the length of the forearm, a Y-axis is established along the direction perpendicular to the palm plane, and a Z-axis is established along the direction parallel to the palm plane and perpendicular to the X-axis.

[0013] A three-axis accelerometer collects the real-time velocity of the user's arm in the X, Y, and Z directions;

[0014] Based on the real-time velocities of the user's arm in the X, Y, and Z directions, the real-time distance of movement of the user's arm in the X, Y, and Z directions is obtained by integration.

[0015] Based on the real-time movement distance of the user's arm in the X, Y, and Z directions and the length of the user's arm, the real-time rotation angle of the user's arm in the X, Y, and Z directions is obtained.

[0016] Furthermore, the photoelectric sensor acquires pulse signals and blood oxygen signals, including:

[0017] The photoelectric sensor uses infrared light to illuminate the skin on the inside of the arm, and uses a PIN diode to convert the reflected light signal into a current signal. A transimpedance amplifier and an instrumentation amplifier are then used to amplify the current signal and convert it into a voltage signal.

[0018] The photoelectric sensor passes the voltage signal through a high-pass filter and a low-pass filter respectively to obtain the pulse signal;

[0019] The photoelectric sensor subtracts the voltage signal from the standard voltage to obtain the blood oxygen signal;

[0020] The photoelectric sensor inputs the pulse signal and blood oxygen signal into the preprocessing module.

[0021] Furthermore, the pressure sensor acquires the user's vascular systolic pressure and changes in external pressure, including:

[0022] The pressure sensor uses a MEMS piezoresistive chip to apply external pressure to the skin at the artery on the inside of the arm. As the external pressure decreases, the blood vessel changes from a collapsed state to a state where blood flow begins. The external pressure value at this time is recorded as the systolic pressure.

[0023] The external pressure continues to decrease and the changes in external pressure are recorded; the pressure sensor outputs the contraction pressure and external pressure change data to the preprocessing module.

[0024] Further, in the preprocessing module, the user's systolic blood pressure is obtained through preprocessing; the optimal total cost of the pulse signal is calculated based on the reference pulse signal to obtain the user's diastolic blood pressure and the pulse signal segment corresponding to the diastolic blood pressure time; the total cost of the blood oxygen signal segment for each cycle is calculated and arranged by cycle to obtain the blood oxygen cost time series; and the user's posture sequence is obtained through preprocessing, including:

[0025] The preprocessing module converts the received systolic pressure into the user's systolic blood pressure based on a conversion function;

[0026] The preprocessing module receives external pressure change data and simultaneously records the pulse signal output by the photoelectric sensor; it aligns the external pressure change data and the pulse signal in time; it divides the pulse signal by period to obtain pulse signal segments and records the start time of each pulse signal segment.

[0027] Significant feature points are extracted from the current pulse signal segment and the reference pulse signal, respectively, to construct the pulse signal segment feature sequence and the reference pulse signal feature sequence; for each pair of feature points in the pulse signal segment feature sequence and the reference pulse signal feature sequence, the cost function of the pulse signal segment is calculated based on the timestamp and amplitude.

[0028] Based on the cost function of the pulse signal segment, a cost function matrix of the pulse signal segment is constructed; the path with the minimum total cost from the starting point to the ending point of the cost function matrix of the pulse signal segment is found, and the cost functions on the path with the minimum total cost are summed to obtain the total cost of the pulse signal segment; the starting point and the ending point of the cost function matrix of the pulse signal segment are the first row and first column elements and the last row and last column elements of the cost function matrix of the pulse signal segment, respectively.

[0029] The preprocessing module calculates the total cost function value of the pulse signal segment in the current cycle. The total cost function value compared to the previous pulse signal segment If a comparison is made, ≦ The total cost of updating the optimal pulse signal is then... And record the start time of the current cycle signal segment as the diastolic blood pressure time, and the pulse signal segment of the current cycle as the pulse signal segment corresponding to the diastolic blood pressure time; if > The total cost of maintaining the optimal pulse signal is The starting time of the previous cycle signal segment is kept as the diastolic pressure time, and the pulse signal segment of the previous cycle is the pulse signal segment corresponding to the diastolic pressure time. Based on the diastolic pressure time, the external pressure value corresponding to the time in the external pressure change data is found and recorded as the diastolic pressure. Based on the conversion function, the diastolic pressure is converted into the user's diastolic blood pressure.

[0030] The preprocessing module divides the blood oxygen signal output by the photoelectric sensor into blood oxygen signal segments according to the period and records the start time of each blood oxygen signal segment; it extracts significant feature points from the current blood oxygen signal segment and the reference blood oxygen signal respectively, and constructs the blood oxygen signal segment feature sequence and the reference blood oxygen signal feature sequence; for each pair of feature points in the blood oxygen signal segment feature sequence and the reference blood oxygen signal feature sequence, it calculates the cost function of the blood oxygen signal segment based on the timestamp and amplitude.

[0031] Based on the cost function of the blood oxygen signal segment, a cost function matrix of the blood oxygen signal segment is constructed; the path with the minimum total cost from the starting point to the ending point of the cost function matrix of the blood oxygen signal segment is found, and the cost functions on the path with the minimum total cost are summed to obtain the total cost of the blood oxygen signal segment; the starting point and the ending point of the cost function matrix of the blood oxygen signal segment are the first row and first column elements and the last row and last column elements of the cost function matrix of the blood oxygen signal segment, respectively.

[0032] The preprocessing module associates the total cost of each blood oxygen signal segment with the start timestamp of the corresponding period in order from earliest to latest according to the start timestamp of the blood oxygen signal cycle, and arranges them to form a blood oxygen cost time series with the blood oxygen signal cycle as the time dimension.

[0033] The preprocessing module constructs a user posture sequence based on the real-time rotation angles of the user's arm in the X, Y, and Z directions from the triaxial accelerometer.

[0034] The preprocessing module outputs the total cost values ​​of systolic blood pressure, diastolic blood pressure, and optimal pulse signal, as well as the blood oxygen cost time series, the pulse signal segment corresponding to the diastolic blood pressure time, and the user posture sequence to the core control module.

[0035] Furthermore, the core control module initially determines the user's status based on the total cost values ​​of systolic blood pressure, diastolic blood pressure, and the optimal pulse signal; if the user's status is determined to be a key monitoring status, a trained convolutional network is used to determine whether the user is in an abnormal state.

[0036] The core control module sets the normal threshold range for systolic and diastolic blood pressure, and sets the cost threshold for the optimal total cost value of the pulse signal.

[0037] When systolic and diastolic blood pressure are outside the normal threshold range, and the total cost of the best pulse signal exceeds the cost threshold, the core control module initially determines that the user's status is a key monitoring status.

[0038] If the user is in a high-priority monitoring state, the blood oxygen cost time series, the pulse signal segment corresponding to the diastolic blood pressure time, and the user posture sequence are fused together.

[0039] Based on the fusion features, a pre-trained convolutional network is used to determine the user's state; when the user is determined to be in an abnormal state, it is reported.

[0040] Furthermore, the trained convolutional network in the core control module includes:

[0041] Optimize the parameters of the convolutional network based on the training data until the detection rate no longer improves, then stop training the convolutional network;

[0042] Assume the cost threshold of the core control module is ,exist Within the range according to step size Perform a traversal; for each cost threshold, compare the F1 score of the convolutional network; update the cost threshold of the core control module based on the cost threshold with the highest F1 score. ;

[0043] Fine-tune the parameters of the convolutional network and retrain the trained convolutional network until the detection rate no longer improves, then stop training the convolutional network.

[0044] Set detection thresholds and false alarm thresholds;

[0045] Repeatedly optimize the cost threshold using fixed convolutional network parameters Then fix the cost threshold Optimizing convolutional network parameters involves setting a cost threshold for the core control module. The parameters of the convolutional network are adjusted; training is considered complete when the detection rate increase is less than the detection threshold and the false alarm rate decrease is less than the false alarm threshold in two consecutive training rounds; the current cost threshold. The current convolutional network parameters are the optimal convolutional network parameters, representing the optimal cost threshold.

[0046] Set the optimal cost threshold as the cost threshold of the core control module; use the convolutional network with the best convolutional network parameters as the trained convolutional network.

[0047] Beneficial effects:

[0048] The wristband provided by this invention uses a three-axis accelerometer, photoelectric sensor, and pressure sensor to achieve real-time acquisition of user vital signs information, facilitating real-time monitoring and early warning of the user's health status. Data processing based on the preprocessing module helps obtain more accurate user blood pressure data, and a cost function enables scientific assessment of the user's status. The preprocessing module's periodic processing of pulse and blood oxygen signals helps to compare vital signs data under stable conditions with healthy vital signs data. The core control module quantitatively measures the difference between user vital signs data and healthy vital signs data through a cost threshold, achieving preliminary screening of the user's status. The core control module's synchronous training based on convolutional network parameters and cost thresholds helps to improve the detection rate of abnormal user states and reduce the false alarm rate. Attached Figure Description

[0049] Figure 1 This is a schematic diagram of a high-performance wristband structure based on vital sign information detection, provided as an embodiment of the present invention. Detailed Implementation

[0050] Example 1:

[0051] A high-performance wristband based on vital sign information detection, such as Figure 1 As shown, it includes a physiological signal detection module, a core control module, a display module, and a power supply module; the physiological signal detection module consists of a triaxial accelerometer, a photoelectric sensor, a pressure sensor, and a preprocessing module; the preprocessing module is connected to the triaxial accelerometer, photoelectric sensor, and pressure sensor in a star topology; the triaxial accelerometer and the preprocessing module are installed on the inner side of the wristband below the display module, and the photoelectric sensor and pressure sensor are distributed and installed on the inner side of the strap;

[0052] Among them, the triaxial accelerometer collects the real-time rotation angle of the user's arm, the photoelectric sensor collects pulse signals and blood oxygen signals, and the pressure sensor collects the user's vascular systolic pressure and changes in external pressure.

[0053] In the preprocessing module, the user's systolic blood pressure is obtained through preprocessing. The optimal total cost of the pulse signal is calculated based on the reference pulse signal to obtain the user's diastolic blood pressure and the pulse signal segment corresponding to the diastolic blood pressure time. The total cost of the blood oxygen signal segment for each cycle is calculated and arranged by cycle to obtain the blood oxygen cost time series. The user's posture sequence is obtained through preprocessing.

[0054] The core control module includes a main controller and a coprocessor. The main controller is responsible for power management, display management, and user interaction, while the coprocessor is responsible for physiological feature fusion and user status recognition. The core control module is installed below the display module and is connected to the preprocessing module in the physiological signal monitoring module. The core control module initially judges the user status based on the total cost values ​​of systolic blood pressure, diastolic blood pressure, and the optimal pulse signal. If the user status is judged to be a key monitoring status, a trained convolutional network is used to determine whether the user is in an abnormal state.

[0055] The display module has a dynamic brightness adjustment function, which reduces the power consumption of the screen when the ambient light intensity is lower than a preset threshold, and automatically turns off the screen when there is no operation; the display module is installed on the outside of the wristband and connected to the core control module;

[0056] The power module supports stable voltage output.

[0057] The triaxial accelerometer collects the real-time rotation angle of the user's arm; including:

[0058] The triaxial accelerometer is attached to the user's arm. An X-axis is established along the length of the forearm, that is, from the elbow joint to the wrist. A Y-axis is established along the direction perpendicular to the palm plane. A Z-axis is established along the direction parallel to the palm plane and perpendicular to the X-axis.

[0059] The triaxial accelerometer collects the real-time speed of the user's arm in the X, Y, and Z directions;

[0060] Based on the real-time velocities of the user's arm in the X, Y, and Z directions, the real-time distance of movement of the user's arm in the X, Y, and Z directions is obtained by integration.

[0061] Based on the real-time movement distance of the user's arm in the X, Y, and Z directions and the length of the user's arm, the real-time rotation angle of the user's arm in the X, Y, and Z directions is obtained.

[0062] Specifically, in this embodiment of the invention, a low-drift, high-resolution triaxial accelerometer is selected, requiring a zero-drift of less than or equal to 5 milliseconds. In the triaxial accelerometer calibration state, the user wears a wristband and remains stationary while measuring the zero-drift reference value of the triaxial accelerometer. In the triaxial accelerometer acquisition state, the obtained acceleration data needs to be subtracted from the zero-drift reference value to obtain the acquired data of the triaxial accelerometer. The triaxial accelerometer has an algorithm encapsulated inside, which directly outputs the posture of the user's arm, and can obtain the real-time velocity of the user's arm in the X, Y, and Z directions.

[0063] The photoelectric sensor collects pulse signals and blood oxygen signals; including:

[0064] The photoelectric sensor uses infrared light to illuminate the skin on the inside of the arm, and uses a PIN-type optical diode to convert the reflected light signal into a current signal. A transimpedance amplifier and an instrumentation amplifier are used to amplify the current signal and convert it into a voltage signal.

[0065] The photoelectric sensor passes the voltage signal through a high-pass filter and a low-pass filter respectively to obtain the pulse signal;

[0066] The photoelectric sensor subtracts the voltage signal from the standard voltage to obtain the blood oxygen signal;

[0067] The photoelectric sensor inputs the pulse signal and blood oxygen signal into the preprocessing module.

[0068] The pressure sensor collects the user's vascular systolic pressure and changes in external pressure; including:

[0069] The pressure sensor uses a MEMS piezoresistive chip to apply external pressure to the skin at the artery on the inside of the arm. As the external pressure decreases, the blood vessel changes from a collapsed state to a state where blood flow begins. The external pressure value at this time is recorded as the systolic pressure.

[0070] The external pressure continues to decrease and the changes in external pressure are recorded; the pressure sensor outputs the contraction pressure and external pressure change data to the preprocessing module.

[0071] The preprocessing module obtains the user's systolic blood pressure through preprocessing, calculates the optimal total cost of the pulse signal based on the reference pulse signal, and obtains the pulse signal segments corresponding to the user's diastolic blood pressure and diastolic time; calculates the total cost of the blood oxygen signal segments for each cycle and arranges them by cycle to obtain the blood oxygen cost time series; and obtains the user's posture sequence through preprocessing.

[0072] The preprocessing module converts the received systolic pressure into the user's systolic blood pressure based on a conversion function;

[0073] While receiving data on changes in external pressure, the preprocessing module simultaneously records the pulse signal output by the photoelectric sensor.

[0074] The preprocessing module aligns the external pressure change data and pulse signal in time;

[0075] The preprocessing module divides the pulse signal into segments according to the period, and records the start time of each pulse signal segment;

[0076] The preprocessing module extracts significant feature points from the current pulse signal segment and the reference pulse signal, respectively, and constructs the pulse signal segment feature sequence and the reference pulse signal feature sequence.

[0077] For each pair of feature points in the pulse signal segment feature sequence and the reference pulse signal feature sequence, calculate the cost function of the pulse signal segment based on the timestamp and amplitude;

[0078] Based on the cost function of the pulse signal segment, construct the cost function matrix of the pulse signal segment;

[0079] Find the path with the minimum total cost from the starting point (first row, first column element) to the ending point (last row, last column element) of the cost function matrix of the pulse signal segment, sum the cost functions along the path, and use the sum as the total cost of the pulse signal segment;

[0080] The preprocessing module calculates the total cost function value of the pulse signal segment in the current cycle. The total cost function value compared to the previous pulse signal segment If a comparison is made, ≦ The total cost of updating the optimal pulse signal is then... And record the start time of the current cycle signal segment as the diastolic blood pressure time, and the pulse signal segment of the current cycle as the pulse signal segment corresponding to the diastolic blood pressure time; if > The total cost of maintaining the optimal pulse signal is And keep the start time of the previous cycle signal segment as the diastolic pressure time, and the pulse signal segment of the previous cycle as the pulse signal segment corresponding to the diastolic pressure time;

[0081] Based on the diastolic pressure time, find the corresponding external pressure value in the external pressure change data and record it as diastolic pressure; based on the conversion function, convert the diastolic pressure into the user's diastolic blood pressure;

[0082] The preprocessing module divides the blood oxygen signal output by the photoelectric sensor into segments according to the period, and records the start time of each blood oxygen signal segment;

[0083] The preprocessing module extracts significant feature points from the current blood oxygen signal segment and the reference blood oxygen signal, respectively, and constructs the blood oxygen signal segment feature sequence and the reference blood oxygen signal feature sequence.

[0084] For each pair of feature points in the blood oxygen signal segment feature sequence and the reference blood oxygen signal feature sequence, calculate the cost function of the blood oxygen signal segment based on the timestamp and amplitude;

[0085] Based on the cost function of the blood oxygen signal segment, a cost function matrix for the blood oxygen signal segment is constructed;

[0086] Find the path with the minimum total cost from the starting point (first row, first column element) to the ending point (last row, last column element) of the cost function matrix of the blood oxygen signal segment, sum the cost functions along this path, and use this sum as the total cost of the blood oxygen signal segment.

[0087] The preprocessing module associates the total cost of each blood oxygen signal segment with the time identifier (start timestamp) of the corresponding period one by one according to the natural time sequence of the blood oxygen signal cycle (i.e., the order of the start timestamp of the blood oxygen signal segment from early to late), and arranges them to form a blood oxygen cost time series with the blood oxygen signal cycle as the time dimension.

[0088] The preprocessing module constructs a user posture sequence based on the real-time rotation angles of the user's arm in the X, Y, and Z directions from a triaxial accelerometer.

[0089] The preprocessing module outputs the total cost values ​​of systolic blood pressure, diastolic blood pressure, and optimal pulse signal, as well as the blood oxygen cost time series, the pulse signal segment corresponding to the diastolic blood pressure time, and the user posture sequence to the core control module.

[0090] Specifically, in this embodiment of the invention, an experimental environment with no electromagnetic interference and a temperature of (23±2℃) and humidity of (45%~65%) was selected; the experimental subjects were 100 volunteers covering different blood pressure ranges (normal, high blood pressure, low blood pressure);

[0091] The systolic and diastolic blood pressure values ​​from the pressure sensor and the reference blood pressure values ​​from the medical mercury sphygmomanometer were collected simultaneously from volunteers. Based on all the collected systolic and diastolic blood pressure values, a blood pressure sequence was constructed. Based on all the collected reference blood pressure values, a reference blood pressure sequence was constructed. The blood pressure sequence and the reference blood pressure sequence were fitted using the least squares method to construct a transformation function. The root mean square error of the fitting was ensured to be less than the acceptable blood pressure error. The obtained transformation function was stored in the pressure sensor.

[0092] In this embodiment of the invention, under normal user conditions, the preprocessing module corrects the period of the pulse signal and the blood oxygen signal to determine the period of the pulse signal and the period of the blood oxygen signal, respectively.

[0093] The preprocessing module first corrects the pulse signal period, using the corrected pulse signal period as the initial value of the blood oxygen signal period, and then corrects the blood oxygen signal period.

[0094] Significant feature points of the current pulse signal segment and the reference pulse signal include: the point where the signal rises to its highest point, the point where the signal falls to its lowest point, the starting inflection point where the signal rises rapidly from the baseline, the secondary small peaks that appear near the lowest point, the point where the first derivative of the signal is 0, and the point where the second derivative of the signal is 0.

[0095] Significant characteristic points of the current blood oxygen signal segment and the reference blood oxygen signal include: the point where the signal rises to its highest point, the point where the signal falls to its lowest point, the starting inflection point where the signal rises rapidly from the baseline, the secondary small peaks that appear near the lowest point, the point where the first derivative of the signal is 0, and the point where the second derivative of the signal is 0.

[0096] In this embodiment of the invention, the path with the minimum total cost from the starting point (first row, first column element) to the ending point (last row, last column element) of the cost function matrix of the pulse signal segment is found, and the cost functions along this path are summed to obtain the total cost of the pulse signal segment.

[0097] Based on the cost function matrix of pulse signal segments Construct the cumulative cost matrix of the pulse signal segment. ;

[0098] The first row and first column of the cumulative cost matrix The element in the first row and first column of the cost function matrix equals the pulse signal segment. ;

[0099] The first row of the cumulative cost matrix Column elements ,in This represents the first row of the cumulative cost matrix. Column elements, The first row of the cost function matrix representing the pulse signal segment Column elements;

[0100] Cumulative cost matrix element in row 1 column ,in The cumulative cost matrix represents the first... The element in the first column of the row, The cost function matrix of the pulse signal segment represents the first... The element in the first column of the row;

[0101] For the cumulative cost matrix, the first Line number List element Represented as:

[0102]

[0103] in, express , , The minimum value in, The cumulative cost matrix represents the first... Line number Column elements; ; indicates the cumulative cost matrix of the th Line number Column elements; The cumulative cost matrix represents the first... Line number Column elements, The cost function matrix of the pulse signal segment represents the first... Line number Column elements;

[0104] The value of the last element in the last row and last column of the cumulative cost matrix is ​​the total cost of the pulse signal segment, and the corresponding path is the path with the minimum total cost.

[0105] In this embodiment of the invention, the path with the minimum total cost from the starting point (first row, first column element) to the ending point (last row, last column element) of the cost function matrix of the blood oxygen signal segment is found, and the cost functions along this path are summed to obtain the total cost of the blood oxygen signal segment.

[0106] Based on the cost function matrix of blood oxygen signal segment Construct the cumulative cost matrix of the blood oxygen signal segment. ;

[0107] The first row and first column of the cumulative cost matrix The elements in the first row and first column of the cost function matrix equal to the blood oxygen signal segment ;

[0108] The first row of the cumulative cost matrix Column elements ,in This represents the first row of the cumulative cost matrix. Column elements, The first row of the cost function matrix represents the blood oxygen signal segment. Column elements;

[0109] Cumulative cost matrix element in row 1 column ,in The cumulative cost matrix represents the first... The element in the first column of the row, The cost function matrix representing the blood oxygen signal segment is... The element in the first column of the row;

[0110] For the cumulative cost matrix, the first Line number List element Represented as:

[0111]

[0112] in, express , , The minimum value in; The cumulative cost matrix represents the first... Line number Column elements; The cumulative cost matrix represents the first... Line number Column elements; The cumulative cost matrix represents the first... Line number Column elements, The cost function matrix representing the blood oxygen signal segment is... Line number Column elements;

[0113] The value of the last element in the last row and last column of the cumulative cost matrix is ​​the total cost of the blood oxygen signal segment, and the corresponding path is the path with the minimum total cost.

[0114] The core control module initially determines the user's status based on the total cost values ​​of systolic blood pressure, diastolic blood pressure, and the optimal pulse signal. If the core control module determines that the user's status is a key monitoring status, it uses a trained convolutional network to determine whether the user is in an abnormal state.

[0115] The core control module sets the normal threshold range for systolic and diastolic blood pressure, and sets the cost threshold for the total cost value of the optimal pulse signal.

[0116] When systolic and diastolic blood pressure are outside the normal threshold range, and the total cost of the best pulse signal exceeds the cost threshold, the core control module initially determines that the user's status is a key monitoring status.

[0117] If the user is in a high-priority monitoring state, the blood oxygen cost time series, the pulse signal segment corresponding to the diastolic blood pressure time, and the user posture sequence are fused together.

[0118] Based on the fusion features, a pre-trained convolutional network is used to determine the user's state; when the user is determined to be in an abnormal state, it is reported.

[0119] In this embodiment of the invention, in the core control module, the first, second, and third convolutional layers of the convolutional network are each equipped with three sets of convolutional kernels of different sizes. The convolutional kernels of each layer extract fine-grained, medium-grained, and coarse-grained information, respectively. The outputs of each convolutional layer are concatenated according to the channel dimension and then compressed through a max pooling layer to retain key features. The size of the convolutional kernels of the first, second, and third convolutional layers of the convolutional network is guaranteed to be in a 1:2:4 ratio. The three convolutional layers of the convolutional network are followed by two fully connected layers and a Dropout layer.

[0120] The pre-trained convolutional network in the core control module includes:

[0121] Optimize the parameters of the convolutional network based on the training data until the detection rate no longer improves, then stop training the convolutional network;

[0122] Assume the cost threshold of the core control module is ,exist Within the range according to step size Perform a traversal; for each cost threshold, compare the F1 score of the convolutional network; update the cost threshold of the core control module based on the cost threshold with the highest F1 score. ;

[0123] Fine-tune the parameters of the convolutional network and retrain the trained convolutional network until the detection rate no longer improves, then stop training the convolutional network.

[0124] Set detection thresholds and false alarm thresholds;

[0125] Repeatedly optimize the cost threshold using fixed convolutional network parameters Then fix the cost threshold Optimizing convolutional network parameters involves setting a cost threshold for the core control module. The parameters of the convolutional network are adjusted; training is considered complete when, in two consecutive training rounds, the improvement in detection rate is less than the detection threshold and the decrease in false alarm rate is less than the false alarm threshold; the cost threshold at this point is... The current convolutional network parameters are the optimal convolutional network parameters, representing the optimal cost threshold.

[0126] Set the optimal cost threshold as the cost threshold of the core control module; use the convolutional network with the best convolutional network parameters as the trained convolutional network.

[0127] Specifically, in this embodiment of the invention, we define the case where the convolutional network determines that the user is in an abnormal state, and the user is actually in an abnormal state, as a true positive example; the case where the convolutional network determines that the user is in an abnormal state, and the user is actually not in an abnormal state (is in a normal state), as a false positive example; the case where the convolutional network determines that the user is not in an abnormal state (is in a normal state), and the user is actually in an abnormal state, as a false negative example; and the case where the convolutional network determines that the user is not in an abnormal state (is in a normal state), and the user is actually not in an abnormal state (is in a normal state), as a true negative example.

[0128] The detection rate represents the proportion of samples in which a user is actually in an abnormal state, and which the convolutional network correctly identifies as being in an abnormal state; the false alarm rate represents the proportion of samples in which a user is actually not in an abnormal state (is in a normal state), and which the convolutional network incorrectly identifies as being in an abnormal state.

[0129] Detection rate = Number of true negatives / (Number of true negatives + Number of false negatives);

[0130] False alarm rate = Number of false positives / (Number of false positives + Number of true negatives);

[0131] In this embodiment of the invention, the F1 score of the convolutional network is represented as:

[0132]

[0133] in, Indicates the detection rate. This indicates the false alarm rate.

[0134] It should be noted that the F1 score of a convolutional network is a key indicator for comprehensively evaluating the model's recognition accuracy. The value ranges from 0 to 1, and a higher value indicates that the model has achieved better overall performance in terms of precision and recall. Precision measures the proportion of samples that the model predicts as positive to actually be positive, reflecting the accuracy of recognition. Recall measures the proportion of all true positive samples that are correctly predicted by the model, reflecting the completeness of recognition. In this invention, the F1 score is used as an objective and quantitative technical metric to evaluate the performance of convolutional networks in task processing.

[0135] It should be noted that the sequence numbers of the above embodiments of the present invention are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0136] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0137] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A high-performance wristband based on vital sign information detection, characterized in that, It includes a physiological signal detection module, a core control module, a display module, and a power supply module; The physiological signal detection module consists of a triaxial accelerometer, a photoelectric sensor, a pressure sensor, and a preprocessing module. The preprocessing module is connected to the triaxial accelerometer, photoelectric sensor, and pressure sensor in a star topology. The triaxial accelerometer and the preprocessing module are installed inside the wristband below the display module, while the photoelectric sensor and pressure sensor are distributed and installed inside the watchband. The system comprises a triaxial accelerometer to acquire the user's real-time arm rotation angle, a photoelectric sensor to acquire pulse and blood oxygenation signals, and a pressure sensor to acquire the user's vascular systolic pressure and changes in external pressure. Specifically, the photoelectric sensor uses infrared light to illuminate the inner skin of the arm, converts the reflected light signal into a current signal using a PIN photodiode, and then amplifies the current signal into a voltage signal using a transimpedance amplifier and an instrumentation amplifier. The photoelectric sensor then passes the voltage signal through a high-pass filter and a low-pass filter to obtain the pulse signal. The photoelectric sensor subtracts the voltage signal from the standard voltage to obtain the blood oxygenation signal. Finally, the photoelectric sensor inputs the pulse and blood oxygenation signals into a preprocessing module. In the preprocessing module, the user's systolic blood pressure is obtained through preprocessing. The optimal total cost of the pulse signal is calculated based on the reference pulse signal to obtain the user's diastolic blood pressure and the pulse signal segment corresponding to the diastolic blood pressure time. The total cost of the blood oxygen signal segment for each cycle is calculated and arranged by cycle to obtain the blood oxygen cost time series. The user's posture sequence is obtained through preprocessing. Specifically: The preprocessing module converts the received systolic pressure into the user's systolic blood pressure based on a conversion function; The preprocessing module receives external pressure change data and simultaneously records the pulse signal output by the photoelectric sensor; it aligns the external pressure change data and the pulse signal in time; it divides the pulse signal by period to obtain pulse signal segments and records the start time of each pulse signal segment. Significant feature points are extracted from the current pulse signal segment and the reference pulse signal, respectively, to construct the pulse signal segment feature sequence and the reference pulse signal feature sequence; for each pair of feature points in the pulse signal segment feature sequence and the reference pulse signal feature sequence, the cost function of the pulse signal segment is calculated based on the timestamp and amplitude. Based on the cost function of the pulse signal segment, a cost function matrix of the pulse signal segment is constructed; the path with the minimum total cost from the starting point to the ending point of the cost function matrix of the pulse signal segment is found, and the cost functions on the path with the minimum total cost are summed to obtain the total cost of the pulse signal segment; the starting point and the ending point of the cost function matrix of the pulse signal segment are the first row and first column elements and the last row and last column elements of the cost function matrix of the pulse signal segment, respectively. The preprocessing module calculates the total cost function value of the pulse signal segment in the current cycle. The total cost function value compared to the previous pulse signal segment If a comparison is made, ≦ The total cost of updating the optimal pulse signal is then... And record the start time of the current cycle signal segment as the diastolic blood pressure time, and the pulse signal segment of the current cycle as the pulse signal segment corresponding to the diastolic blood pressure time; if > The total cost of maintaining the optimal pulse signal is The starting time of the previous cycle signal segment is kept as the diastolic pressure time, and the pulse signal segment of the previous cycle is the pulse signal segment corresponding to the diastolic pressure time. Based on the diastolic pressure time, the external pressure value corresponding to the time in the external pressure change data is found and recorded as the diastolic pressure. Based on the conversion function, the diastolic pressure is converted into the user's diastolic blood pressure. The core control module includes a main controller and a coprocessor. The main controller is responsible for power management, display management, and user interaction, while the coprocessor is responsible for physiological feature fusion and user status recognition. The core control module is installed below the display module and is connected to the preprocessing module in the physiological signal detection module. The core control module initially judges the user status based on the total cost values ​​of systolic blood pressure, diastolic blood pressure, and the optimal pulse signal. If the user status is judged to be a key monitoring status, a trained convolutional network is used to determine whether the user is in an abnormal state. The display module has a dynamic brightness adjustment function, which reduces the power consumption of the screen when the ambient light intensity is lower than a preset threshold, and automatically turns off the screen when there is no operation; the display module is installed on the outside of the wristband and connected to the core control module; The power module supports stable voltage output.

2. The high-performance wristband based on vital sign information detection according to claim 1, characterized in that, The triaxial accelerometer collects the real-time rotation angle of the user's arm, including: The triaxial accelerometer is attached to the user's arm. An X-axis is established along the length of the forearm, a Y-axis is established along the direction perpendicular to the palm plane, and a Z-axis is established along the direction parallel to the palm plane and perpendicular to the X-axis. A three-axis accelerometer collects the real-time velocity of the user's arm in the X, Y, and Z directions; Based on the real-time velocities of the user's arm in the X, Y, and Z directions, the real-time distance of movement of the user's arm in the X, Y, and Z directions is obtained by integration. Based on the real-time movement distance of the user's arm in the X, Y, and Z directions and the length of the user's arm, the real-time rotation angle of the user's arm in the X, Y, and Z directions is obtained.

3. The high-performance wristband based on vital sign information detection according to claim 1, characterized in that, The pressure sensor collects the user's vascular systolic pressure and changes in external pressure, including: The pressure sensor uses a MEMS piezoresistive chip to apply external pressure to the skin at the artery on the inside of the arm. As the external pressure decreases, the blood vessel changes from a collapsed state to a state where blood flow begins. The external pressure value at this time is recorded as the systolic pressure. The external pressure continues to decrease and the changes in external pressure are recorded; the pressure sensor outputs the contraction pressure and external pressure change data to the preprocessing module.

4. The high-performance wristband based on vital sign information detection according to claim 1, characterized in that, In the preprocessing module, the user's systolic blood pressure is obtained through preprocessing. The optimal total cost of the pulse signal is calculated based on a reference pulse signal to obtain the user's diastolic blood pressure and the pulse signal segment corresponding to the diastolic blood pressure time. The total cost of the blood oxygen signal segment for each cycle is calculated and arranged by cycle to obtain the blood oxygen cost time series. The user's posture sequence is obtained through preprocessing, specifically: The preprocessing module divides the blood oxygen signal output by the photoelectric sensor into blood oxygen signal segments according to the period and records the start time of each blood oxygen signal segment; it extracts significant feature points from the current blood oxygen signal segment and the reference blood oxygen signal respectively, and constructs the blood oxygen signal segment feature sequence and the reference blood oxygen signal feature sequence; for each pair of feature points in the blood oxygen signal segment feature sequence and the reference blood oxygen signal feature sequence, it calculates the cost function of the blood oxygen signal segment based on the timestamp and amplitude. Based on the cost function of the blood oxygen signal segment, a cost function matrix for the blood oxygen signal segment is constructed; Find the path with the minimum total cost from the starting point to the ending point of the cost function matrix of the blood oxygen signal segment, and sum the cost functions on the path with the minimum total cost to obtain the total cost of the blood oxygen signal segment; the starting point and the ending point of the cost function matrix of the blood oxygen signal segment are the first row and first column elements and the last row and last column elements of the cost function matrix of the blood oxygen signal segment, respectively. The preprocessing module associates the total cost of each blood oxygen signal segment with the start timestamp of the corresponding period in order from earliest to latest according to the start timestamp of the blood oxygen signal cycle, and arranges them to form a blood oxygen cost time series with the blood oxygen signal cycle as the time dimension. The preprocessing module constructs a user posture sequence based on the real-time rotation angles of the user's arm in the X, Y, and Z directions from the triaxial accelerometer. The preprocessing module outputs the total cost values ​​of systolic blood pressure, diastolic blood pressure, and optimal pulse signal, as well as the blood oxygen cost time series, the pulse signal segment corresponding to the diastolic blood pressure time, and the user posture sequence to the core control module.

5. The high-performance wristband based on vital sign information detection according to claim 1, characterized in that, The core control module initially determines the user's status based on the total cost values ​​of systolic blood pressure, diastolic blood pressure, and the optimal pulse signal. If the user's status is determined to be a key monitoring status, a trained convolutional network is used to determine whether the user is in an abnormal state. The core control module sets the normal threshold range for systolic and diastolic blood pressure, and sets the cost threshold for the optimal total cost value of the pulse signal. When systolic and diastolic blood pressure are outside the normal threshold range, and the total cost of the best pulse signal exceeds the cost threshold, the core control module initially determines that the user's status is a key monitoring status. If the user is in a high-priority monitoring state, the blood oxygen cost time series, the pulse signal segment corresponding to the diastolic blood pressure time, and the user posture sequence are fused together. Based on the fusion features, a pre-trained convolutional network is used to determine the user's state; when the user is determined to be in an abnormal state, it is reported.

6. The high-performance wristband based on vital sign information detection according to claim 5, characterized in that, The pre-trained convolutional network in the core control module includes: Optimize the parameters of the convolutional network based on the training data until the detection rate no longer improves, then stop training the convolutional network; Assume the cost threshold of the core control module is ,exist Within the range according to step size Perform a traversal; for each cost threshold, compare the F1 score of the convolutional network; update the cost threshold of the core control module based on the cost threshold with the highest F1 score. ; Fine-tune the parameters of the convolutional network and retrain the trained convolutional network until the detection rate no longer improves, then stop training the convolutional network. Set detection thresholds and false alarm thresholds; Repeatedly optimize the cost threshold using fixed convolutional network parameters Then fix the cost threshold Optimizing convolutional network parameters involves setting a cost threshold for the core control module. The parameters of the convolutional network are adjusted; training is considered complete when the detection rate increase is less than the detection threshold and the false alarm rate decrease is less than the false alarm threshold in two consecutive training rounds; the current cost threshold. The current convolutional network parameters are the optimal convolutional network parameters, representing the optimal cost threshold. Set the optimal cost threshold as the cost threshold of the core control module; use the convolutional network with the best convolutional network parameters as the trained convolutional network.

Images