Training method of blood pressure prediction model and blood pressure data prediction method

Abstract

The invention provides a training method of a blood pressure prediction model and a blood pressure data prediction method.The training method of the blood pressure prediction model comprises the steps that data of wearable equipment is obtained, and an original data sequence is obtained; preprocessing the original data sequence to obtain a standard data sequence; features of the standard pressure signal sequence and the standard blood pressure sequence are extracted based on the peak value and the valley value respectively, and a pressure feature sequence, a blood pressure feature sequence and a pulse wave propagation time sequence are obtained; extracting features of the standard acceleration sequence and the standard angular velocity sequence based on the change rate and the frequency to obtain a motion intensity sequence; fusing the pressure feature sequence, the blood pressure feature sequence, the pulse wave propagation time sequence and the exercise intensity sequence to form a multi-dimensional feature sequence; and training the blood pressure prediction pre-training model based on the multi-dimensional feature sequence to obtain a blood pressure prediction model. The blood pressure prediction model obtained by the method has higher precision when being used for predicting the blood pressure.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of blood pressure measurement, in particular to a training method of a blood pressure prediction model of a wearable device and a blood pressure data prediction method. BACKGROUND

[0002] Blood pressure, as an important indicator of cardiovascular health, its accurate monitoring is of key significance for the prevention and management of hypertension and other cardiovascular diseases. The traditional blood pressure measurement method mainly relies on the cuff device, which measures the systolic and diastolic pressure through the intermittent inflation and deflation process, and has the shortcomings of inconvenient operation, discontinuous measurement and limitation on user activity. With the development of wearable technology, continuous and real-time blood pressure monitoring using wearable devices has become a research hotspot. However, the existing wearable blood pressure monitoring methods still have many challenges in data collection accuracy, real-time processing capability and prediction accuracy.

[0003] In the prior art, a single sensor often cannot fully reflect the complex physiological mechanism of blood pressure changes, and the data processing method of multi-sensor fusion has been improved, but the data preprocessing, feature extraction and model prediction still need to be further optimized. In addition, the traditional machine learning algorithm often has difficulty in fully capturing the deep features in the data when processing high-dimensional and time-series data, resulting in insufficient accuracy and stability of blood pressure prediction.

[0004] Therefore, there is an urgent need for a new data processing method that can effectively integrate multi-sensor data, improve the accuracy and real-time performance of blood pressure prediction, and have good generalization ability to adapt to the needs of different user groups. SUMMARY

[0005] The present application provides a blood pressure prediction model training method and a blood pressure data prediction method to solve the problems in the prior art.

[0006] To solve the above technical problems, the present application solves the problems by the following technical solutions: A blood pressure prediction model training method, comprising the following steps: Obtaining data of a wearable device to obtain an original data sequence, wherein the data of the wearable device at least includes a pressure signal, a blood pressure signal, an acceleration and an angular velocity; Preprocessing the original data sequence to obtain a standard data sequence; Extracting features of the standard pressure signal sequence and the standard blood pressure sequence based on peak and valley to obtain a pressure feature sequence, a blood pressure feature sequence and a pulse wave propagation time sequence; Extracting features of the standard acceleration sequence and the standard angular velocity sequence based on the rate of change and the frequency to obtain a motion intensity sequence; Fusing the pressure feature sequence, the blood pressure feature sequence, the pulse wave propagation time sequence, and the motion intensity sequence to form a multi-dimensional feature sequence; Training the blood pressure prediction pre-training model based on the multi-dimensional feature sequence to obtain a blood pressure prediction model.

[0007] As an implementable manner, the pressure feature sequence at least includes a pressure peak value sequence, a pressure valley value sequence, a pressure rising time sequence, and a pressure falling time sequence. The pressure feature sequence is obtained by the following steps: Extracting the peak value and the valley value of each heartbeat cycle of the standard pressure signal sequence to obtain a pressure peak-valley value sequence, a pressure peak value sequence, and a pressure valley value sequence; Subtracting the timestamp of a target peak value in the pressure peak-valley value sequence from the timestamp of a valley value before the target peak value to obtain a pressure rising time sequence; Subtracting the timestamp of a target valley value in the pressure peak-valley value sequence from the timestamp of a peak value before the target valley value to obtain a pressure falling time sequence.

[0008] As an implementable manner, the blood pressure feature sequence at least includes a blood voltage peak value sequence, a blood voltage rising rate sequence, and a blood voltage falling rate sequence. The blood pressure feature sequence is obtained by the following steps: Extracting the peak value and the valley value of each blood volume change cycle of the standard blood pressure sequence to obtain a blood voltage peak-valley value sequence and a blood voltage peak value sequence; Calculating the rate at which the blood voltage rises from a valley value to a subsequent peak value in the blood voltage peak-valley value sequence to obtain a blood voltage rising rate sequence; Calculating the rate at which the blood voltage falls from a peak value to a subsequent valley value in the blood voltage peak-valley value sequence to obtain a blood voltage falling rate sequence.

[0009] As an implementable manner, the pulse wave propagation time sequence is obtained by the following steps: Extracting the peak value of each heartbeat cycle of the standard pressure signal sequence to obtain a pressure peak value sequence; Extracting the peak value of each blood volume change cycle of the standard blood pressure sequence to obtain a blood voltage peak value sequence; Subtracting the pressure peak value sequence from the blood voltage peak value sequence based on the timestamp to obtain a pulse wave propagation time sequence.

[0010] As an implementable manner, the motion intensity sequence at least includes a motion intensity level sequence, an acceleration change rate sequence, and a motion frequency sequence. The motion intensity sequence is obtained by the following steps: Summing the standard acceleration sequence and the standard angular velocity sequence based on the rate to obtain a comprehensive rate sequence; based on the integrated rate sequence, the first preset rate threshold and the second preset rate threshold, a motion intensity level sequence is obtained; The rate of change of acceleration in the standard acceleration sequence is calculated to obtain an acceleration rate of change sequence; Based on the Fourier transform, the frequency extraction in each preset period of the standard acceleration sequence is performed to obtain an acceleration frequency feature sequence; The maximum frequency in each preset period in the acceleration frequency feature sequence, i.e. the motion frequency, is extracted to obtain a motion frequency sequence.

[0011] As an implementable manner, the blood pressure prediction pre-training model comprises a convolutional recurrent fusion module, an attention weighting module and a prediction module; The convolutional recurrent fusion module performs convolution operation and bidirectional recurrent operation on the input vector respectively to obtain a convolutional feature sequence and a bidirectional feature sequence; and fuses the convolutional feature sequence and the bidirectional feature sequence to obtain an integrated feature sequence; The attention weighting module performs attention mechanism processing on the integrated feature sequence to obtain an attention weight sequence, and performs weighting processing on the integrated feature sequence and the attention weight sequence to obtain a weighted feature sequence; The prediction module maps the weighted feature sequence through a fully connected layer to obtain a blood pressure data prediction value; The blood pressure data prediction value is represented as follows:

[0012] wherein, represents the systolic blood pressure prediction value, represents the diastolic blood pressure prediction value, represents the matrix transpose, represents the fully connected layer weight, represents the weighted feature sequence, represents the fully connected layer bias.

[0013] A blood pressure data prediction method, comprising the following steps: Obtaining target data of a wearable device; Processing the target data based on a blood pressure prediction model to obtain a blood pressure data prediction value; The blood pressure prediction model is obtained by using the training method of the blood pressure prediction model of any one of the above.

[0014] A blood pressure prediction model training system for implementing the training method of the blood pressure prediction model of any one of the above, the training system comprising: A data acquisition module, configured to acquire data from a wearable device to obtain an original data sequence, wherein the data from the wearable device includes at least a pressure signal, a blood pressure signal, acceleration, and angular velocity; The data processing module is used to preprocess the original data sequence to obtain a standard data sequence; A first feature extraction module is used to extract features of the standard pressure signal sequence and the standard blood pressure sequence based on peak values ​​and valley values, respectively, to obtain a pressure feature sequence, a blood pressure feature sequence, and a pulse wave propagation time sequence; The second feature extraction module is used to extract features of the standard acceleration sequence and the standard angular velocity sequence based on the change rate and frequency to obtain a motion intensity sequence; Feature fusion module, used to fuse pressure feature sequence, blood pressure feature sequence, pulse wave propagation time sequence and exercise intensity sequence to form a multi-dimensional feature sequence; The model training module is used to train the blood pressure prediction pre-training model based on the multi-dimensional feature sequence to obtain the blood pressure prediction model.

[0015] A blood pressure data prediction system, the prediction system comprising: An acquisition module is used to obtain target data of the wearable device; A prediction module is used to infer the target data based on a blood pressure prediction model to obtain a predicted value of the blood pressure data; wherein the blood pressure prediction model is a model obtained by using any of the above-mentioned blood pressure prediction model training methods.

[0016] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements any one of the above-mentioned methods for training a blood pressure prediction model.

[0017] A blood pressure prediction model training device comprises a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, the blood pressure prediction model training method described in any one of the above is implemented.

[0018] Due to the adoption of the above technical solutions, the present invention has significant technical effects: the present invention provides a blood pressure prediction model training method and a blood pressure data prediction method, which effectively overcomes the limitations of a single sensor data source through the collaborative processing and deep fusion of multi-sensor data, and significantly improves the comprehensiveness and reliability of data acquisition. Through innovative model training strategies and data processing procedures, the system can adapt to user groups with different physiological characteristics and maintain stable prediction performance in a variety of application scenarios. The various technical features of the present invention support each other and are closely integrated in function, which improves the accuracy and real-time performance of the blood pressure prediction model in predicting blood pressure. At the same time, the blood pressure prediction model has good generalization capabilities and can adapt to the needs of different user groups. It has significant technical advantages and application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0020] Figure 1 1 is a flow chart of an embodiment of a method for training a blood pressure prediction model according to the present invention; Figure 2 1 is an overall schematic diagram of an embodiment of a training system for a blood pressure prediction model according to the present invention; Figure 3 1 is a flow chart of an embodiment of a method for predicting blood pressure data according to the present invention; Figure 4 It is an overall schematic diagram of an embodiment of the blood pressure data prediction system of the present invention. DETAILED DESCRIPTION

[0021] The present invention will be further described in detail below with reference to the embodiments. The following embodiments are for explanation of the present invention and the present invention is not limited to the following embodiments. In the absence of conflict, the features in the following embodiments may be combined with each other.

[0022] Example 1: A training method for a blood pressure prediction model, such as Figure 1 As shown, the following steps are included: S110: Acquire data from the wearable device to obtain an original data sequence, wherein the data from the wearable device at least includes a pressure signal, a blood pressure signal, acceleration, and angular velocity; S120: Preprocessing the original data sequence to obtain a standard data sequence; S130: Extract features of the standard pressure signal sequence and the standard blood pressure sequence based on the peak value and the valley value respectively, to obtain a pressure feature sequence, a blood pressure feature sequence, and a pulse wave propagation time sequence; S140: Extract features of the standard acceleration sequence and the standard angular velocity sequence based on the change rate and the frequency, to obtain a motion intensity sequence; S150: Fuse the pressure feature sequence, the blood pressure feature sequence, the pulse wave propagation time sequence, and the motion intensity sequence to form a multi-dimensional feature sequence; S160: Train the blood pressure prediction pre-trained model based on the multi-dimensional feature sequence to obtain a blood pressure prediction model.

[0023] In S110, data of the wearable device is acquired to obtain an original data sequence, wherein the data of the wearable device at least includes a pressure signal, a blood pressure signal, acceleration, and angular velocity. The data of the wearable device is collected by a built-in sensor of the wearable device, and the built-in sensor at least includes a pressure sensor, a blood volume sensor, an acceleration sensor, and an angular velocity sensor. Specifically as follows: (1) Pressure sensor The pressure sensor is used to acquire data related to the pressure signal of the wearable device, for example, MEMS (Micro Electro Mechanical System) pressure sensor, PVDF film pressure sensor, etc., and a high-sensitivity MEMS pressure sensor die is preferably used as the main pressure detection element. The pressure sensor is usually located at the wrist wearing part, and the watch bottom directly contacts the skin to collect the pressure applied by the user in real time.

[0024] The pressure collected by the film is conducted to the MEMS pressure sensor die through a small fluid pipeline by using internal silicone oil (silicone oil has good viscosity and stability). The silicone oil conduction path is a cylindrical pipeline. Under the condition of steady flow, the relationship between the flow rate and the pressure gradient in the silicone oil is: wherein the flow rate in the silicone oil is in cubic meters per second. The pressure gradient represents the pressure difference between the two ends of the silicone oil conduction path, i.e. the pressure applied by the film and the pressure received by the sensor, i.e. . represents the radius of the silicone oil conduction path, in meters. represents the dynamic viscosity of the silicone oil, in Pascal seconds. represents the length of the silicone oil conduction path, in meters. Therefore, the pressure received by the sensor is: Since the pressure will be attenuated to a certain extent during conduction, it is necessary to ensure that is close to , you need to meet the following requirements: , which requires As small as possible, and As big as possible.

[0025] The relationship between the output voltage and input pressure of the pressure sensor is: ,in, Indicates the voltage signal output by the sensor in millivolts. Indicates the sensitivity of the sensor in mV / Pa (millivolt / Pa). It indicates the pressure received by the sensor, that is, the actual pressure applied to the sensor, in Pa.

[0026] Pressure transmitted to the film The force applied by the user to the membrane The relationship between them is: ,in, The effective contact area of ​​the film is expressed in square meters. The force applied by the user to the film is expressed in Newtons. The pressure transmitted to the film is expressed in Pascals.

[0027] After the pressure collected by the film is transmitted to the MEMS sensor through silicone oil, the sensor converts the pressure into a voltage signal. The voltage signal output by the pressure sensor Expressed as: The unit of the voltage signal output by the pressure sensor is mV, so the voltage signal output by the pressure sensor is The force applied by the user to the film associated.

[0028] It should be noted that those skilled in the art are well aware that although the sensor collects pressure data, in the context of blood pressure measurement, pressure is a form of pressure. Therefore, pressure and pressure are actually only conceptual conversions, not numerical conversions. Based on the habits of those skilled in the art, the text in this application is still expressed as pressure, but the symbol is represented as (pressure).

[0029] (2) Blood volume sensor The blood volume sensor is used to measure the change of blood volume, for example, a photoplethysmography sensor PPG, a bioimpedance sensor BIA, etc. In this embodiment, a PPG sensor is used to measure the change of blood volume, and its output is a PPG signal voltage recorded as ,but ,in, Indicates the voltage signal output by the PPG sensor, in mV. Indicates the sensitivity of the PPG sensor in units of Unit light intensity change. Indicates the change in light intensity caused by changes in blood volume.

[0030] (3) Accelerometer There are many types of accelerometers, which are used to measure the acceleration of the user. The accelerometer outputs the acceleration vector ,but ,in, Represents the acceleration sensor measured in The acceleration value in the direction of . Indicates the sensitivity of the accelerometer (unit: ). Represents the output voltage of the acceleration sensor, then , Represents the acceleration sensor measured in The output voltage value in the direction is in mV. Indicates transpose.

[0031] (4) Angular velocity sensor Angular velocity sensors are used to measure angular velocity. They are mainly divided into different types of gyroscopes. The output of the angular velocity sensor is the angular velocity vector measured by the gyroscope. ,but ,in, Represents the gyroscope measurement The angular velocity value of the direction, in rad / s. Indicates the sensitivity of the gyroscope in units of . represents the output voltage of the gyroscope, then , Represents the gyroscope measurement The output voltage value in the direction is in mV.

[0032] In S120, the original data sequence is preprocessed to obtain a standard data sequence, including the following steps: S121: Perform bandpass filtering on the original pressure signal sequence and the original blood voltage signal sequence to obtain a bandpass pressure signal sequence and a bandpass blood voltage signal sequence. The details are as follows: (1) Identify the frequency of the original pressure signal sequence and obtain the pressure signal frequency sequence.

[0033] (2) Based on the pressure frequency range formed by the first preset pressure frequency threshold and the second preset pressure frequency threshold, if the frequency in the pressure signal frequency sequence is within the pressure frequency range, the corresponding transfer value is 1, and then the pressure transfer value sequence is obtained. The pressure transfer value sequence can be expressed as: ,in, Indicates the first value in the pressure transfer value sequence Pass value, Indicates the first frequencies, They represent the first preset pressure-frequency threshold and the second preset pressure-frequency threshold respectively.

[0034] (3) Perform inverse Fourier transform on the transfer value with a value of 1 in the pressure transfer value sequence to obtain the pressure time domain response value sequence. The pressure time domain response value sequence can be expressed as: ,in, Indicates the first value in the pressure time domain response value sequence Time domain response values, represents the imaginary unit, Indicates frequency, Indicates time.

[0035] (4) Convolution processing is performed on the pressure time domain response value sequence and the original pressure signal sequence to achieve bandpass filtering processing, thereby obtaining a bandpass pressure signal sequence; the bandpass pressure signal sequence is expressed as follows: ,in, represents the bandpass pressure signal sequence, represents the original pressure signal sequence, represents the pressure time domain response value sequence, Represents convolution processing.

[0036] (5) Perform bandpass filtering on the original blood voltage signal sequence to obtain a bandpass blood voltage signal sequence.

[0037] The original pressure signal sequence and the original blood voltage signal sequence are subjected to band-pass filtering to obtain a band-pass pressure signal sequence and a band-pass blood voltage signal sequence.

[0038] Bandpass filtering can remove power supply interference (such as 50 Hz) and high-frequency noise in the original pressure signal sequence and the original blood voltage signal sequence, and retain low-frequency components related to blood pressure (such as 0.5 Hz to 5 Hz).

[0039] S122: Low-pass filtering is performed on the original acceleration sequence and the original angular velocity sequence to obtain a low-pass acceleration sequence and a low-pass angular velocity sequence, as follows: (1) Identify the frequency of the original acceleration sequence and obtain the acceleration frequency sequence.

[0040] (2) If the frequency in the acceleration frequency sequence is lower than or equal to the preset acceleration frequency threshold, the corresponding transfer value is 1, and the acceleration transfer value sequence is obtained. The acceleration transfer value sequence can be expressed as: ,in, Indicates the first value in the acceleration transfer value sequence Pass value, Indicates the first acceleration frequency in the sequence frequencies, Indicates the preset acceleration frequency threshold.

[0041] (3) Perform inverse Fourier transform on the transfer value with a value of 1 in the acceleration transfer value sequence to obtain the acceleration time domain response value sequence. The acceleration time domain response value sequence can be expressed as: ,in, Indicates the first value in the acceleration time domain response value sequence Time domain response values, represents the imaginary unit, Indicates frequency, Indicates time.

[0042] (4) Convolution processing is performed on the acceleration time domain response value sequence and the original acceleration sequence to achieve low-pass filtering processing, thereby obtaining a low-pass acceleration sequence; the low-pass acceleration sequence is expressed as: ,in, represents the low-pass acceleration sequence, represents the original acceleration sequence, represents the acceleration time domain response value sequence, Represents convolution processing.

[0043] Low-pass filtering can remove high-frequency motion noise in the original acceleration sequence and the original angular velocity sequence, and retain the user's main motion state information (such as stillness, walking, and running).

[0044] Data sequence filtering aims to remove noise and irrelevant frequency components from the data sequence, improving data quality. This is a key step in ensuring that the raw data collected from the sensor can be effectively utilized by the blood pressure prediction pretrained model. Based on the actual application environment and sensor characteristics, appropriate bandpass and lowpass filters can be selected to optimize data sequence quality. Using digital filters (such as FIR or IIR filters) to filter the data sequence reduces computational complexity. Butterworth filters can be used to design bandpass and / or lowpass filters, offering advantages such as a smooth frequency response and a lack of noticeable ripple.

[0045] In the present application, band-pass filtering and low-pass filtering are respectively performed on different types of raw data to improve data quality and the accuracy of the blood pressure prediction model. Band-pass filtering is mainly used to process pressure signals and blood pressure signals, which can remove high-frequency noise and low-frequency interference in the original signals, retain the key low-frequency components of blood pressure changes, and effectively enhance the effectiveness of the data and the prediction accuracy. Low-pass filtering is applied to acceleration signals and angular velocity signals to remove high-frequency motion noise and retain low-frequency information related to the user's motion state. This step helps to accurately capture changes in motion intensity and improve the model's ability to recognize motion states, thereby enhancing the stability and reliability of the blood pressure prediction model. Through the joint application of band-pass filtering and low-pass filtering, the data preprocessing process can be significantly optimized, and the overall performance of the blood pressure prediction model can be improved.

[0046] S123: Data alignment (synchronization) processing. If the sampling frequencies of the band-pass pressure signal sequence, the band-pass blood voltage signal sequence, the low-pass acceleration sequence, and the low-pass angular velocity sequence are inconsistent, data alignment processing based on a unified timestamp is performed on the band-pass pressure signal sequence, the band-pass blood voltage signal sequence, the low-pass acceleration sequence, and the low-pass angular velocity sequence. Specifically as follows: (1) Obtain the minimum timestamp and the maximum timestamp in the band-pass pressure signal sequence, the band-pass blood voltage signal sequence, the low-pass acceleration sequence, and the low-pass angular velocity sequence, and then obtain a unified timestamp sequence in combination with a pre-set time interval.

[0047] (2) Based on the unified timestamp sequence, perform alignment processing on the band-pass pressure signal sequence, the band-pass blood voltage signal sequence, the low-pass acceleration sequence, and the low-pass angular velocity sequence, which includes one or more of interpolation, forward padding, backward padding, and resampling.

[0048] Data synchronization ensures that data from different sensors are aligned in time, facilitating subsequent feature extraction and model training.

[0049] S124: Standardization processing. Perform standardization processing on the band-pass pressure signal sequence, the band-pass blood voltage signal sequence, the low-pass acceleration sequence, and the low-pass angular velocity sequence respectively to obtain a standard pressure signal sequence, a standard blood voltage signal sequence, a standard acceleration sequence, and a standard angular velocity sequence, including the following steps: (1) Based on the band-pass pressure signal sequence, obtain the mean of the pressure signal and the standard deviation of the pressure signal.

[0050] (2) Based on the band-pass pressure signal sequence, the mean of the pressure signal, and the standard deviation of the pressure signal, obtain the standard pressure signal sequence, which can be represented as: wherein, represents the i-th pressure signal in the standard pressure signal sequence, represents the i-th pressure signal in the standard pressure signal sequence, Indicates the first A pressure signal, represents the mean value of the pressure signal, represents the standard deviation of the pressure signal, represents a constant, represents a learnable scaling parameter, represents the learnable translation parameters.

[0051] (3) The band-pass blood voltage signal sequence, low-pass acceleration sequence and low-pass angular velocity sequence are standardized to obtain the standard blood voltage signal sequence, standard acceleration sequence and standard angular velocity sequence.

[0052] Normalization converts data from different sensors and features into the same dimension, eliminating the dimensional differences between different features, thereby improving the efficiency and stability of blood pressure prediction pre-training model training.

[0053] In S130, the features of the standard pressure signal sequence and the standard blood pressure sequence are extracted based on the peak values ​​and the valley values, respectively, to obtain the pressure feature sequence, the blood pressure feature sequence, and the pulse wave propagation time sequence, including the following steps: S131: Obtain the pressure characteristic sequence, which includes at least a pressure peak sequence, a pressure valley sequence, a pressure rise time sequence, and a pressure fall time sequence, specifically including: (1) The peak value (systolic pressure) and valley value (diastolic pressure) of each heartbeat cycle are extracted from the standard pressure signal sequence to obtain a pressure peak-valley value sequence, a pressure peak value sequence, and a pressure valley value sequence. The pressure peak-valley value sequence is a sequence containing the peak value and valley value of each heartbeat cycle.

[0054] (2) The timestamp of the target peak value in the pressure peak-valley value sequence and the timestamp of the valley value before the target peak value are subtracted to obtain the pressure rise time series. The pressure rise time represents the time required for the pressure to rise from the valley value to the peak value.

[0055] (3) The timestamp of the target valley value in the pressure peak-valley value sequence is subtracted from the timestamp of the peak value before the target valley value to obtain the pressure drop time series. The pressure drop time represents the time required for the pressure to drop from the peak value to the valley value of the next cycle.

[0056] The pressure rise time series can be expressed as: , the pressure drop time series can be expressed as: ,in, Indicates the first Rise time, Indicates the first value in the pressure peak-valley sequence The timestamp corresponding to the pressure peak, Indicates the first value in the pressure peak-valley sequence The timestamp corresponding to the pressure valley value before the pressure peak, Indicates the first The fall time, Indicates the first value in the pressure peak-valley sequence The timestamp corresponding to the pressure valley value, Indicates the first value in the pressure peak-valley sequence The timestamp corresponding to the previous pressure peak of the pressure valley.

[0057] S132: Obtain the blood pressure characteristic sequence, which includes at least a blood voltage peak sequence, a blood voltage rising rate sequence, and a blood voltage falling rate sequence, as follows: (1) The peak and valley values ​​of each blood volume change cycle of the standard blood pressure sequence are extracted to obtain the blood voltage peak-valley value sequence and the blood voltage peak value sequence.

[0058] (2) Calculate the rate at which the blood voltage rises from the valley value to the next peak value in the blood voltage peak-valley value sequence to obtain the blood voltage rise rate sequence.

[0059] (3) Calculate the rate at which the blood voltage drops from the peak value to the next valley value in the blood voltage peak-valley value sequence to obtain the blood voltage drop rate sequence.

[0060] Among them, the blood voltage rising rate series can be expressed as: , the blood voltage drop rate series can be expressed as: ,in, Indicates the first in the blood voltage rising rate sequence The blood voltage rise rate, Indicates the peak-valley value sequence of blood voltage Blood voltage valley, Indicates the peak-valley value sequence of blood voltage The next blood voltage peak after the blood voltage valley, Indicates the timestamp corresponding to the next blood voltage peak value in the blood voltage peak-to-valley value sequence, Indicates the peak-valley value sequence of blood voltage The timestamp corresponding to the blood voltage valley value, Indicates the peak-valley value sequence of blood voltage Peak blood voltage, Indicates the peak-valley value sequence of blood voltage The blood voltage valley after the blood voltage peak, Indicates the peak-valley value sequence of blood voltage The timestamp corresponding to the blood voltage peak, Indicates the timestamp corresponding to the next blood voltage valley value in the blood voltage peak-valley value sequence.

[0061] The blood voltage peak is used to characterize the peak position of the blood voltage waveform. The blood voltage rise rate and the blood voltage fall rate are characteristics related to the shape of the blood voltage waveform, reflecting the dynamic characteristics of blood volume changes.

[0062] S133: Obtain the pulse wave propagation time sequence, as follows: (1) Extract the peak value of each heartbeat cycle from the standard pressure signal sequence to obtain the pressure peak sequence.

[0063] (2) The peak value of each blood volume change cycle of the standard blood pressure sequence is extracted to obtain the blood voltage peak sequence.

[0064] (3) Perform time stamp-based difference processing on the blood voltage peak sequence and pressure peak sequence to obtain the pulse wave propagation time series, which can be expressed as: ,in, Indicates the first pulse in the pulse wave propagation time series propagation time, Indicates the first peak in the pressure sequence The timestamp corresponding to the pressure peak, Indicates the peak value of blood voltage The timestamp corresponding to the blood voltage peak.

[0065] The heartbeat is the driving force behind blood volume changes, which are the result of the heartbeat's propagation through the vascular system. The pulse wave propagation time (PWT) represents the time it takes the pulse wave to travel along the conduction path (from the heartbeat to the blood volume change), reflecting the elasticity and hemodynamic properties of the arteries. Because the heartbeat precedes the blood volume change, the PWT is positive. If the PWT is negative due to the timing of the acquisition, the standard blood pressure sequence for one blood volume change cycle is discarded. Under physiological conditions, the heartbeat cycle and the blood volume change cycle are equal in length. Because they are both driven by the heartbeat and are different manifestations of the same physiological process, the PWT can be considered to be derived based on the heartbeat cycle.

[0066] In S140, features of the standard acceleration sequence and the standard angular velocity sequence are extracted based on the change rate and frequency to obtain a motion intensity sequence, wherein the motion intensity sequence includes at least a motion intensity level sequence, an acceleration change rate sequence, and a motion frequency sequence, including the following steps: S141: Perform rate-based summation processing on the standard acceleration sequence and the standard angular velocity sequence to obtain a comprehensive rate sequence. The comprehensive rate sequence can be expressed as: ,in, Indicates the first The rate of timestamps, Represents the first The acceleration of the timestamp is The acceleration value in the axis direction, Represents the first The angular velocity of the timestamp is The angular velocity value in the axis direction.

[0067] S142: Based on the comprehensive rate sequence, the first preset rate threshold, and the second preset rate threshold, an exercise intensity level sequence is obtained. The exercise intensity level sequence can be expressed as: ,in, Indicates the first level in the sequence of exercise intensity levels. The exercise intensity level value of the timestamp, represents the first preset rate threshold, The second preset speed threshold is represented by the motion intensity level value of 0, which indicates that the user is stationary, 1, which indicates that the user is walking, and 2, which indicates that the user is running.

[0068] S143: Calculate the rate of change of acceleration in the standard acceleration sequence to obtain an acceleration rate of change sequence. The acceleration rate of change sequence is expressed as: ,in, Indicates the acceleration rate of change sequence The acceleration rate of each timestamp, Indicates the first acceleration in the standard acceleration sequence The acceleration of the timestamp, Indicates the first acceleration in the standard acceleration sequence The acceleration of the timestamp, Indicates the first acceleration in the standard acceleration sequence The timestamp corresponding to the acceleration of the timestamp, Indicates the first acceleration in the standard acceleration sequence The timestamp corresponding to the acceleration of the timestamp.

[0069] S144: Based on Fourier transform, extract the frequency of the standard acceleration sequence within each preset period to obtain an acceleration frequency characteristic sequence.

[0070] S145: Extract the maximum frequency in each preset period of the acceleration frequency characteristic sequence, i.e., the motion frequency, and then obtain the motion frequency sequence. The motion frequency can be expressed as: ,in, represents the movement frequency, represents the Fourier transform, Indicates the range of acceleration frequency in the acceleration frequency characteristic sequence, The acceleration frequency corresponding to the maximum value of the function is represented by the Fourier transform, which is used to perform frequency domain analysis on the acceleration data to extract the main motion frequency components. Preferably, the fast Fourier transform (FFT) can be used.

[0071] The preset period is an integer multiple of the heartbeat period ≥ 2. The timestamp corresponding to the heartbeat period can be obtained through the standard pressure signal sequence, and the timestamp corresponding to the preset period can be extracted accordingly.

[0072] Acceleration rate and motion frequency are used to capture the dynamic changes in the user's motion state, thereby further refining the impact of exercise on blood pressure.

[0073] The order of step S130 and step S140 is not limited.

[0074] In S150, the pressure feature sequence, blood pressure feature sequence, pulse wave propagation time sequence and exercise intensity sequence are fused to form a multi-dimensional feature sequence Each value in the pressure feature sequence (pressure peak sequence, pressure trough sequence, pressure rise time sequence, and pressure fall time sequence) is obtained for each cardiac cycle, and the length of each pressure feature sequence is consistent with the total number of cardiac cycles. In contrast, each value in the blood pressure feature sequence (blood voltage peak sequence, blood voltage rise rate sequence, and blood voltage fall rate sequence) is obtained for each blood volume change cycle, and the length of each blood pressure feature sequence is consistent with the total number of blood volume change cycles. As mentioned above, each value in the pulse wave propagation time series can be considered to be obtained for each cardiac cycle. Since the cardiac cycle and blood volume change cycle are of equal length, the lengths of the pressure feature sequence, blood pressure feature sequence, and pulse wave propagation time series are generally identical, occasionally differing by one value. In such cases, the first or last value of the longer sequence is discarded.

[0075] Each value in the exercise intensity level sequence and acceleration change rate sequence in the exercise intensity sequence is obtained based on each timestamp, and the sequence length is significantly greater than the total number of heartbeat cycles. This is processed as follows: (1) extract the timestamp of each heartbeat cycle in the standard pressure signal sequence; (2) calculate the average acceleration change rate in each heartbeat cycle based on the timestamp of each heartbeat cycle, and obtain the average acceleration change rate sequence; (3) obtain the exercise intensity level that appears most frequently in each heartbeat cycle based on the timestamp of each heartbeat cycle, and obtain the main exercise intensity level sequence. In this way, the processed exercise intensity level sequence and acceleration change rate sequence are consistent with the sequence length of the pressure feature sequence / blood pressure feature sequence / pulse wave propagation time sequence.

[0076] The motion frequency sequence is derived based on a preset period, which is an integer multiple of the heartbeat cycle ≥ 2. Therefore, the length of each sequence in the pressure signature sequence is also an integer multiple of the length of the motion frequency sequence. Therefore, the motion frequency within the corresponding preset period in the motion frequency sequence is simply expanded to the corresponding heartbeat cycle to obtain an extended motion frequency sequence. The length of the extended motion frequency sequence is consistent with the length of the other (processed) sequences.

[0077] The lengths of the processed sequences are the same, and the fusion process can be performed by matching the values ​​of each sequence in order.

[0078] In S160, the blood pressure prediction pre-training model is trained based on the multi-dimensional feature sequence to obtain a blood pressure prediction model, including the following steps: S161: Construct a blood pressure prediction pre-training model. The blood pressure prediction pre-training model can directly apply an existing artificial intelligence model or be self-built based on artificial intelligence technology. Self-building a blood pressure prediction pre-training model can better balance performance, efficiency, and scenario requirements. For example, the model architecture can be customized based on the specific sensor data of the wearable device (such as MEMS pressure sensor, PPG, etc.), and targeted optimization can be performed to improve prediction accuracy. In this embodiment, a method of self-building a pre-training model is adopted. The blood pressure prediction pre-training model includes a convolutional recursive fusion module, an attention weighting module, and a prediction module, as follows: (1) Convolution-recursion fusion module, which performs convolution operation and bidirectional recursion operation on the input vector respectively to obtain convolution feature sequence and bidirectional feature sequence; and fuses the convolution feature sequence and bidirectional feature sequence to obtain a comprehensive feature sequence. In this embodiment, the convolution operation is implemented by convolutional neural network (CNN), and the bidirectional recursion operation is implemented by bidirectional recursive neural network (Bi-RNN). CNN processes the waveform features of PPG signals and extracts local time-frequency features. Bi-RNN can process the temporal features of pressure signals and motion data to capture long-term dependencies.

[0079] Then the convolution feature sequence can be expressed as: , the bidirectional feature sequence can be expressed as: , the comprehensive feature sequence can be expressed as: ,in, represents the convolution feature sequence, represents the input vector, represents the activation function, represents the convolution kernel weight, represents the bias term, represents a bidirectional feature sequence, Represents a bidirectional recursive operation, represents the comprehensive feature sequence, represents the convolution operation, Represents a fusion operation.

[0080] (2) Attention weighting module: The attention mechanism is used to process the comprehensive feature sequence to weight the importance of different features, and the attention weight sequence is obtained. The comprehensive feature sequence and the attention weight sequence are weighted to obtain the weighted feature sequence. The attention weight sequence can be expressed as: , the weighted feature sequence can be expressed as: ,in, Indicates the first attention weights, Indicates the Importance score, , Both represent learnable parameters, represents the matrix transpose, Indicates the first Features, represents the dimension of the comprehensive feature sequence, represents a weighted feature sequence.

[0081] (3) The prediction module maps the weighted feature sequence through the fully connected layer to obtain the predicted value of blood pressure data. The predicted value of blood pressure data can be expressed as: ,in, represents the predicted value of blood pressure data, represents the predicted value of systolic blood pressure, represents the predicted value of diastolic blood pressure, represents the weight of the fully connected layer, represents the bias of the fully connected layer.

[0082] S162: Based on a preset partitioning ratio, the multidimensional feature sequence is partitioned to obtain a training dataset, a test dataset, and a validation dataset. A user dataset encompassing different ages, genders, health conditions, and lifestyles can be constructed to improve the model's generalization capabilities. Furthermore, data augmentation methods (such as noise addition and time warping) can be applied to expand the training dataset, improving the model's anti-interference ability and robustness.

[0083] S163: Input the training dataset into the blood pressure prediction pre-training model. The model's predicted output is obtained through a forward propagation algorithm. Combined with the actual value, a loss function value is obtained. The loss function value is the sum of the diastolic mean square error, the systolic mean square error, and the regularization value. This prevents overfitting and improves the model's generalization ability. Furthermore, the shared hidden layer can improve the model's overall understanding of blood pressure changes.

[0084] Then the loss function value can be expressed as: , the mean square error of systolic blood pressure can be expressed as: , the mean square error of diastolic pressure can be expressed as: , the regularization value can be expressed as: ,in, represents the loss function value, represents the mean square error of systolic blood pressure, are the mean square error of diastolic blood pressure, They represent the weight coefficients of the mean square error of systolic blood pressure and the mean square error of diastolic blood pressure, represents the number of samples in the training dataset, They represent the systolic pressure prediction value and diastolic pressure prediction value predicted by the blood pressure prediction pre-training model respectively. Represent the actual values ​​of systolic and diastolic blood pressure, respectively. represents the regularization value, represents the regularization strength, represents the regularization weight, represents the regularization bias.

[0085] S164: Calculate the gradient of the blood pressure prediction pre-training model parameters based on the loss function value through the back propagation algorithm.

[0086] S165: Based on the learning rate of the training cycle, the parameters of the blood pressure prediction pre-training model are updated. Specifically, the Adam optimizer is used for efficient gradient descent to quickly converge to the optimal solution. The learning rate decay strategy is also used to gradually reduce the learning rate, stabilize the training process, and improve model performance. The learning rate can be expressed as: ,in, Indicates the The learning rate for each training cycle, represents the initial learning rate, represents the learning rate decay coefficient, Represents the training cycle.

[0087] S166: Systematically adjust hyperparameters (e.g., learning rate, batch size, number of network layers) to find the optimal model configuration using grid search or Bayesian optimization. Evaluate the generalization ability of the pre-trained blood pressure prediction model using a validation dataset.

[0088] S167: Iterate the training until an iteration termination condition is met to obtain a blood pressure prediction model, wherein the iteration termination condition includes the convergence of the loss function value or reaching a preset number of training times.

[0089] Example 2: A blood pressure prediction model training system is used to implement the blood pressure prediction model training method of any of the above embodiments, such as Figure 2 As shown, the training system includes: The data acquisition module 110 is used to acquire data from the wearable device to obtain an original data sequence, wherein the data from the wearable device includes at least a pressure signal, a blood pressure signal, acceleration, and angular velocity; The data processing module 120 is used to pre-process the original data sequence to obtain a standard data sequence; A first feature extraction module 130 is configured to extract features of a standard pressure signal sequence and a standard blood pressure sequence based on peak values ​​and valley values, respectively, to obtain a pressure feature sequence, a blood pressure feature sequence, and a pulse wave propagation time sequence; A second feature extraction module 140 is configured to extract features of the standard acceleration sequence and the standard angular velocity sequence based on the change rate and frequency to obtain a motion intensity sequence; A feature fusion module 150 is used to fuse the pressure feature sequence, the blood pressure feature sequence, the pulse wave propagation time sequence, and the motion intensity sequence to form a multi-dimensional feature sequence; The model training module 160 is used to train the blood pressure prediction pre-training model based on the multi-dimensional feature sequence to obtain a blood pressure prediction model.

[0090] Example 3: A blood pressure data prediction method, such as Figure 3 As shown, the following steps are included: S210: Acquire target data of the wearable device; S220: Processing the target data based on the blood pressure prediction model to obtain a predicted value of the blood pressure data; The blood pressure prediction model is a model obtained by using the blood pressure prediction model training method of any one of the above embodiments.

[0091] Example 4: A blood pressure data prediction system, such as Figure 4 As shown, the prediction system includes: An acquisition module 210 is configured to acquire target data of a wearable device; A prediction module 220 is configured to infer the target data based on a blood pressure prediction model to obtain a predicted value of the blood pressure data; The blood pressure prediction model is a model obtained by using the blood pressure prediction model training method of any one of the above embodiments.

[0092] Various changes and modifications can be made without departing from the spirit and scope of the present invention, and all equivalent technical solutions also fall within the scope of the present invention.

[0093] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be referenced to each other.

[0094] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, apparatus, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0095] The present invention is described with reference to the flowcharts and / or block diagrams of the method, terminal device (system), and computer program product according to the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the process in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0096] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1the function specified in the one or more blocks.

[0097] These computer program instructions can also be loaded into computer or other programmable data processing terminal devices, so that a series of operational steps are performed on the computer or other programmable terminal devices to generate a computer implemented process, so that the instructions executed on the computer or other programmable terminal devices provide a process for implementing the flow Figure 1 the flow or flows and / or blocks Figure 1 the steps of the function specified in the one or more blocks.

[0098] It should be noted that: The phrase "one embodiment" or "an embodiment" as used throughout this specification means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. The appearances of the phrase "one embodiment" or "an embodiment" in various places throughout the specification are not necessarily all referring to the same embodiment.

[0099] Furthermore, it should be noted that the specific embodiments described in the specification, the shape of the zero, components, the name taken, etc. can be different. Any equivalent or simple change made in accordance with the structure, features and principles described in the patent concept of the present application is included in the protection scope of the present application. Those skilled in the art can make various modifications or supplements to the described specific embodiments or use similar ways to replace them, as long as they do not deviate from the structure of the present application or exceed the scope defined by the claims.

Claims

1. A method for training a blood pressure prediction model, characterized in that: The following steps are involved: Acquire data from a wearable device to obtain an original data sequence, wherein the data from the wearable device includes at least a pressure signal, a blood pressure signal, acceleration, and angular velocity; Preprocess the original data sequence to obtain a standard data sequence; The features of the standard pressure signal sequence and the standard blood pressure sequence are extracted based on the peak and valley values, respectively, to obtain the pressure feature sequence, the blood pressure feature sequence and the pulse wave propagation time series; The features of the standard acceleration sequence and the standard angular velocity sequence are extracted based on the change rate and frequency to obtain the motion intensity sequence; Fusing pressure feature sequence, blood pressure feature sequence, pulse wave propagation time sequence and exercise intensity sequence to form a multi-dimensional feature sequence; The blood pressure prediction pre-training model is trained based on the multidimensional feature sequence to obtain a blood pressure prediction model.

2. The blood pressure prediction model training method according to claim 1, characterized in that: The pressure characteristic sequence includes at least a pressure peak sequence, a pressure valley sequence, a pressure rise time sequence, and a pressure fall time sequence. Obtaining the pressure characteristic sequence includes the following steps: Extracting the peak and valley values ​​of each heartbeat cycle from the standard pressure signal sequence to obtain a pressure peak-valley value sequence, a pressure peak value sequence, and a pressure valley value sequence; The timestamp of the target peak value in the pressure peak-valley value sequence is subtracted from the timestamp of the valley value before the target peak value to obtain the pressure rise time series; The timestamp of the target valley value in the pressure peak-valley value sequence is subtracted from the timestamp of the peak value before the target valley value to obtain the pressure drop time series.

3. The blood pressure prediction model training method according to claim 1, characterized in that: The blood pressure characteristic sequence includes at least a blood voltage peak sequence, a blood voltage rising rate sequence, and a blood voltage falling rate sequence. Obtaining the blood pressure characteristic sequence includes the following steps: Extract the peak and valley values ​​of each blood volume change cycle from the standard blood pressure sequence to obtain a blood voltage peak-valley value sequence and a blood voltage peak value sequence; Calculating the rate at which the blood voltage rises from a valley value to a subsequent peak value in the blood voltage peak-valley value sequence to obtain a blood voltage rise rate sequence; The rate at which the blood voltage drops from a peak value to a subsequent valley value in the blood voltage peak-valley value sequence is calculated to obtain a blood voltage drop rate sequence.

4. The blood pressure prediction model training method according to claim 1, characterized in that: Obtaining the pulse wave propagation time series includes the following steps: Extract the peak value of each heartbeat cycle from the standard pressure signal sequence to obtain a pressure peak sequence; Extract the peak value of each blood volume change cycle from the standard blood pressure sequence to obtain the blood voltage peak sequence; The pressure peak sequence and the blood voltage peak sequence are subjected to time stamp-based difference processing to obtain the pulse wave propagation time series.

5. The blood pressure prediction model training method according to claim 1, characterized in that: The exercise intensity sequence includes at least an exercise intensity level sequence, an acceleration change rate sequence, and an exercise frequency sequence. Obtaining the exercise intensity sequence includes the following steps: Perform rate-based summation on the standard acceleration sequence and the standard angular velocity sequence to obtain a comprehensive rate sequence; Obtaining an exercise intensity level sequence based on the comprehensive rate sequence, the first preset rate threshold, and the second preset rate threshold; Calculate the rate of change of acceleration in the standard acceleration sequence to obtain the acceleration change rate sequence; Based on Fourier transform, the frequency of each preset period of the standard acceleration sequence is extracted to obtain the acceleration frequency characteristic sequence; The maximum frequency in each preset period of the acceleration frequency characteristic sequence is extracted, that is, the motion frequency, to obtain a motion frequency sequence.

6. The blood pressure prediction model training method according to claim 1, characterized in that: The blood pressure prediction pre-training model includes a convolutional recursive fusion module, an attention weighting module and a prediction module; The convolution-recursion fusion module performs convolution operation and bidirectional recursion operation on the input vector respectively to obtain a convolution feature sequence and a bidirectional feature sequence; and fuses the convolution feature sequence and the bidirectional feature sequence to obtain a comprehensive feature sequence; The attention weighting module performs attention mechanism processing on the comprehensive feature sequence to obtain an attention weight sequence, and performs weighted processing on the comprehensive feature sequence and the attention weight sequence to obtain a weighted feature sequence; The prediction module maps the weighted feature sequence through a fully connected layer to obtain a predicted value of the blood pressure data; The predicted value of blood pressure data is expressed as follows: in, represents the predicted value of systolic blood pressure, represents the predicted value of diastolic blood pressure, represents the matrix transpose, represents the weight of the fully connected layer, represents the weighted feature sequence, represents the bias of the fully connected layer.

7. A blood pressure data prediction method, characterized in that: The following steps are involved: Obtain target data from wearable devices; Processing the target data based on a blood pressure prediction model to obtain a predicted value of the blood pressure data; The blood pressure prediction model is a model obtained by using the blood pressure prediction model training method according to any one of claims 1 to 6.

8. A blood pressure prediction model training system, characterized in that: A training method for implementing the blood pressure prediction model according to any one of claims 1 to 6, wherein the training system comprises: A data acquisition module, configured to acquire data from a wearable device to obtain an original data sequence, wherein the data from the wearable device includes at least a pressure signal, a blood pressure signal, acceleration, and angular velocity; The data processing module is used to preprocess the original data sequence to obtain a standard data sequence; A first feature extraction module is used to extract features of the standard pressure signal sequence and the standard blood pressure sequence based on peak values ​​and valley values, respectively, to obtain a pressure feature sequence, a blood pressure feature sequence, and a pulse wave propagation time sequence; The second feature extraction module is used to extract features of the standard acceleration sequence and the standard angular velocity sequence based on the change rate and frequency to obtain a motion intensity sequence; Feature fusion module, used to fuse pressure feature sequence, blood pressure feature sequence, pulse wave propagation time sequence and exercise intensity sequence to form a multi-dimensional feature sequence; The model training module is used to train the blood pressure prediction pre-training model based on the multi-dimensional feature sequence to obtain the blood pressure prediction model.

9. A blood pressure data prediction system, characterized in that: The prediction system includes: An acquisition module is used to obtain target data of the wearable device; A prediction module is used to infer the target data based on a blood pressure prediction model to obtain a predicted value of the blood pressure data; wherein the blood pressure prediction model is a model obtained by using the training method of the blood pressure prediction model according to any one of claims 1 to 6.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the training method of the blood pressure prediction model according to any one of claims 1 to 6 is implemented.

11. A blood pressure prediction model training device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the training method of the blood pressure prediction model according to any one of claims 1 to 6 is implemented.

Images