Intelligent blood pressure measuring method and system based on electronic Korotkoff sound method

Abstract

The invention provides an intelligent blood pressure measuring method and system based on an electronic Korotkoff sound method, and relates to the technical field of medical instruments.The method comprises the steps that in the process that a cuff wound around the upper arm of a person to be measured is deflated, multiple Korotkoff sound signals from a brachial artery and the air pressure, corresponding to the Korotkoff sound signals, in the cuff are obtained; determining the initial systolic pressure of the to-be-tested person according to the first air pressure in the cuff when the Korotkoff sound signal appears for the first time, and determining the initial diastolic pressure of the to-be-tested person according to the second air pressure in the cuff when the Korotkoff sound signal appears for the last time; performing feature value extraction on the plurality of Korotkoff sound signals to obtain a plurality of feature parameters; and inputting the plurality of characteristic parameters, the initial systolic pressure and the initial diastolic pressure into a preset neural network model to obtain target systolic pressure and target diastolic pressure of the to-be-measured person, and taking the target systolic pressure and the target diastolic pressure as blood pressure measurement results of the to-be-measured person. The method has the technical effect that the accuracy of blood pressure measurement is improved.

Description

Technical Field

[0001] The present application relates to the technical field of medical devices, and in particular to an intelligent blood pressure measurement method and system based on an electronic Korotkoff sound method. Background Art

[0002] Blood pressure is one of the important physiological indicators of the human body, and its measurement accuracy is directly related to the diagnosis and treatment of diseases. At present, electronic sphygmomanometers based on the Korotkoff sound method are widely used because of their simple operation and low cost. However, during the measurement process, they are easily affected by various factors such as environmental noise, human activities, and measurement posture, making it difficult to ensure the accuracy and stability of the measurement results.

[0003] In the prior art, a method of setting a fixed threshold is usually used to identify Korotkoff sound signals and determine blood pressure values. That is, when the detected signal amplitude exceeds the preset threshold, it is determined to be a valid Korotkoff sound signal, and the corresponding cuff pressure when it first appears and disappears is used as the systolic pressure and diastolic pressure, respectively. However, due to the differences in physiological characteristics of different individuals, the intensity and characteristics of Korotkoff sound signals are also different. It is difficult to adapt to such individual differences by using a unified fixed threshold, which is easy to cause blood pressure measurement deviation. Summary of the invention

[0004] The present application provides an intelligent blood pressure measurement method based on the electronic Korotkoff sound method, which is used to improve the accuracy of blood pressure measurement.

[0005] In a first aspect, the present application provides an intelligent blood pressure measurement method based on an electronic Korotkoff sound method, the method comprising: increasing the air pressure in a cuff wrapped around the upper arm of a person to be measured to a preset pressure, and then deflating the cuff; in the process of deflating the cuff, obtaining multiple Korotkoff sound signals from the brachial artery, and the air pressure in the cuff corresponding to each of the Korotkoff sound signals; determining the initial systolic pressure of the person to be measured based on the first air pressure in the cuff when a Korotkoff sound signal appears for the first time among the multiple Korotkoff sound signals, and determining the initial diastolic pressure of the person to be measured based on the second air pressure in the cuff when a Korotkoff sound signal appears for the last time among the multiple Korotkoff sound signals; extracting feature values ​​of the multiple Korotkoff sound signals to obtain multiple feature parameters; inputting the multiple feature parameters, the initial systolic pressure, and the initial diastolic pressure into a preset neural network model to obtain the target systolic pressure and target diastolic pressure of the person to be measured, and using the target systolic pressure and the target diastolic pressure as the blood pressure measurement results of the person to be measured.

[0006] By adopting the above technical solution, by obtaining multiple Korotkoff sound signals and the corresponding cuff air pressure during the deflation process, not only can the initial systolic pressure and initial diastolic pressure be determined according to the first and last appearance times of the Korotkoff sound signals, but also the acquired Korotkoff sound signals are subjected to deep feature extraction to obtain multiple feature parameters including frequency domain eigenvalues, amplitude eigenvalues ​​and time domain eigenvalues. These extracted feature parameters are input into the preset neural network model together with the initial systolic pressure and initial diastolic pressure. Through the deep learning ability of the model, the rich feature information in the Korotkoff sound signal is fully mined, and the accurate mapping from the multidimensional feature space to the blood pressure value is realized. This method not only overcomes the problem of poor adaptability of the traditional fixed threshold method to individual differences, but also can adaptively learn the signal characteristics under different measurement conditions, thereby improving the accuracy of blood pressure measurement.

[0007] Optionally, the acquiring of multiple Korotkoff sound signals from the brachial artery, and the air pressure inside the cuff corresponding to each of the Korotkoff sound signals, includes: during the process of deflating the cuff, receiving the multiple Korotkoff sound signals generated by the pulse beating of the brachial artery from a sound sensor disposed in the cuff, and receiving the air pressure inside the cuff corresponding to each of the Korotkoff sound signals sent by a pressure sensor disposed in the cuff.

[0008] By adopting the above technical solution, the sound sensor and pressure sensor are set in the cuff to achieve the synchronous acquisition of Korotkoff sound signal and corresponding cuff air pressure. The sound sensor can accurately capture the Korotkoff sound signal generated by the beating of the brachial artery pulse, while the pressure sensor monitors the air pressure changes in the cuff in real time. The combination of the two ensures the timing of signal acquisition and the accuracy of the corresponding relationship, providing a reliable raw data basis for subsequent feature extraction and blood pressure calculation.

[0009] Optionally, the feature value extraction of the multiple Korotkoff sound signals to obtain multiple feature parameters includes: performing Fourier transform on each of the Korotkoff sound signals to obtain frequency domain feature values ​​of each of the Korotkoff sound signals; performing amplitude normalization processing on each of the Korotkoff sound signals to obtain amplitude feature values ​​of each of the Korotkoff sound signals; obtaining time domain feature values ​​of each of the Korotkoff sound signals based on the duration of each of the Korotkoff sound signals; and using the frequency domain feature values, the amplitude feature values ​​and the time domain feature values ​​as multiple feature parameters of each of the Korotkoff sound signals.

[0010] By adopting the above technical solution, the frequency domain eigenvalues, amplitude eigenvalues ​​and time domain eigenvalues ​​of the signal are obtained by performing Fourier transform, amplitude normalization processing and time feature analysis on the Korotkoff sound signal, and the multi-dimensional feature extraction of the Korotkoff sound signal is realized. This multi-dimensional feature extraction method can fully capture the frequency distribution, amplitude change and timing characteristics of the signal, so that the extracted feature parameters more completely characterize the characteristics of the Korotkoff sound signal, provide rich input information for the neural network model, and thus improve the accuracy of blood pressure measurement.

[0011] Optionally, before inputting the multiple characteristic parameters, the initial systolic pressure and the initial diastolic pressure into the preset neural network model, it also includes: acquiring multiple groups of sample data, each group of the sample data including characteristic parameters of multiple Korotkoff sound signals, initial systolic pressure, initial diastolic pressure, and standard systolic pressure and standard diastolic pressure measured by a standard blood pressure measurement device; using the characteristic parameters, the initial systolic pressure and the initial diastolic pressure in each group of the sample data as input features, and using the standard systolic pressure and the standard diastolic pressure as output features; training the neural network structure according to the input features and the output features to obtain a preset neural network model.

[0012] By adopting the above technical solution, a large amount of sample data is collected for neural network training, where each set of sample data contains complete feature parameters, initial blood pressure values ​​and standard blood pressure values, so that the neural network can learn the mapping relationship between feature parameters and actual blood pressure values. This training method based on the calibration of standard blood pressure measurement equipment enables the preset neural network model to have the ability to convert multi-dimensional features into accurate blood pressure values, effectively improving the generalization performance and prediction accuracy of the model, and ensuring the consistency of the final blood pressure measurement results with standard equipment.

[0013] Optionally, after obtaining the preset neural network model, it also includes: obtaining multiple groups of verification sample data, each group of the verification sample data includes verification characteristic parameters of multiple Korotkoff sound signals, verification initial systolic pressure, verification initial diastolic pressure, and verification standard systolic pressure and verification standard diastolic pressure measured by a standard blood pressure measurement device; inputting the verification characteristic parameters, the verification initial systolic pressure and the verification initial diastolic pressure in each group of the verification sample data into the preset neural network model to obtain predicted systolic pressure and predicted diastolic pressure; calculating a first difference between the predicted systolic pressure and the verification standard systolic pressure, and a second difference between the predicted diastolic pressure and the verification standard diastolic pressure; when the first difference and the second difference both meet the preset requirements, determining that the preset neural network model meets the usage requirements, if the first difference and / or the second difference does not meet the preset requirements, retraining the neural network model until a preset neural network model that meets the usage requirements is obtained.

[0014] By adopting the above technical solution, the performance of the trained neural network model is verified by introducing independent verification sample data, and the accuracy of the model is evaluated by calculating the difference between the predicted value and the standard value, thus establishing a strict model quality control mechanism. When the difference between the predicted systolic and diastolic blood pressure and the verification standard value does not meet the preset requirements, the model retraining process is triggered. This iterative optimization mechanism ensures that the neural network model finally put into use has reliable prediction accuracy, significantly improving the accuracy and credibility of blood pressure measurement results.

[0015] Optionally, the step of inputting the plurality of characteristic parameters, the initial systolic pressure and the initial diastolic pressure into a preset neural network model to obtain the target systolic pressure and the target diastolic pressure of the person to be tested comprises: inputting the plurality of characteristic parameters, the initial systolic pressure and the initial diastolic pressure into the preset neural network model through the input layer of the preset neural network model; performing feature extraction and nonlinear transformation on the input plurality of characteristic parameters, the initial systolic pressure and the initial diastolic pressure using the hidden layer of the preset neural network model; and obtaining the target systolic pressure and the target diastolic pressure of the person to be tested through the output layer of the preset neural network model.

[0016] By adopting the above technical solution and the three-layer structure design of the neural network model, an efficient conversion process from original features to blood pressure values ​​is achieved. The input layer receives multi-dimensional feature parameters and initial blood pressure values, the hidden layer fully explores the deep associations between features through feature extraction and nonlinear transformation, and the output layer finally generates target systolic and diastolic pressures. This hierarchical data processing mechanism can effectively capture the complex mapping relationship between feature parameters and blood pressure values, and improve the accuracy and stability of blood pressure prediction.

[0017] Optionally, after taking the target systolic pressure and the target diastolic pressure as the blood pressure measurement results of the person to be measured, it also includes: comparing the target systolic pressure and the target diastolic pressure with a preset normal range of systolic pressure and a preset normal range of diastolic pressure, respectively; when the target systolic pressure exceeds the preset normal range of systolic pressure and / or the target diastolic pressure exceeds the preset normal range of diastolic pressure, generating blood pressure abnormality prompt information; and displaying the target systolic pressure, the target diastolic pressure and the blood pressure abnormality prompt information on a display screen.

[0018] By adopting the above technical solution, the measured target systolic and diastolic blood pressures are compared with the preset normal range in real time, and prompt information is automatically generated when abnormal blood pressure is found, and the blood pressure measurement results and abnormal prompt information are simultaneously presented on the display screen, realizing the intelligent monitoring and early warning function of blood pressure measurement results. This timely abnormal prompt mechanism can help users quickly identify abnormal blood pressure conditions, help users take corresponding health management measures in a timely manner, and improve the practical value of blood pressure measurement and health monitoring effects.

[0019] In a second aspect, the present application provides an intelligent blood pressure measurement system based on the electronic Korotkoff sound method, the system comprising: a control module, an acquisition module, a determination module, an extraction module and an output module; wherein the control module is used to increase the air pressure in the cuff wrapped around the upper arm of the person to be measured to a preset pressure and then deflate the cuff; the acquisition module is used to acquire multiple Korotkoff sound signals from the brachial artery and the air pressure in the cuff corresponding to each of the Korotkoff sound signals during the process of deflation of the cuff; the determination module is used to determine the air pressure in the cuff according to the pressure in the cuff when the Korotkoff sound signal first appears in the multiple Korotkoff sound signals. The first air pressure is used to determine the initial systolic pressure of the person to be tested, and the initial diastolic pressure of the person to be tested is determined according to the second air pressure in the cuff when the Korotkoff sound signal appears last among the multiple Korotkoff sound signals; the extraction module is used to extract feature values ​​of the multiple Korotkoff sound signals to obtain multiple feature parameters; the output module is used to input the multiple feature parameters, the initial systolic pressure and the initial diastolic pressure into a preset neural network model to obtain the target systolic pressure and target diastolic pressure of the person to be tested, and use the target systolic pressure and the target diastolic pressure as the blood pressure measurement results of the person to be tested.

[0020] In a third aspect, the present application provides an electronic device, adopting the following technical solution: comprising a processor, a memory, a user interface and a network interface, the memory being used to store instructions, the user interface and the network interface being used to communicate with other devices, and the processor being used to execute the instructions stored in the memory so that the electronic device executes a computer program such as any of the above-mentioned intelligent blood pressure measurement methods based on the electronic Korotkoff sound method.

[0021] In a fourth aspect, the present application provides a computer-readable storage medium, which adopts the following technical solution: storing a computer program that can be loaded by a processor and execute any of the above-mentioned intelligent blood pressure measurement methods based on the electronic Korotkoff sound method.

[0022] In summary, the present application includes at least one of the following beneficial technical effects: 1. By acquiring multiple Korotkoff sound signals and the corresponding cuff air pressure during the deflation process, not only can the initial systolic pressure and initial diastolic pressure be determined based on the first and last occurrence times of the Korotkoff sound signals, but also the acquired Korotkoff sound signals can be subjected to deep feature extraction to obtain multiple feature parameters including frequency domain eigenvalues, amplitude eigenvalues, and time domain eigenvalues. These extracted feature parameters are input into the preset neural network model together with the initial systolic pressure and initial diastolic pressure. Through the deep learning ability of the model, the rich feature information in the Korotkoff sound signals is fully mined, and accurate mapping from multidimensional feature space to blood pressure values ​​is achieved. This method not only overcomes the problem of poor adaptability of the traditional fixed threshold method to individual differences, but can also adaptively learn the signal characteristics under different measurement conditions, thereby improving the accuracy of blood pressure measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flow chart of an intelligent blood pressure measurement method based on the electronic Korotkoff sound method provided in an embodiment of the present application; Figure 2 is a structural schematic diagram of an intelligent blood pressure measurement device based on the electronic Korotkoff sound method provided in an embodiment of the present application; Figure 3 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application.

[0024] Description of reference numerals: 1000, electronic device; 1001, processor; 1002, communication bus; 1003, user interface; 1004, network interface; 1005, memory. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0026] In the description of the embodiments of the present application, words such as "illustrative", "for example" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "illustrative", "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "illustrative", "for example" or "for example" is intended to present related concepts in a concrete way.

[0027] Figure 1 FIG. 1 is a flow chart of an intelligent blood pressure measurement method based on the electronic Korotkoff sound method provided in an embodiment of the present application. Figure 1 As shown, the method includes S101-S105: S101, after increasing the air pressure in the cuff wrapped around the upper arm of the person to be tested to a preset pressure, the cuff is deflated.

[0028] When implementing the present invention, the step of pressurization and deflation is first performed, which aims to create necessary conditions for obtaining Korotkoff sound signals later by controlling the change of air pressure in the cuff.

[0029] Specifically, the cuff of the electronic sphygmomanometer is wrapped around the upper arm of the person to be tested, so that the center of the cuff is aligned with the brachial artery. By starting the air pump, the air pressure in the cuff is quickly increased to a preset pressure, which is usually set to 180 mmHg, or increased by 20-30 mmHg according to the systolic blood pressure value of the person to be tested last measured. When the air pressure in the cuff reaches the preset value, the brachial artery is completely compressed and the blood flow is temporarily blocked. Subsequently, the solenoid valve is controlled to deflate the cuff. A stable pressure change environment is provided for the subsequent acquisition of Korotkoff sound signals, which helps to improve the accuracy of blood pressure measurement.

[0030] S102 , during the process of deflating the cuff, obtaining a plurality of Korotkoff sound signals from the brachial artery and the air pressure in the cuff corresponding to each Korotkoff sound signal.

[0031] Korotkoff sound is a characteristic sound produced by blood flow during blood pressure measurement.

[0032] When the cuff pressure is higher than the systolic pressure, the brachial artery is completely compressed and no blood flows through it, so no sound can be heard. As the cuff pressure gradually decreases and falls below the systolic pressure, a small amount of blood will flow turbulently through the partially compressed brachial artery each time the heart contracts. This turbulence causes the blood vessel wall to vibrate, producing detectable sounds, namely Korotkoff sounds.

[0033] There are five characteristic stages of Korotkoff sounds: Stage 1 (emergence stage): When the cuff pressure is slightly lower than the systolic pressure, a clear knocking sound appears for the first time, and this pressure value corresponds to the systolic pressure; Stage 2 (enhancement stage): The sound becomes louder and longer, with a gurgling sound; Stage 3 (loud stage): The sound reaches its maximum intensity; The fourth stage (weakening stage): the sound suddenly becomes dull and soft; the fifth stage (disappearance stage): the sound disappears completely, and the pressure value at this time corresponds to the diastolic pressure.

[0034] Characteristics of Korotkoff sound signals: Frequency characteristics: mainly concentrated in the range of 20-200Hz; Amplitude characteristics: the sound intensity shows a trend of first increasing and then decreasing with the change of cuff pressure; Time domain characteristics: the duration and occurrence time of Korotkoff sound in each pulse cycle.

[0035] Specifically, during the cuff deflation process, the Korotkoff sound signal and the corresponding air pressure value in the cuff need to be collected synchronously. This is because there is a direct correspondence between the appearance and disappearance of the Korotkoff sound signal and the air pressure in the cuff, and the blood pressure value can be accurately determined through this correspondence.

[0036] In the specific implementation, a sound sensor and a pressure sensor are set in the cuff, wherein the sound sensor is preferably a high-sensitivity electret microphone for collecting Korotkoff sound signals from the brachial artery; the pressure sensor uses a piezoresistive sensor with an accuracy of 0.1 mmHg for real-time monitoring of the air pressure changes in the cuff. When the air pressure in the cuff gradually decreases, the sound sensor begins to capture the Korotkoff sound signal generated when the brachial artery blood flows through, and the pressure sensor continues to collect the air pressure value in the cuff at this time. The sampling frequency is set to 1kHz to ensure that the subtle changes in the Korotkoff sound signal can be accurately captured. The collected Korotkoff sound signal is amplified by a preamplifier, and the environmental noise and other interference signals are filtered out by a bandpass filter (20Hz-200Hz). The collected Korotkoff sound signal is paired and stored with the air pressure value in the cuff at the corresponding time by a microprocessor to form time series data. This synchronous acquisition method can not only accurately record the appearance and disappearance time of the Korotkoff sound signal, but also ensure that each Korotkoff sound signal has its accurately corresponding air pressure value in the cuff, providing a reliable data basis for subsequent blood pressure calculations.

[0037] Based on the above embodiment, as an optional implementation, in S102, obtaining multiple Korotkoff sound signals from the brachial artery and the air pressure in the cuff corresponding to each Korotkoff sound signal specifically includes: During the process of deflation of the cuff, a sound sensor disposed in the cuff collects multiple Korotkoff sound signals generated by the pulse beating of the brachial artery, and a pressure sensor disposed in the cuff receives the air pressure in the cuff corresponding to each Korotkoff sound signal.

[0038] S103, determining the initial systolic pressure of the person to be measured according to the first air pressure in the cuff when a Korotkoff sound signal appears for the first time among the multiple Korotkoff sound signals, and determining the initial diastolic pressure of the person to be measured according to the second air pressure in the cuff when a Korotkoff sound signal appears for the last time among the multiple Korotkoff sound signals.

[0039] In one example, after obtaining a complete Korotkoff sound signal sequence, it is necessary to determine the initial systolic blood pressure and initial diastolic blood pressure of the person to be measured, which is a basic step in blood pressure measurement.

[0040] First, the collected Korotkoff sound signals are processed, including bandpass filtering and envelope extraction of the signals, to highlight the characteristics of Korotkoff sounds and eliminate interference. By setting an appropriate amplitude threshold (for example, 3 times the mean of the background noise), the appearance and disappearance of the Korotkoff sound signals can be accurately identified. When the first Korotkoff sound signal exceeding the threshold is detected, the first air pressure in the cuff at that moment is recorded and used as the initial systolic pressure of the person to be tested; when the last Korotkoff sound signal exceeding the threshold is detected, the second air pressure in the cuff at that moment is recorded and used as the initial diastolic pressure of the person to be tested.

[0041] In order to improve the reliability of the measurement, a verification mechanism of three consecutive cardiac cycles is used in the recognition process, that is, only when the Korotkoff sound signal is detected in three consecutive cycles can it be confirmed as a valid signal, which can effectively avoid misjudgment caused by occasional noise. This method of determining the initial blood pressure based on the characteristic points of the Korotkoff sound signal can provide reliable initial input parameters for the subsequent neural network model, while also retaining the basic principles of the traditional Korotkoff sound method measurement. The accurate determination of the initial systolic pressure and the initial diastolic pressure lays the foundation for obtaining more accurate blood pressure measurement results through the subsequent neural network model.

[0042] S104, extracting feature values ​​of the multiple Korotkoff sound signals to obtain multiple feature parameters.

[0043] Specifically, in order to fully utilize the blood pressure characteristic information contained in the Korotkoff sound signal, it is necessary to perform comprehensive feature value extraction on the acquired multiple Korotkoff sound signals. The time domain, frequency domain and amplitude characteristics of each Korotkoff sound signal are extracted to obtain a complete feature parameter set.

[0044] First, the Korotkoff sound signal is subjected to Fourier transform, the spectral characteristics of the signal are calculated, the energy distribution of the main frequency components (within the range of 20-200Hz) is extracted, and the frequency domain eigenvalues ​​are obtained, including the main frequency of the signal, the bandwidth and the energy proportion of each frequency band. Secondly, the Korotkoff sound signal is subjected to amplitude standardization processing, the signal amplitude is mapped to a unified range (such as between 0-1), and the amplitude eigenvalues ​​are extracted, including the maximum amplitude, average amplitude, amplitude variance and other parameters of the signal. This standardization process can eliminate the difference in signal strength under different measurement environments. At the same time, the time domain characteristics of the Korotkoff sound signal are analyzed, and the time domain eigenvalues ​​such as the duration, rise time, and peak occurrence time of each Korotkoff sound signal are measured. These parameters reflect the vascular elasticity and hemodynamic characteristics. The obtained frequency domain eigenvalues, amplitude eigenvalues ​​and time domain eigenvalues ​​are combined to form a characteristic parameter set. These characteristic parameters fully reflect the various characteristics of the Korotkoff sound signal, provide rich input features for the subsequent neural network model, and help improve the accuracy of blood pressure measurement. Through this multi-dimensional feature extraction method, not only can the basic features that traditional measurement methods rely on be captured, but also more potential features reflecting blood pressure changes can be obtained, providing more comprehensive data support for intelligent blood pressure measurement.

[0045] Based on the above embodiment, as an optional implementation, in S104, feature value extraction is performed on multiple Korotkoff sound signals to obtain multiple feature parameters, specifically including S401-S404: S401, performing Fourier transform on each Korotkoff sound signal to obtain a frequency domain eigenvalue of each Korotkoff sound signal.

[0046] Specifically, each collected Korotkoff sound signal is preprocessed, including removing the DC component and windowing, and the Hanning window function is selected to reduce spectrum leakage. Subsequently, the processed signal is subjected to a fast Fourier transform (FFT) to convert the time domain signal into a frequency domain representation.

[0047] Since the effective frequency components of Korotkoff sound signals are mainly distributed in the range of 20-200Hz, the spectral characteristics within this frequency band are analyzed in detail. By analyzing the spectrum, the following frequency domain eigenvalues ​​are extracted: main frequency (frequency corresponding to the maximum spectral energy), bandwidth (frequency range where the spectral energy exceeds 50% of the maximum value), spectral center of gravity (reflecting the center position of spectral energy distribution), energy proportion of each frequency band (dividing 20-200Hz into several sub-bands, calculating the proportion of energy of each sub-band to the total energy), and spectrum peak ratio (ratio of secondary peak to main peak). These frequency domain eigenvalues ​​reflect the frequency composition characteristics of Korotkoff sound signals, which can reflect the changes in the vibration characteristics of the vascular wall and the blood flow state, and provide important frequency domain information input for the subsequent neural network model.

[0048] S402, performing amplitude normalization processing on each Korotkoff sound signal to obtain an amplitude characteristic value of each Korotkoff sound signal.

[0049] Specifically, each Korotkoff sound signal is normalized, and the maximum and minimum normalization method is adopted to map the signal amplitude to the interval [0, 1]. The calculation formula is: X_normalized = (X-X_min) / (X_max-X_min), where X is the original signal value, X_min and X_max are the minimum and maximum values ​​of the signal, respectively.

[0050] After normalization, the following amplitude characteristic values ​​are calculated: the maximum amplitude of the signal (normalized peak amplitude), the average amplitude (mean of the normalized signal), the amplitude standard deviation (reflecting the discrete degree of the signal amplitude), the waveform factor (ratio of the effective value to the average value), the peak factor (ratio of the maximum value to the effective value), and the pulse factor (ratio of the maximum value to the average value). These amplitude characteristic values ​​can fully reflect the intensity change characteristics of the Korotkoff sound signal, eliminate the signal amplitude difference caused by different measurement conditions, and make the Korotkoff sound signals obtained at different times and different measurement environments comparable. Through this standardization and feature extraction method, the amplitude change characteristics of the signal are retained, and the stability of the feature parameters is improved, providing a reliable amplitude feature input for the neural network model, which helps to improve the accuracy and repeatability of blood pressure measurement.

[0051] S403: Obtaining a time domain characteristic value of each Korotkoff sound signal according to the duration of each Korotkoff sound signal.

[0052] Specifically, an amplitude threshold (usually 10% of the maximum amplitude of the signal) is set to determine the starting point and ending point of each Korotkoff sound signal, thereby calculating the duration of the signal. Based on the determined time range, the following time domain feature values ​​are extracted: signal duration (the time interval from the starting point to the ending point), rise time (the time from the starting point to the peak), fall time (the time from the peak to the ending point), peak occurrence time (the time position relative to the starting point of the signal), zero crossing rate (the number of times the signal crosses the zero point per unit time) and the first-order difference of the signal (reflecting the rate of change of the signal).

[0053] These time-domain eigenvalues ​​reflect the temporal characteristics of vascular wall vibration and hemodynamic characteristics, especially the signal duration and rise time are closely related to vascular elasticity and blood pressure level. In actual processing, the sliding window method is used to analyze the signal in segments, and the window length is set to one cardiac cycle to ensure the continuity and integrity of feature extraction. By analyzing the changing trend of these time-domain eigenvalues, the dynamic change characteristics of the vascular state during the blood pressure change process can be reflected, which provides important time-domain information for the neural network model and helps to improve the accuracy and reliability of blood pressure measurement.

[0054] S404: taking the frequency domain eigenvalues, the amplitude eigenvalues ​​and the time domain eigenvalues ​​as a plurality of characteristic parameters of each Korotkoff sound signal.

[0055] On the basis of the above embodiment, as an optional implementation, before inputting the multiple characteristic parameters, the initial systolic pressure and the initial diastolic pressure into the preset neural network model, the following steps are specifically included: S501-S503: S501, obtaining multiple groups of sample data, each group of sample data including characteristic parameters of multiple Korotkoff sound signals, initial systolic pressure, initial diastolic pressure, and standard systolic pressure and standard diastolic pressure measured by a standard blood pressure measuring device.

[0056] S502, taking the characteristic parameters, initial systolic pressure and initial diastolic pressure in each group of sample data as input features, and taking the standard systolic pressure and standard diastolic pressure as output features.

[0057] Specifically, multiple groups of sample data are collected through clinical experiments, and the experimental subjects include subjects of different age groups, genders and health conditions. Each subject is measured using the measurement method of the present invention and a medical mercury sphygmomanometer at the same time. For each group of sample data, characteristic parameters (including frequency domain characteristic values, amplitude characteristic values ​​and time domain characteristic values) extracted from the Korotkoff sound signal, as well as the initial systolic pressure and initial diastolic pressure determined based on the first and last Korotkoff sounds are recorded, and the standard systolic pressure and standard diastolic pressure measured by the medical mercury sphygmomanometer are recorded as the true value.

[0058] The sample data collection process strictly follows the medical measurement standards to ensure the accuracy and reliability of the data. In the data preprocessing stage, all feature parameters are normalized so that they are distributed in the same numerical range to avoid the impact of large differences in numerical ranges on the model training effect. The processed feature parameter set and the initial systolic pressure and initial diastolic pressure form the input feature vector, and the corresponding standard systolic pressure and standard diastolic pressure are used as the output feature vector to form a complete training data pair.

[0059] S503, training the neural network structure according to the input features and the output features to obtain a preset neural network model.

[0060] In one example, to build an intelligent model that can accurately predict blood pressure values, it is necessary to train the neural network based on the training data prepared in the previous steps.

[0061] The neural network adopts a multi-layer perceptron structure, which includes an input layer, two hidden layers and an output layer. The number of neurons in the input layer corresponds to the dimension of the characteristic parameters (including frequency domain eigenvalues, amplitude eigenvalues, time domain eigenvalues, initial systolic pressure and initial diastolic pressure). The two hidden layers are configured with 64 and 32 neurons respectively. The output layer contains 2 neurons corresponding to the target systolic pressure and target diastolic pressure respectively.

[0062] The network uses ReLU as the activation function of the hidden layer, the output layer uses a linear activation function, and the loss function uses the mean square error (MSE) to evaluate the difference between the predicted value and the true value. The training process uses a small batch stochastic gradient descent algorithm, the batch size is set to 64, the initial value of the learning rate is set to 0.001, and the Adam optimizer is used to adaptively adjust the learning rate. To prevent overfitting, a Dropout layer is added between the two hidden layers, the dropout rate is set to 0.3, and the L2 regularization technique is used. The overall data is divided into training set, validation set, and test set in a ratio of 7:2:1. The model performance is monitored by the validation set during the training process. When the validation set loss has not improved for 5 consecutive epochs, the early stopping mechanism is activated. After repeated training and tuning, the mean absolute error of the final preset neural network model on the test set is controlled within ±3mmHg, meeting the requirements of medical device standards.

[0063] In order to build a high-precision blood pressure prediction model, the neural network structure is systematically trained. The specific implementation process is as follows: First, the basic architecture of the neural network is constructed. The input layer contains 22 neurons, corresponding to 22-dimensional input features: frequency domain eigenvalues ​​(main frequency, bandwidth, spectrum center of gravity, frequency band energy ratio, spectrum peak ratio), amplitude eigenvalues ​​(signal maximum amplitude, average amplitude, amplitude standard deviation, waveform factor, peak factor, pulse factor), time domain eigenvalues ​​(signal duration, rise time, fall time, peak occurrence time, zero crossing rate, first-order difference), and initial systolic pressure and initial diastolic pressure.

[0064] The first hidden layer is configured with 64 neurons, hidden layer: ReLU activation function, f(x) = max(0, x), x is the neuron input value, and Dropout technology is applied (discard rate 0.3). The second hidden layer is configured with 32 neurons, also using ReLU activation function and Dropout (discard rate 0.3). Output layer: linear activation function f(x) = x, the output layer contains 2 neurons, corresponding to the target systolic pressure and target diastolic pressure, respectively, using linear activation function.

[0065] In the data preprocessing stage, the input features are standardized, and StandardScaler is used to convert each feature into a standard normal distribution with a mean of 0 and a standard deviation of 1: X_standardized = (X-μ) / σ, where X is the original feature value, μ is the feature mean, μ = (1 / n)Σxi, n is the number of samples, and σ is the feature standard deviation: The processed dataset is randomly divided into training set, validation set and test set in the ratio of 7:2:1.

[0066] The training process uses a small batch stochastic gradient descent algorithm, and the specific configuration is as follows: The batch size (batch_size) is set to 64, which means that 64 samples are used each time the parameters are updated.

[0067] The initial learning rate is set to 0.001, and the Adam optimizer is used (parameter configuration: beta1=0.9, beta2=0.999, epsilon=1e-8).

[0068] The loss function uses the mean square error: MSE = (1 / n)Σ(y_pred-y_true)2, where n is the number of samples, y_pred is the blood pressure value predicted by the model, and y_true is the standard blood pressure value.

[0069] L2 regularization term: L2_loss = λ*Σ(w2), where: λ is the regularization coefficient, set to 0.01, w is the weight parameter of each layer of the network, and the total loss function: Total_loss = MSE + L2_loss.

[0070] The training iteration process includes: Forward propagation: The input features are calculated through each layer of the network to obtain the predicted blood pressure value.

[0071] Back propagation: Calculate the gradient of the loss function with respect to the parameters of each layer.

[0072] Parameter update: Use the Adam optimizer to update the network parameters according to the calculated gradients.

[0073] Validation evaluation: After each epoch, the loss value is calculated on the validation set.

[0074] Early stopping mechanism: When the validation set loss does not improve for 5 consecutive epochs, save the current optimal model parameters and stop training.

[0075] Final model evaluation criteria: Mean absolute error (MAE): systolic blood pressure error ≤ ±3mmHg, diastolic blood pressure error ≤ ±2mmHg.

[0076] Root mean square error (RMSE): systolic blood pressure RMSE ≤ 4 mmHg, diastolic blood pressure RMSE ≤ 3 mmHg.

[0077] Correlation coefficient (R 2 ): R of systolic and diastolic blood pressure 2 Both should be greater than 0.95.

[0078] The trained preset neural network model should meet the following performance indicators: Accuracy on the test set: More than 95% of the measurement results are within the range of ±5mmHg.

[0079] Model stability: The standard deviation of consecutive measurements is less than 2 mmHg.

[0080] Generalization ability: The measurement accuracy can be kept stable for testers of different ages, genders and physical conditions.

[0081] Assume that the measurement data of 1000 subjects are collected, and each subject is measured 3 times, and a total of 3000 sets of sample data are obtained. The input features of each set of samples include: frequency domain eigenvalues: main frequency (35Hz), bandwidth (45Hz), spectrum center of gravity (42Hz), frequency band energy ratio (0.65), spectrum peak ratio (0.78); amplitude eigenvalues: signal maximum amplitude (0.95), average amplitude (0.45), amplitude standard deviation (0.15), waveform factor (1.25), peak factor (2.1), pulse factor (2.3); time domain eigenvalues: signal duration (0.8s), rise time (0.3s), fall time (0.5s), peak occurrence time (0.25s), zero crossing rate (120Hz), first order difference (0.05).

[0082] Initial blood pressure values: initial systolic pressure (135 mmHg), initial diastolic pressure (85 mmHg), the corresponding output characteristics are measured by a standard sphygmomanometer: standard systolic pressure (130 mmHg) and standard diastolic pressure (80 mmHg) After data preprocessing, the 3000 groups of samples were divided into 7:2:1 ratio: training set: 2100 groups, validation set: 600 groups, and test set: 300 groups.

[0083] Model training process record: First round of training (Epoch1): training set loss: MSE = 25.6; Validation set loss: MSE = 23.8; systolic MAE = ±5.2 mmHg; diastolic MAE = ±4.1 mmHg.

[0084] Mid-training (Epoch50): training set loss: MSE = 12.3, validation set loss: MSE = 11.9, systolic blood pressure MAE = ± 3.8 mmHg, diastolic blood pressure MAE = ± 2.9 mmHg.

[0085] Training completed (Epoch 100): training set loss: MSE = 8.2, validation set loss: MSE = 8.5, systolic blood pressure MAE = ± 2.8 mmHg, diastolic blood pressure MAE = ± 1.9 mmHg.

[0086] Final performance evaluation on the test set: Systolic blood pressure measurement results: mean absolute error: ±2.8mmHg, root mean square error: 3.2mmHg, correlation coefficient R 2 :0.96.

[0087] Diastolic blood pressure measurement results: mean absolute error: ±1.9 mmHg, root mean square error: 2.3 mmHg, correlation coefficient R 2 :0.97.

[0088] Test case verification: Input new measurement data: Frequency domain eigenvalues: main frequency (38Hz), bandwidth (42Hz), spectrum center of gravity (40Hz), frequency band energy ratio (0.68), spectrum peak ratio (0.75); Amplitude eigenvalues: maximum signal amplitude (0.92), average amplitude (0.48), amplitude standard deviation (0.14), waveform factor (1.28), peak factor (2.0), pulse factor (2.2); Time domain eigenvalues: signal duration (0.75s), rise time (0.28s), fall time (0.47s), peak occurrence time (0.23s), zero crossing rate (125Hz), first-order difference (0.04).

[0089] Initial blood pressure values: initial systolic pressure (140 mmHg), initial diastolic pressure (88 mmHg).

[0090] Model prediction results: predicted systolic blood pressure: 133 mmHg (actual value: 132 mmHg, error: +1 mmHg); predicted diastolic blood pressure: 82 mmHg (actual value: 81 mmHg, error: +1 mmHg).

[0091] After obtaining the preset neural network model, it also includes: obtaining multiple groups of verification sample data, each group of verification sample data includes verification characteristic parameters of multiple Korotkoff sound signals, verification initial systolic pressure, verification initial diastolic pressure, and verification standard systolic pressure and verification standard diastolic pressure measured by standard blood pressure measurement equipment; inputting the verification characteristic parameters, verification initial systolic pressure and verification initial diastolic pressure in each group of verification sample data into the preset neural network model to obtain predicted systolic pressure and predicted diastolic pressure; calculating a first difference between the predicted systolic pressure and the verification standard systolic pressure, and a second difference between the predicted diastolic pressure and the verification standard diastolic pressure; when the first difference and the second difference both meet the preset requirements, determining that the preset neural network model meets the use requirements, if the first difference and / or the second difference does not meet the preset requirements, retraining the neural network model until the preset neural network model that meets the use requirements is obtained.

[0092] In one example, an independent validation sample data set is collected, which comes from subjects who did not participate in the model training to ensure the objectivity of the validation. For each group of validation samples, the validation characteristic parameters (including frequency domain characteristic values, amplitude characteristic values ​​and time domain characteristic values) of the Korotkoff sound signal are extracted using the same method as the training data, and the validation initial systolic pressure and the validation initial diastolic pressure are recorded. At the same time, the validation standard systolic pressure and the validation standard diastolic pressure are measured using a medical mercury column sphygmomanometer as reference values. The validation characteristic parameters, the validation initial systolic pressure and the validation initial diastolic pressure are input into the trained preset neural network model to obtain the predicted systolic pressure and the predicted diastolic pressure. According to the requirements of the medical device standard, the difference between the predicted value and the standard value is calculated: the first difference is the absolute error between the predicted systolic pressure and the validation standard systolic pressure, and the second difference is the absolute error between the predicted diastolic pressure and the validation standard diastolic pressure. The preset requirement is set as follows: in more than 95% of the measurement results, the first difference does not exceed ±8mmHg, the second difference does not exceed ±5mmHg, and the average errors do not exceed ±5mmHg and ±3mmHg respectively. When the measurement results of all verification samples meet these preset requirements, it can be confirmed that the preset neural network model meets the usage requirements; if there is a situation that does not meet the preset requirements, it is necessary to analyze the cause of the error. Possible optimization measures include: increasing the number of training samples, adjusting network structure parameters, optimizing feature extraction methods or modifying training strategies, etc., and then retraining the model until it meets the usage requirements.

[0093] S105, inputting multiple characteristic parameters, initial systolic pressure and initial diastolic pressure into a preset neural network model to obtain a target systolic pressure and a target diastolic pressure of the person to be measured, and using the target systolic pressure and the target diastolic pressure as the blood pressure measurement results of the person to be measured.

[0094] Specifically, in order to improve the accuracy of blood pressure measurement, a preset neural network model is used to deeply analyze and process the characteristic parameters and initial blood pressure values ​​obtained in the above steps.

[0095] The preset neural network model adopts a multi-layer perceptron structure, including an input layer, two hidden layers and an output layer. The number of neurons in the input layer corresponds to the dimension of the characteristic parameters, the two hidden layers contain 64 and 32 neurons respectively, and the output layer contains 2 neurons corresponding to the target systolic blood pressure and target diastolic blood pressure respectively. The model uses ReLU as the activation function, is trained by the back propagation algorithm, and uses the mean square error as the loss function.

[0096] In actual application, the extracted characteristic parameters (including frequency domain eigenvalues, amplitude eigenvalues ​​and time domain eigenvalues) are normalized together with the initial systolic pressure and initial diastolic pressure and then input into the neural network model. The model will calculate a more accurate blood pressure value based on the complex nonlinear relationship of these input parameters. The neural network model uses a large amount of clinical verification data during the training phase, including patient data of different ages, genders and physical conditions, to ensure that the model has good generalization ability. The target systolic pressure and target diastolic pressure output by the model are used as the final blood pressure measurement results. This blood pressure measurement method based on a neural network has higher accuracy than the traditional Korotkoff sound method, and can better handle the influence of individual differences and the measurement environment, while maintaining the stability and repeatability of the measurement process.

[0097] In this way, not only the accuracy of blood pressure measurement is improved, but also the reliability of the measurement results is enhanced, providing a more valuable reference for clinical diagnosis.

[0098] Based on the above embodiment, as an optional implementation, in S105, multiple characteristic parameters, initial systolic pressure and initial diastolic pressure are input into a preset neural network model to obtain the target systolic pressure and target diastolic pressure of the person to be tested, which specifically includes S601-S603: S601, inputting a plurality of characteristic parameters, initial systolic pressure and initial diastolic pressure into a preset neural network model through an input layer of the preset neural network model.

[0099] S602, using the hidden layer of the preset neural network model, performing feature extraction and nonlinear transformation on the input multiple feature parameters, the initial systolic pressure and the initial diastolic pressure.

[0100] S603, obtaining the target systolic blood pressure and target diastolic blood pressure of the person to be tested through the output layer of the preset neural network model.

[0101] Specifically, the extracted 22-dimensional feature parameters (including frequency domain feature values, amplitude feature values, and time domain feature values) as well as the initial systolic pressure and initial diastolic pressure are input into the input layer of the neural network, and each input neuron corresponds to a feature parameter. After the data enters the input layer, it is first standardized so that all feature data are distributed in the same numerical range.

[0102] Next, the data flows through two hidden layers for deep feature extraction and nonlinear transformation. The 64 neurons in the first hidden layer process the input data through the ReLU activation function to extract high-dimensional feature representation; the 32 neurons in the second hidden layer further reduce the dimension and combine the features, and enhance the nonlinear expression ability of the model through the ReLU activation function. During the hidden layer processing, the Dropout mechanism randomly blocks some neurons (discard rate 0.3) to prevent the model from overfitting, and the L2 regularization technology controls the scale of the weight parameters. Finally, the feature data processed by the hidden layer is passed to the output layer, and the two neurons in the output layer calculate the target systolic pressure and target diastolic pressure respectively through the linear activation function.

[0103] Taking an actual case as an example, when the input features include 22 parameters such as the main frequency of 35Hz and the bandwidth of 45Hz, after standardization of the input layer and feature extraction and nonlinear transformation of the hidden layer, the prediction results of the target systolic pressure of 132mmHg and the target diastolic pressure of 81mmHg are finally obtained in the output layer. This multi-layer processing method based on deep neural networks can make full use of the multi-dimensional feature information in the blood pressure measurement process, realize the precise mapping from feature space to blood pressure value, and ensure the accuracy and reliability of blood pressure prediction.

[0104] After the target systolic blood pressure and the target diastolic blood pressure are used as the blood pressure measurement results of the person to be tested, it also includes: The target systolic blood pressure and target diastolic blood pressure are compared with the preset normal range of systolic blood pressure and the normal range of diastolic blood pressure respectively; when the target systolic blood pressure exceeds the preset normal range of systolic blood pressure and / or the target diastolic blood pressure exceeds the preset normal range of diastolic blood pressure, an abnormal blood pressure prompt message is generated; the target systolic blood pressure, target diastolic blood pressure and abnormal blood pressure prompt message are displayed on the display screen.

[0105] In one example, first, according to the blood pressure classification standard of the World Health Organization (WHO), the normal range of systolic blood pressure is set to 90-140mmHg, and the normal range of diastolic blood pressure is set to 60-90mmHg. After the target systolic blood pressure and target diastolic blood pressure are calculated by the neural network model, the system automatically compares these measured values ​​with the preset normal range. To illustrate with an actual case, suppose that the target systolic blood pressure is 155mmHg and the target diastolic blood pressure is 85mmHg in a certain measurement. Since the target systolic blood pressure of 155mmHg exceeds the preset upper limit of the normal range of 140mmHg, and the target diastolic blood pressure of 85mmHg is within the normal range, the system will generate a blood pressure abnormality prompt message of "Warning: Systolic blood pressure is high". At the same time, the measurement results are clearly displayed on the display: systolic blood pressure 155mmHg (marked in red), diastolic blood pressure 85mmHg, and blood pressure abnormality prompt information is displayed in conjunction. This real-time monitoring and prompt mechanism can help users understand their blood pressure status in a timely manner.

[0106] Based on the above method, the present application also discloses an intelligent blood pressure measurement system based on the electronic Korotkoff sound method, such as Figure 2 As shown, Figure 2 The present invention is a schematic diagram of the structure of an intelligent blood pressure measurement system based on the electronic Korotkoff sound method provided in an embodiment of the present invention, wherein the device comprises: a control module, an acquisition module, a determination module, an extraction module and an output module; wherein the control module is used to increase the air pressure in the cuff wrapped around the upper arm of the person to be measured to a preset pressure and then deflate the cuff; the acquisition module is used to acquire multiple Korotkoff sound signals from the brachial artery and the air pressure in the cuff corresponding to each Korotkoff sound signal during the process of deflation of the cuff; the determination module is used to obtain the first occurrence of a plurality of Korotkoff sound signals according to the pressure in the cuff; the output ... The first air pressure in the cuff when a Korotkoff sound signal appears is used to determine the initial systolic pressure of the person to be tested, and the initial diastolic pressure of the person to be tested is determined according to the second air pressure in the cuff when the last Korotkoff sound signal appears among multiple Korotkoff sound signals; an extraction module is used to extract feature values ​​of multiple Korotkoff sound signals to obtain multiple feature parameters; an output module is used to input the multiple feature parameters, the initial systolic pressure and the initial diastolic pressure into a preset neural network model to obtain the target systolic pressure and target diastolic pressure of the person to be tested, and use the target systolic pressure and target diastolic pressure as the blood pressure measurement results of the person to be tested.

[0107] It should be noted that: when the device provided in the above embodiment realizes its function, only the division of the above functional modules is used as an example. In actual application, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0108] See also Figure 3 , is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 3 As shown, the electronic device 1000 may include: at least one processor 1001 , at least one network interface 1004 , a user interface 1003 , a memory 1005 , and at least one communication bus 1002 .

[0109] The communication bus 1002 is used to realize the connection and communication between these components.

[0110] The user interface 1003 may include a display screen (Display) and a camera (Camera), and the optional user interface 1003 may also include a standard wired interface and a wireless interface.

[0111] The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0112] Among them, the processor 1001 may include one or more processing cores. The processor 1001 uses various interfaces and lines to connect various parts in the entire server, and executes various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 1005, and calling data stored in the memory 1005. Optionally, the processor 1001 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 1001 can integrate one or more combinations of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 1001, and it can be implemented separately through a chip.

[0113] Among them, the memory 1005 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read-Only Memory). Optionally, the memory 1005 includes a non-transitory computer-readable storage medium. The memory 1005 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 1005 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 1005 may optionally be at least one storage device located away from the aforementioned processor 1001. As Figure 3 As shown, the memory 1005 as a computer storage medium may include an operating system, a network communication module, a user interface module, and an application program of an intelligent blood pressure measurement method based on the electronic Korotkoff sound method.

[0114] exist Figure 3In the electronic device 1000 shown, the user interface 1003 is mainly used to provide an input interface for the user and obtain data input by the user; and the processor 1001 can be used to call an application program stored in the memory 1005 for an intelligent blood pressure measurement method based on the electronic Korotkoff sound method. When executed by one or more processors, the electronic device executes one or more methods described in the above embodiments.

[0115] An electronic device readable storage medium stores instructions, which, when executed by one or more processors, enable the electronic device to execute one or more of the methods described in the above embodiments.

[0116] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the order of the actions described, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required for the present application.

[0117] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0118] In the several embodiments provided in the present application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of the units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some service interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0119] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0120] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0121] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a memory and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned memory includes: various media that can store program codes, such as USB flash drives, mobile hard drives, magnetic disks or optical disks.

[0122] The above is only an exemplary embodiment of the present disclosure, and the scope of the present disclosure cannot be limited thereto. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the specification and practicing the disclosure here, those skilled in the art will easily think of other embodiments of the present disclosure. This application is intended to cover any modification, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the technical field not recorded in the present disclosure. The description and examples are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. An intelligent blood pressure measurement method based on electronic Korotkoff sound method, characterized in that: The method comprises: After increasing the air pressure in the cuff wrapped around the upper arm of the person to be tested to a preset pressure, the cuff is deflated; In the process of deflation of the cuff, a plurality of Korotkoff sound signals from the brachial artery and the air pressure in the cuff corresponding to each of the Korotkoff sound signals are obtained; Determine the initial systolic pressure of the person to be measured according to a first air pressure in the cuff when a Korotkoff sound signal appears for the first time among the multiple Korotkoff sound signals, and determine the initial diastolic pressure of the person to be measured according to a second air pressure in the cuff when a Korotkoff sound signal appears for the last time among the multiple Korotkoff sound signals; Extracting characteristic values ​​of the multiple Korotkoff sound signals to obtain multiple characteristic parameters; The plurality of characteristic parameters, the initial systolic pressure and the initial diastolic pressure are input into a preset neural network model to obtain the target systolic pressure and the target diastolic pressure of the person to be measured, and the target systolic pressure and the target diastolic pressure are used as the blood pressure measurement results of the person to be measured.

2. The intelligent blood pressure measurement method based on the electronic Korotkoff sound method according to claim 1, characterized in that: The step of obtaining a plurality of Korotkoff sound signals from the brachial artery and the air pressure in the cuff corresponding to each of the Korotkoff sound signals comprises: During the process of deflation of the cuff, a sound sensor disposed in the cuff collects the multiple Korotkoff sound signals generated by the pulse beating of the brachial artery, and a pressure sensor disposed in the cuff receives the air pressure in the cuff corresponding to each Korotkoff sound signal.

3. The intelligent blood pressure measurement method based on the electronic Korotkoff sound method according to claim 1, characterized in that: The extracting characteristic values ​​of the multiple Korotkoff sound signals to obtain multiple characteristic parameters includes: Performing Fourier transform on each of the Korotkoff sound signals to obtain frequency domain eigenvalues ​​of each of the Korotkoff sound signals; Performing amplitude normalization processing on each of the Korotkoff sound signals to obtain an amplitude characteristic value of each of the Korotkoff sound signals; Obtaining a time domain characteristic value of each Korotkoff sound signal according to the duration of each Korotkoff sound signal; The frequency domain eigenvalue, the amplitude eigenvalue and the time domain eigenvalue are used as a plurality of characteristic parameters of each of the Korotkoff sound signals.

4. The intelligent blood pressure measurement method based on the electronic Korotkoff sound method according to claim 1, characterized in that: Before inputting the plurality of characteristic parameters, the initial systolic pressure and the initial diastolic pressure into the preset neural network model, the method further includes: Acquire multiple groups of sample data, each group of sample data includes characteristic parameters of multiple Korotkoff sound signals, initial systolic pressure, initial diastolic pressure, and standard systolic pressure and standard diastolic pressure measured by a standard blood pressure measuring device; Taking the characteristic parameters, the initial systolic pressure and the initial diastolic pressure in each group of the sample data as input features, and taking the standard systolic pressure and the standard diastolic pressure as output features; The neural network structure is trained according to the input features and the output features to obtain a preset neural network model.

5. The intelligent blood pressure measurement method based on the electronic Korotkoff sound method according to claim 4, characterized in that: After obtaining the preset neural network model, the method further includes: Acquire multiple groups of verification sample data, each group of the verification sample data includes verification characteristic parameters of multiple Korotkoff sound signals, verification initial systolic pressure, verification initial diastolic pressure, and verification standard systolic pressure and verification standard diastolic pressure measured by a standard blood pressure measurement device; Inputting the verification characteristic parameters, the verification initial systolic pressure and the verification initial diastolic pressure in each group of the verification sample data into the preset neural network model to obtain the predicted systolic pressure and the predicted diastolic pressure; Calculating a first difference between the predicted systolic pressure and the verification standard systolic pressure, and a second difference between the predicted diastolic pressure and the verification standard diastolic pressure; When both the first difference and the second difference meet the preset requirements, it is determined that the preset neural network model meets the usage requirements. If the first difference and / or the second difference does not meet the preset requirements, the neural network model is retrained until the preset neural network model that meets the usage requirements is obtained.

6. The intelligent blood pressure measurement method based on the electronic Korotkoff sound method according to claim 1, characterized in that: The step of inputting the plurality of characteristic parameters, the initial systolic pressure, and the initial diastolic pressure into a preset neural network model to obtain the target systolic pressure and the target diastolic pressure of the person to be tested comprises: Inputting the plurality of characteristic parameters, the initial systolic pressure and the initial diastolic pressure into the preset neural network model through the input layer of the preset neural network model; Using the hidden layer of the preset neural network model, feature extraction and nonlinear transformation are performed on the input multiple characteristic parameters, the initial systolic pressure and the initial diastolic pressure; The target systolic blood pressure and target diastolic blood pressure of the person to be tested are obtained through the output layer of the preset neural network model.

7. The intelligent blood pressure measurement method based on the electronic Korotkoff sound method according to claim 1, characterized in that: After taking the target systolic pressure and the target diastolic pressure as the blood pressure measurement result of the person to be measured, the method further includes: Comparing the target systolic pressure and the target diastolic pressure with a preset normal range of systolic pressure and a preset normal range of diastolic pressure, respectively; When the target systolic pressure exceeds the preset normal range of systolic pressure and / or the target diastolic pressure exceeds the preset normal range of diastolic pressure, generating abnormal blood pressure prompt information; The target systolic pressure, the target diastolic pressure and the abnormal blood pressure prompt information are displayed on a display screen.

8. An intelligent blood pressure measurement system based on the electronic Korotkoff sound method, characterized in that: The system includes: a control module, an acquisition module, a determination module, an extraction module and an output module; wherein, The control module is used to increase the air pressure in the cuff wrapped around the upper arm of the person to be tested to a preset pressure and then deflate the cuff; The acquisition module is used to acquire multiple Korotkoff sound signals from the brachial artery and the air pressure in the cuff corresponding to each Korotkoff sound signal during the process of deflation of the cuff; The determination module is used to determine the initial systolic pressure of the person to be tested according to the first air pressure in the cuff when the Korotkoff sound signal appears for the first time among the multiple Korotkoff sound signals, and to determine the initial diastolic pressure of the person to be tested according to the second air pressure in the cuff when the Korotkoff sound signal appears for the last time among the multiple Korotkoff sound signals; The extraction module is used to extract characteristic values ​​of the multiple Korotkoff sound signals to obtain multiple characteristic parameters; The output module is used to input the multiple characteristic parameters, the initial systolic pressure and the initial diastolic pressure into a preset neural network model to obtain the target systolic pressure and target diastolic pressure of the person to be tested, and use the target systolic pressure and the target diastolic pressure as the blood pressure measurement results of the person to be tested.

9. An electronic device, characterized in that: It includes a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: A computer program is stored which can be loaded by a processor and execute the method according to any one of claims 1 to 7.

Images