Precise blood pressure measuring method based on portable pulse diagnosis instrument

Abstract

According to the precise blood pressure measuring method based on the portable pulse diagnosis instrument, the pulse and blood pressure data of the subject are collected, and systematic data processing and analysis are carried out, so that precise prediction of the blood pressure is achieved. The method comprises the following steps: collecting left and right hand pulse data by using a portable pulse diagnosis instrument, and synchronously collecting blood pressure data to establish a mapping relation between pulse and blood pressure; the data processing step comprises data cleaning, Shannon sampling and wavelet threshold denoising algorithm, and accuracy and reliability of pulse signals are ensured. By constructing a machine learning regression model improved based on Hunter Prey Optimization (HPO) and combining an XGBoost algorithm for prediction, the generalization ability and stability of the model are improved. According to the model, a mean square error (MSE) is adopted as a loss function, iterative optimization is performed through an adaptive learning rate adjustment strategy, overfitting is effectively reduced, and prediction precision is improved. Finally, blood pressure prediction is conducted on the pulse data of the subject through the model, error analysis is conducted, and the clinical application effect of the model is verified. The system has high efficiency, accuracy and stability, and a non-contact and convenient technical scheme is provided for portable blood pressure measurement.

Description

Technical Field

[0001] The present invention relates to the technical field of computer vision applications; in particular, it relates to a method for accurately measuring blood pressure based on a portable pulse diagnosis instrument. Technical Background

[0002] Hypertension, as a common chronic disease, is an important risk factor for cardiovascular and cerebrovascular diseases. Therefore, accurate and convenient blood pressure monitoring is crucial for the prevention and control of hypertension. However, traditional blood pressure measurement methods mainly rely on mercury sphygmomanometers or electronic blood pressure monitors. Such devices are usually large in size, poor in portability, and require professional guidance for operation, making it difficult to achieve daily convenient monitoring. To meet the growing personalized health needs of people, portable blood pressure monitoring devices have gradually attracted attention in recent years.

[0003] Most of the existing portable blood pressure measurement devices are based on the principle of cuff pressurization. Although the size has been reduced, wearing and operation are still relatively complex, and they are easily affected by factors such as posture and wearing position, resulting in unstable measurement accuracy. Some cuffless photoelectric blood pressure measurement devices have improved portability, but problems such as data accuracy and device cost limit their popular application in home and daily health monitoring.

[0004] Pulse diagnosis is an important method in traditional Chinese medicine diagnosis for judging the health status of the human body. By touching the patient's pulse and observing the characteristics of the pulse, such as strength, depth, and speed, traditional Chinese medicine can judge the patient's blood pressure status. According to different manifestations of the pulse, traditional Chinese medicine can identify the running status of qi and blood, as well as the health of the heart and blood vessels. Modern pulse diagnosis instruments combine traditional pulse diagnosis methods with sensor technology, and analyze indicators such as the frequency, amplitude, and waveform changes of the pulse through data analysis algorithms to help judge the blood pressure status in a digital and objective manner. This method can continuously monitor pulse fluctuations, quantify pulse diagnosis information, and assist in judging blood pressure conditions, providing scientific support for health management.

[0005] The present invention proposes a blood pressure measurement method based on a portable pulse diagnosis instrument. This method combines the advantages of traditional Chinese medicine pulse diagnosis and modern sensor technology, accurately collects pulse data through highly sensitive inertial sensors, and combines machine learning algorithms for blood pressure measurement, realizing a cuffless portable design for convenient daily use by users. At the same time, the pulse diagnosis instrument can capture the subtle changes in pulse fluctuations, digitalize and standardize the data, greatly improving the objectivity and consistency of measurement. Compared with traditional cuff-type blood pressure monitors, this portable pulse diagnosis instrument not only does not need to pressurize the arm, avoiding discomfort during use, but also reduces external interference by directly reading pulse characteristics, providing continuous and real-time blood pressure data, which is suitable for dynamic monitoring and trend analysis. While improving the user experience, this method takes into account the accuracy and portability of blood pressure measurement, providing an efficient and innovative solution for personalized health management and disease prevention. Summary of the Invention

[0006] The present invention proposes a method for accurately measuring blood pressure based on a portable pulse diagnosis instrument. This method processes the data by collecting the pulse data and blood pressure data of the subject, and using techniques such as data cleaning, Shannon sampling, and wavelet threshold denoising algorithms. And by constructing a machine learning regression model improved based on Hunter Prey Optimization (HPO) to extract the characteristics of pulse data, accurate blood pressure prediction is achieved. This method has high efficiency, accuracy, and stability, and can provide reliable support for clinical blood pressure measurement.

[0007] The specific technical solution of the present invention is as follows:

[0008] Collect the left and right hand pulse data of the subject through a portable pulse diagnosis instrument, and synchronously collect the corresponding blood pressure data through an OMRON U30 electronic sphygmomanometer to establish a mapping relationship between the pulse and blood pressure. At the same time, record the blood pressure measurement time, pulse data, and the physiological state of the subject to provide background information to assist data analysis.

[0009] Clean the collected pulse data, set the upper and lower threshold values, remove outliers and noise interference to ensure the accuracy of the data. Based on the Shannon sampling theorem, set an appropriate sampling frequency, and retain the pulse signal in the frequency band from 0 to 50 Hz to improve the accuracy and reliability of the signal. Use the wavelet threshold denoising algorithm to filter the pulse data to remove noise and obtain a smooth and accurate pulse signal. By selecting the bior3.5 basis wavelet for six-layer wavelet decomposition, extract the signal characteristics of different frequency bands, and use the soft threshold function for signal smoothing to reduce the influence of noise. In addition, dynamically adjust the threshold size according to the noise level and signal characteristics to optimize the noise removal effect.

[0010] Divide the obtained final pulse data set into a training set and a test set according to a ratio of 8:2 to ensure the sample independence of model training and testing and enhance the generalization ability of the model.

[0011] Construct a machine learning regression model based on the HPO algorithm and combine it with the eXtreme Gradient Boosting (XGBoost) regression algorithm to improve the prediction performance of the model. Iteratively optimize the parameters of the improved regression model, and apply the adaptive learning rate dynamic adjustment method to reduce overfitting and improve the stability and accuracy of the model. Use the mean square error (MSE) as the loss function, and optimize the model training process by adjusting the learning rate decay strategy.

[0012] Input the training set into the improved regression model for training to improve the prediction accuracy of the model. Use the trained regression model to predict the blood pressure of the subjects' pulse data in the test set, compare the prediction results with the actual blood pressure, and conduct error analysis to provide reliable prediction support for clinical applications.

[0013] Through the above technical solutions, the present invention realizes accurate blood pressure prediction based on pulse data, has high measurement accuracy and stability, can be widely applied in portable devices, and provides technical support for non-contact and convenient blood pressure measurement.

[0014] Beneficial technical effects

[0015] 1. The present invention collects the pulse data and blood pressure data of the subjects through a portable pulse diagnosis instrument, and uses the Shannon sampling theorem and frequency band selection method to retain the key frequency band information (0 to 50 Hz) of the pulse signal, effectively avoiding the influence of low-frequency noise and high-frequency interference on the blood pressure prediction results, providing a good data basis for the accurate extraction and subsequent analysis of the signal, and ensuring the stability of blood pressure prediction.

[0016] 2. The present invention uses the wavelet threshold denoising algorithm to process the pulse data, successfully extracts high-quality pulse signals, removes outliers and noise interference, provides accurate and reliable input data for the subsequent blood pressure prediction model, and significantly improves the accuracy and precision of blood pressure measurement.

[0017] 3. The present invention combines the HPO algorithm with XGBoost regression to construct an efficient blood pressure prediction model, and adopts an adaptive learning rate optimization strategy to improve the generalization ability and robustness of the model, reduce the overfitting phenomenon, and ensure the blood pressure prediction accuracy under different individuals and physiological states.

[0018] 4. The present invention first applies the machine learning regression model to the accurate measurement of blood pressure by a portable pulse diagnosis instrument. By automatically extracting the characteristics of pulse data, it successfully eliminates the influence of environmental interference, measurement errors and data noise on the prediction results, provides an efficient, intelligent and stable blood pressure prediction method, and provides a new technical solution for clinical and convenient blood pressure measurement.

[0019] 5. The present invention breaks through the limitations of traditional blood pressure measurement methods, uses a portable pulse diagnosis instrument for non-invasive blood pressure prediction, improves the measurement accuracy while reducing the cost and improving the efficiency, and is particularly suitable for large-scale screening, elderly health monitoring and daily convenient blood pressure management. Description of the drawings

[0020] Figure 1 It is a flowchart of the accurate blood pressure measurement method based on a portable pulse diagnosis instrument provided by an embodiment of the present invention;

[0021] Figure 2 Schematic diagram of the structure of a portable pulse diagnosis instrument for left - hand pulse detection adopted by the present invention.

[0022] Wherein: 1 is an inertial sensor located at the left cun pulse; 2 is an inertial sensor located at the left guan pulse; 3 is an inertial sensor located at the left chi pulse; 4 is a PCB circuit board equipped with a Bluetooth chip; 5 is a lithium battery for powering the pulse diagnosis instrument; 6 is an FPC flexible circuit board;

[0023] Figure 3 OMRON U30 electronic sphygmomanometer adopted by the present invention;

[0024] Figure 4 Flowchart of the wavelet threshold denoising algorithm adopted by the present invention;

[0025] Figure 5 Data comparison chart before and after the wavelet threshold denoising algorithm adopted by the present invention;

[0026] Figure 6 Architectural diagram of the machine - learning model combining the HPO algorithm and XGBoost regression adopted by the present invention;

[0027] Figure 7 Comparison result of the true blood - pressure value and the predicted value after regression prediction using the machine - learning model adopted by the present invention; Specific implementation manner

[0028] The following further elaborates on the present invention in detail in conjunction with the attached drawings and embodiments.

[0029] Figure 1 Flowchart of the method for accurate blood - pressure measurement based on a portable pulse diagnosis instrument. The specific steps are as follows:

[0030] Step 1: Obtain the pulse and blood - pressure data of the subject's left and right hands, which specifically includes the following steps:

[0031] Step 1.1: As Figure 2 shown, collect the left - and right - hand pulse data of 100 subjects through a portable pulse diagnosis instrument. The pulse diagnosis instrument accurately records the pulse - waveform data through sensors, and the sampling frequency is set to 100 Hz to ensure capturing the subtle changes of the pulse. Each subject will have their pulse collected on both the left and right hands respectively to ensure comparative analysis of the differences in the left - and right - hand pulses. Meanwhile, Figure 3 as shown in the electronic sphygmomanometer, blood - pressure measurement will be carried out synchronously. The airbag pressurizes the subject's upper arm to measure the systolic and diastolic blood - pressure values, and the results will be stored together with the pulse data to establish the corresponding relationship between the pulse and blood pressure.

[0032] Step 1.2: Record the blood pressure measurement time, pulse data, and corresponding physiological states of the subjects to provide background information for data analysis. In addition to recording the time of each blood pressure measurement, it is also necessary to record in detail the physiological states of the subjects during the measurement process, such as emotional states (relaxed, tense, etc.) and activity states before measurement (whether exercising, etc.). These information help to analyze the potential relationship between blood pressure and pulse waveforms and provide more comprehensive background information for subsequent data modeling. By combining physiological state data, the accuracy of the model can be further optimized and the reliability of blood pressure prediction can be improved.

[0033] Step 2: Preprocess the left and right hand pulse data obtained in Step 1 to obtain the final pulse dataset, which specifically includes the following steps:

[0034] Step 2.1: Clean the pulse data collected in Step 1.1, set the upper and lower threshold values of the pulse, and remove outliers and interference factors to ensure the accuracy of the data. First, through the time-domain analysis of the pulse signal, identify and eliminate the sudden abnormal data caused by equipment failures or external interferences, such as the maximum or minimum values in the pulse data, which usually exceed the reasonable physiological range. Set the upper and lower threshold values of the pulse signal, usually set according to the normal physiological range (60 to 100 beats per minute). Through data screening, remove abnormal fluctuations and interferences caused by misoperations or environmental factors to ensure the purity of the data. In addition, by observing the periodicity and regularity of the pulse waveform, eliminate sudden instantaneous signal errors to ensure that the data quality meets the requirements of subsequent analysis.

[0035] Step 2.2: Set the sampling frequency based on the Shannon sampling theorem, and retain the pulse signals in the 0 to 50 Hz frequency band of the data obtained in Step 2.1 to improve the accuracy and reliability of pulse detection. According to the Shannon sampling theorem, the sampling frequency needs to be at least twice the signal frequency to avoid aliasing. In the case of pulse signals, the frequency of physiological pulses is usually between 0 and 2 Hz, and since the components of the pulse waveform usually include lower frequency components (such as breathing, body position changes, etc.), the sampling rate is set above 100 Hz to ensure that various changes in the pulse signal can be completely collected. By filtering, retaining the pulse signals between 0 and 50 Hz can effectively eliminate interference signals outside the frequency band and further improve the clarity of the pulse signal.

[0036] Step 2.3: Use the wavelet threshold denoising algorithm to filter the pulse data in Step 2.2. The wavelet threshold denoising algorithm is a signal denoising method based on wavelet transform. It separates noise and effective signals through threshold processing to improve data quality and obtain smooth and accurate pulse signals. The flow chart of the wavelet threshold denoising algorithm is as Figure 4As shown. In this process, the pulse signal can be decomposed into different frequency sub-bands through wavelet transform, and then noise suppression can be carried out on each sub-band. First, perform multi-layer wavelet decomposition on the signal, and gradually extract different characteristic information from low frequency to high frequency. Next, use the threshold denoising algorithm to denoise the decomposed signal, and use the wavelet function to maximize the retention of the useful components of the signal while removing noise to avoid distortion. The specific steps are as follows:

[0037] Step 2.3.1: To evaluate the noise reduction effect of different wavelet bases at different decomposition levels, the signal-to-noise ratio (SNR) and root mean square error (RMSE) are used to evaluate the signal. SNR is the ratio of signal power to noise power, which is used to measure the signal quality. RMSE is the square root of the mean square of the deviation between the predicted value and the true value, which is used to evaluate the prediction accuracy of the model. By using these two methods, the noise reduction performance can be objectively measured to ensure that the signal quality is effectively improved. Among them, the SNR formula is shown in Equation (1), and the RMSE formula is shown in Equation (2):

[0038]

[0039] Among them, x(i) is the original signal component, and y(i) is the denoised signal component. For the bior3.5 wavelet basis: when using the SNR evaluation method, the result at level 2 is 31.1, at level 4 is 30.9, and at level 6 is 31.5. When using the RMSE evaluation method, the result at level 2 is 3.96654, at level 4 is 4.57983, and at level 6 is 2.05465. The larger the value of SNR, the better the denoising effect; the smaller the RMSE, the closer the actually reconstructed signal is to the original signal, and the better the denoising effect.

[0040] Step 2.3.2: To better extract the signal characteristics of different frequency bands, the present invention uses the bior3.5 wavelet basis to perform six-layer wavelet decomposition on the pulse signal. The bior3.5 wavelet basis is a biorthogonal wavelet, which has excellent time-frequency localization ability and linear phase characteristics. Its asymmetric design can effectively extract the local characteristics of non-stationary signals such as pulse waves. In this process, the pulse signal undergoes six-layer wavelet decomposition, and the signal is decomposed into multiple frequency bands, so that the wavelet coefficients of each layer can represent the characteristics of the pulse signal at different scales. Through this decomposition method, the important characteristics and noise in the pulse signal can be more accurately identified and separated, providing an effective basis for subsequent noise removal and signal smoothing.

[0041] Step 2.3.3: Use a soft-thresholding function to smoothly reduce the signal amplitude close to the threshold in Step 2.3.2, thereby reducing the influence of noise and making the signal transition smoother. The soft-thresholding function is a commonly used wavelet denoising method. It reduces the interference of noise on the signal by smoothing the wavelet coefficients. The formula of the soft-thresholding function is shown in Equation (3).

[0042]

[0043] Where t is the wavelet coefficient and λ is the threshold, representing the degree of denoising. In this process, first process each layer of wavelet coefficients according to the preset threshold. Zeroize the coefficients below the threshold, and smoothly reduce the amplitude of the coefficients above the threshold, thereby removing the influence of noise. This operation can effectively suppress high-frequency noise, make the transition of the pulse signal smoother, and at the same time retain the information of the effective signal.

[0044] Step 2.3.4: By calculating the noise estimation value of each layer, dynamically adjust the threshold size in Step 2.3.3 according to the noise level and signal characteristics, so that more noise can be removed in the high-frequency layer, while more signal information can be retained in the low-frequency layer. In the denoising process, the estimated value of noise is very important for adjusting the threshold. According to the distribution of each layer of wavelet coefficients, first estimate the noise, usually represented by the standard deviation of the signal. For the high-frequency part, there is more noise component, so a lower threshold needs to be set to effectively remove the noise. For the low-frequency part, the main components of the signal usually exist in the low-frequency band, so a higher threshold is needed to retain more signal information. In this way, the threshold can be dynamically adjusted according to the noise characteristics of different frequency components to achieve a more accurate denoising effect. The comparison chart of data before and after the wavelet threshold denoising algorithm is as Figure 5 shown.

[0045] Step 3: Divide the final pulse data set obtained in Step 2 into a training set and a test set according to a ratio of 8:2, ensure the independence of the samples for model training and testing, and enhance the generalization ability of the model.

[0046] Step 4: Build a machine learning regression model improved based on HPO to improve the accuracy and stability of blood pressure prediction. The specific implementation includes the following steps:

[0047] Step 4.1: Construct a machine learning regression model based on the HPO algorithm. The HPO algorithm is a bionic optimization algorithm that simulates the behavioral strategies between predators and prey. It can search for the optimal solution in high-dimensional space and improve the convergence speed and global search ability of the algorithm through group cooperation and competition mechanisms. First, define the model structure and optimization process of HPO and use it for model parameter selection in the blood pressure regression task. The HPO algorithm screens the optimal individuals through the fitness evaluation method and iteratively searches for the best parameter combination, enabling the regression model to fully fit the characteristics of blood pressure data. Through the global search ability of the HPO algorithm, it can effectively avoid the local optimal trap in traditional algorithms and ensure better prediction results in complex datasets.

[0048] Step 4.2: Combine the HPO algorithm in Step 4.1 with the XGBoost regression model to effectively improve the performance of the model. The machine learning model framework diagram is as Figure 6 shown. XGBoost is a high-performance machine learning algorithm based on the gradient boosting framework. It significantly improves the prediction accuracy and generalization ability of the model through parallel computing and regularization optimization. With its efficient distributed computing architecture and built-in feature importance evaluation mechanism, this algorithm demonstrates excellent performance in regression tasks. By combining HPO with XGBoost and using HPO for parameter optimization of XGBoost, the prediction performance of the model is enhanced. The global optimization feature of HPO brings higher accuracy to XGBoost, while enhancing the convergence and stability of the algorithm, enabling the model to maintain a high prediction accuracy under different physiological states and interference environments.

[0049] Step 4.3: Iteratively optimize the parameters of the improved machine learning regression model in Step 4.2 and apply an adaptive learning rate to dynamically adjust the learning rate, reduce overfitting, and improve the stability of the model. The adaptive learning rate adopts the dynamic decay mechanism shown in formula (4):

[0050]

[0051] where new_lr is the new learning rate, initial_lr is the initial learning rate, γ is the decay factor, epoch is the current training epoch, and step_size is the interval for learning rate update. By dynamically adjusting the learning rate, the model can converge quickly in the early learning stage and gradually reduce the learning rate in the later stage to avoid parameter fluctuations, thereby reducing the overfitting problem. At the same time, through the adaptive learning rate strategy, the model can dynamically adjust the optimization step size according to different data complexities, making the model training more stable and improving the overall prediction performance. In the selection of the loss function, MSE is used as the evaluation criterion. MSE is the average of the squared differences between the predicted value and the true value and is used to measure the error size of the prediction model, expressed by formula (5):

[0052]

[0053] where \(n\) is the number of samples, \(x(i)\) is the true value, is the predicted value of the model. MSE can effectively reflect the fitting degree of the model to the real data. As the optimization objective of the regression model, it helps the HPO algorithm to finely adjust the parameters, enabling the model to better capture the subtle differences in blood pressure data. In this way, the improved model not only achieves high-precision blood pressure prediction but also further enhances its anti-interference ability and robustness in complex scenarios.

[0054] Step 5: Use the trained machine learning regression model to predict the blood pressure of the subjects. Specifically, it includes the following steps:

[0055] Step 5.1: Input the training set preprocessed in Step 3 into the optimized regression model in Step 4.3. Through multiple iterative trainings, the model gradually enhances its ability to identify and extract the characteristics of the pulse signal. A strategy of dynamically adjusting the learning rate is adopted to improve the training efficiency, and the sample ratio is controlled in each training batch so that the samples with systolic blood pressure less than 120 mmHg, 120 - 140 mmHg, and greater than 140 mmHg account for 30%, 50%, and 20% respectively to optimize the data distribution. When the loss reduction rate for 10 consecutive iterations is less than 1%, it is determined that the model converges and the training is terminated. During the training process, the convergence of the model is judged by monitoring the downward trend of the loss function to ensure that the model stops training after reaching the best prediction performance, thereby improving the prediction accuracy of blood pressure.

[0056] Step 5.2: Use the trained regression model to predict the blood pressure of the pulse data of the subjects in the test set divided in Step 3, and post-process the prediction results. Automatically eliminate the abnormal prediction values with systolic blood pressure less than 70 mmHg or greater than 200 mmHg and retest them. At the same time, calculate the average value of the continuous 5 prediction values as the final output. Compare the processed prediction results with the actual blood pressure to evaluate the generalization performance and clinical applicability of the model. The comparison results between the true value and the predicted value of the blood pressure after regression prediction using the machine learning model are as Figure 7 shown. Compare the predicted values with the true blood pressure values of the subjects one by one, and evaluate the accuracy and stability of the prediction results through the mean absolute error (MAE). MAE is the average of the absolute differences between the predicted value and the true value, used to measure the accuracy of the prediction, and is expressed by formula (6).

[0057]

[0058] After performing regression prediction using a machine learning model, the MAEs of three algorithms were compared. Among them, the MAE of the Gradient Boosting algorithm was 3.98, the MAE of the Random Forests algorithm was 3.6, and the MAE of the HPO+XGBoost algorithm was 3.07. It can be seen from the results that the HPO+XGBoost algorithm has higher accuracy and stability in predicting blood pressure values.

Claims

1. A method for accurately measuring blood pressure based on a portable pulse diagnosis instrument, characterized in that: It includes the following steps: Step 1: Obtain the pulse and blood pressure data of the subject's left and right hands; Step 2: Preprocess the pulse data of the left and right hands obtained in Step 1 to obtain the final pulse data set; Step 3: Divide the final pulse data set obtained in Step 2 into a training set and a test set at a ratio of 8:2 to ensure the independence of the samples for model training and testing and enhance the generalization ability of the model; Step 4: Build a machine learning regression model improved based on Hunter Prey Optimization (HPO) to improve the accuracy and stability of blood pressure prediction; Step 5: Use the trained machine learning regression model to predict the blood pressure of the subject.

2. The method for accurately measuring blood pressure based on a portable pulse diagnosis instrument according to claim 1, wherein: The specific process of Step 1 includes the following steps: Step 1.1: Collect the pulse waveform data of the subject's left and right hands through a portable pulse diagnosis instrument and synchronously collect the corresponding blood pressure data to establish the corresponding relationship between the pulse and blood pressure; Step 1.2: Record the blood pressure measurement time, pulse data and corresponding physiological status of the subject to provide background information for data analysis.

3. The method for accurately measuring blood pressure based on a portable pulse diagnosis instrument according to claim 1, wherein: The specific process of Step 2 includes the following steps: Step 2.1: Clean the pulse data collected in Step 1.1, set the upper and lower threshold values of the pulse, remove outliers and interference factors to ensure the accuracy of the data; Step 2.2: Set the sampling frequency based on the Shannon sampling theorem, and retain the pulse signals in the frequency band of 0 to 50 Hz in the data obtained in Step 2.1 to improve the accuracy and reliability of pulse detection. Step 2.3: Use the wavelet threshold denoising algorithm to filter the pulse data in Step 2.2 to remove noise and obtain a smooth and accurate pulse signal; Step 2.3.1: Use two methods, signal-to-noise ratio (SNR) and root mean square error (RMSE), to evaluate the denoising effect of different wavelet bases at different decomposition levels, so as to objectively measure the denoising performance and ensure that the quality of the signal is effectively improved. Among them, the SNR formula is shown in Equation (1), and the RMSE formula is shown in Equation (2): Where x(i) is the original signal component and y(i) is the denoised signal component. Step 2.3.2: Perform six-layer multi-scale decomposition on the preprocessed pulse signal using the bior3.5 wavelet to effectively extract signal features through its excellent frequency band division ability. This decomposition strategy can accurately separate the physiological characteristics and noise components of different frequency bands, providing a reliable signal basis for subsequent denoising processing and feature analysis. Step 2.3.3: Use the soft threshold function to smoothly reduce the signal amplitude close to the threshold in Step 2.3.2, thereby reducing the influence of noise and making the signal transition smoother. The soft threshold function formula is shown in Equation (3). Where x is the wavelet coefficient and λ is the threshold, representing the degree of denoising. Step 2.3.4: By calculating the noise estimation value of each layer, dynamically adjust the threshold size in Step 2.3.3 according to the noise level and signal characteristics, so that more noise can be removed in the high-frequency layer, while more signal information can be retained in the low-frequency layer.

4. The method for accurately measuring blood pressure based on a portable pulse diagnosis instrument according to claim 1, wherein: The specific process of Step 4 includes the following steps: Step 4.1: Construct a machine learning regression model based on the HPO algorithm and optimize the weight parameters of the model to improve the convergence speed and accuracy of the model. Step 4.2: Combine the HPO algorithm in Step 4.1 with the eXtreme Gradient Boosting (XGBoost) regression model to effectively improve the performance of the model. Step 4.3: Perform parameter iteration optimization on the improved machine learning regression model in Step 4.2, and apply an adaptive learning rate to dynamically adjust the learning rate, reduce overfitting, and improve the stability of the model. The adaptive learning rate adopts the dynamic decay mechanism shown in formula (4): where new_lr is the new learning rate, initial_lr is the initial learning rate, γ is the decay factor, epoch is the current training epoch, and step_size is the interval for learning rate update. By dynamically adjusting the learning rate, the model can converge quickly in the early learning stage and gradually decrease the learning rate in the later stage to avoid parameter fluctuations, thereby reducing the overfitting problem. At the same time, through the adaptive learning rate strategy, the model can dynamically adjust the optimization step size according to different data complexities, making the model training more stable and improving the overall prediction performance. In the selection of the loss function, the mean squared error (MSE) is used as the evaluation criterion, which is expressed by formula (5): where n is the number of samples, and y i is the true value, is the predicted value of the model. MSE can effectively reflect the fitting degree of the model to the real data. As the optimization objective of the regression model, it helps the HPO algorithm to finely adjust the parameters, enabling the model to better capture the subtle differences in blood pressure data. In this way, the improved model not only achieves high-precision blood pressure prediction but also further enhances its anti-interference ability and robustness in complex scenarios.

5. The method for accurately measuring blood pressure based on a portable pulse diagnosis instrument according to claim 1, characterized in that: The specific process of Step 5 includes the following steps: Step 5.1: Input the training set obtained by preprocessing into the optimized regression model in Step 4.3 to further optimize the prediction accuracy of the model. Through multiple iterative trainings, the model gradually enhances its ability to identify and extract pulse signal features. The strategy of dynamically adjusting the learning rate is adopted to improve the training efficiency, and the sample ratio is controlled in each training batch so that the samples with systolic blood pressure less than 120 mmHg, 120 - 140 mmHg, and greater than 140 mmHg account for 30%, 50%, and 20% respectively to optimize the data distribution. When the loss reduction rate for 10 consecutive iterations is less than 1%, it is determined that the model has converged and the training is terminated. These measures together improve the prediction accuracy of blood pressure. During the training process, the convergence of the model is also judged by monitoring the decreasing trend of the loss function to ensure that the model stops training after reaching the best prediction performance. Step 5.2: Use the trained regression model to predict the blood pressure of the pulse data of the subjects in the test set divided in Step 3, and post-process the prediction results. Automatically eliminate the abnormal prediction values with systolic blood pressure less than 70 mmHg or greater than 200 mmHg and retest them. At the same time, calculate the average value of the continuous 5 prediction values as the final output. Compare the processed prediction results with the actual blood pressure to evaluate the generalization performance and clinical applicability of the model. Compare the predicted values with the true blood pressure values of the subjects one by one, and evaluate the accuracy and stability of the prediction results through the mean absolute error (MAE).

Images