A method, system, device and medium for improving the accuracy of heart rate acquisition

Abstract

The present invention relates to the technical field of smart wearable devices, and discloses a method, system, device and medium for improving the accuracy of heart rate acquisition. The method includes obtaining an electrocardiogram signal and a photoplethysmogram signal; calculating a pulse rate based on the photoplethysmogram signal, and obtaining heart rate variability parameters through heart rate variability analysis; calculating a heart rate based on the electrocardiogram signal; obtaining a heart rate category classification through a linear discriminant analysis algorithm according to the heart rate variability parameters; obtaining a heart rate prediction value through a heart rate model according to the heart rate; and obtaining a final heart rate output value through an error backpropagation neural network according to the heart rate prediction value, the heart rate category classification, the photoplethysmogram signal and the heart rate. The method has the following effects: By considering the heart rate caused by individual differences and the influence of exercise or different intensity actions on the heart rate, the accuracy of heart rate acquisition is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of smart wearable devices, and particularly to a method, system, device and medium for improving the accuracy of heart rate acquisition. Background Art

[0002] Heart rate can be an important indicator reflecting the health status of the human body. In the past, due to technical limitations, heart rate data could only be obtained through professional electrocardiogram acquisition devices. With technological progress, wearable devices such as smart bracelets and smart watches have become mainstream electronic products because of their function of monitoring heart rate and good user experience. Nowadays, most of the smart watch and bracelet products sold on the market have ECG electrocardiogram sensors and PPG photoplethysmogram sensors, which measure heart rate based on different theoretical bases, namely ECG heart rate measured based on electrocardiogram sensors and PPG heart rate measured based on photoplethysmography. However, the accuracy of both is greatly affected by multiple factors such as individual heart rate, exercise intensity, and heart rate variability, resulting in significant differences in heart rate data collected by the same instrument for different individuals or in different scenarios. Therefore, how to comprehensively utilize data and improve the reliability of heart rate data, and improving the algorithm model is one of the keys.

[0003] Currently, relevant research results have been accumulated in terms of the accuracy of heart rate acquisition. For example, algorithms based on the joint processing of ECG and PPG can effectively improve the detection results of heart rate under the cooperation of ECG and PPG, such as improving the anti-noise ability and heart rate detection range. However, the prior art does not consider the influence of heart rate differences between individuals and situational factors on the accuracy of heart rate acquisition, and does not consider the influence of the physiological parameter of heart rate variability (Autonomic Nervous System, ANS) in heart rate estimation. Moreover, its effect on the heart rate estimation deviation caused by individual differences is poor, and the influence of heart rate caused by exercise or different intensity actions is not fully considered, resulting in low accuracy of heart rate acquisition. Summary of the Invention

[0004] The present invention provides a method, system, device and medium for improving the accuracy of heart rate acquisition to achieve the improvement of the accuracy of heart rate acquisition.

[0005] In a first aspect, to solve the above technical problems, the present invention provides a method for improving the accuracy of heart rate acquisition, including:

[0006] Obtaining an electrocardiogram signal and a photoplethysmogram signal;

[0007] Calculating the pulse rate according to the photoplethysmogram signal, and obtaining heart rate variability parameters through heart rate variability analysis;

[0008] Calculating the heart rate according to the electrocardiogram signal;

[0009] Based on the heart rate variability parameters, a heart rate category classification is obtained through a linear discriminant analysis algorithm;

[0010] Based on the heart rate, a heart rate prediction value is obtained through a heart rate model;

[0011] Based on the heart rate prediction value, the heart rate category classification, the photoplethysmogram signal, and the heart rate, a final heart rate output value is obtained through a backpropagation neural network.

[0012] Preferably, calculating the pulse rate according to the photoplethysmogram signal and obtaining the heart rate variability parameters through heart rate variability analysis includes:

[0013] Performing frequency domain filtering processing on the photoplethysmogram signal to obtain a filtered photoplethysmogram signal;

[0014] The pulse rate is calculated through the following formula:

[0015]

[0016] where R bpm is the pulse rate, and T rp is the time interval between two adjacent peaks in the filtered photoplethysmogram signal;

[0017] Based on the filtered photoplethysmogram signal, the heartbeat peak interval of each heartbeat cycle is extracted;

[0018] The heart rate variability parameters include respiratory sinus arrhythmia, the root mean square of the difference in adjacent heartbeat peak intervals, and the standard deviation of normal heartbeat peak intervals;

[0019] Respiratory sinus arrhythmia is calculated through the following formula:

[0020]

[0021] where RSA is respiratory sinus arrhythmia, F H is the high-frequency proportion of the filtered photoplethysmogram signal, and F L is the low-frequency proportion of the filtered photoplethysmogram signal;

[0022] The root mean square of the difference in adjacent heartbeat peak intervals is calculated through the root mean square formula;

[0023] The standard deviation of normal heartbeat peak intervals is calculated through the standard deviation formula.

[0024] Preferably, obtaining the heart rate category classification through the linear discriminant analysis algorithm based on the heart rate variability parameters includes:

[0025] The heart rate category classification includes: resting heart rate, active heart rate, and exercise heart rate;

[0026] When the value of respiratory sinus arrhythmia and the root mean square of the difference in adjacent heartbeat peak intervals in the heart rate variability parameters are higher than a preset threshold, and the standard deviation of normal heartbeat peak intervals is less than the preset threshold, it is determined as the exercise heart rate;

[0027] When the value of respiratory sinus arrhythmia and the root mean square of the difference in adjacent heartbeat peak intervals in the heart rate variability parameters are lower than the preset threshold, and the standard deviation of normal heartbeat peak intervals is greater than the preset threshold, it is determined as the resting heart rate.

[0028] Preferably, obtaining the heart rate prediction value according to the heart rate through a heart rate model includes:

[0029] The heart rate model formula is:

[0030] PPG = c 1 ·HR + c 2 ·HR 2 - c 3 ·HR 3 + c 4 ·HR 4 - c 5 ·HR 5

[0031] where HR is the heart rate, PPG is the predicted heart rate, and c 1 、c 2 、c 3 、c 4 and c 5 are set coefficients;

[0032] Among them, the heart rate model is trained through historical big data, and the heart rate model can obtain the heart rate prediction value according to the input heart rate.

[0033] Preferably, obtaining the final heart rate output value according to the heart rate prediction value, the heart rate category classification, the photoplethysmogram signal, and the heart rate through an error backpropagation neural network includes:

[0034] Taking the heart rate prediction value, the heart rate category classification, the photoplethysmogram signal, and the heart rate as the neurons of the input layer of the error backpropagation neural network;

[0035] Obtaining the neural network predicted heart rate through the error backpropagation neural network;

[0036] When the heart rate category classification is the active heart rate, taking the neural network predicted heart rate as the final heart rate output value.

[0037] Preferably, calculating the heart rate based on the electrocardiogram signal includes:

[0038] Performing frequency-domain filtering on the electrocardiogram signal to obtain a filtered electrocardiogram signal;

[0039] Calculating the heart rate through the following formula:

[0040]

[0041] where H r is the heart rate, and T r is the time interval between two adjacent peaks in the filtered electrocardiogram signal.

[0042] Preferably, the method further includes:

[0043] When the heart rate category is classified as the quiet heart rate, using the heart rate prediction value as the final heart rate output value.

[0044] In a second aspect, the present invention provides a system for improving the accuracy of heart rate acquisition, including:

[0045] A data acquisition module for acquiring an electrocardiogram signal and a photoplethysmogram signal;

[0046] A heart rate parameter module for calculating the pulse rate based on the photoplethysmogram signal and obtaining heart rate variability parameters through heart rate variability analysis;

[0047] A heart rate calculation module for calculating the heart rate based on the electrocardiogram signal;

[0048] A heart rate classification module for obtaining a heart rate category classification through a linear discriminant analysis algorithm based on the heart rate variability parameters;

[0049] A heart rate prediction module for obtaining a heart rate prediction value through a heart rate model based on the heart rate;

[0050] A heart rate output module for obtaining a final heart rate output value through an error backpropagation neural network based on the heart rate prediction value, the heart rate category classification, the photoplethysmogram signal, and the heart rate.

[0051] In a third aspect, the present invention further provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the method for improving the accuracy of heart rate acquisition described in any one of the above is implemented.

[0052] Fourthly, the present invention further provides a computer-readable storage medium, which includes a stored computer program. When the computer program runs, it controls the device where the computer-readable storage medium is located to execute the method for improving the accuracy of heart rate acquisition described in any one of the above.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] The present invention discloses a method for improving the accuracy of heart rate acquisition. The method includes obtaining an electrocardiogram (ECG) signal and a photoplethysmogram (PPG) signal; calculating a pulse rate based on the PPG signal, and obtaining heart rate variability parameters through heart rate variability analysis; calculating a heart rate based on the ECG signal; obtaining a heart rate category classification through a linear discriminant analysis algorithm based on the heart rate variability parameters; obtaining a heart rate prediction value through a heart rate model based on the heart rate; and obtaining a final heart rate output value through an error backpropagation neural network based on the heart rate prediction value, the heart rate category classification, the PPG signal, and the heart rate. By combining two different signal acquisition methods of ECG and PPG, the advantages of each are fully utilized. ECG provides high-precision electrocardiographic information, while PPG is suitable for continuous monitoring. This multi-modal fusion can more comprehensively reflect the heart rate status of the user. At the same time, through a personalized model, it can adapt to the physiological characteristics of different users and reduce the heart rate estimation deviation caused by individual differences. Further, the LDA algorithm is used to classify the heart rate variability parameters to distinguish three states of quiet, active, and exercise, so as to adopt different heart rate calculation methods in different situations and improve the accuracy of data. Description of the Drawings

[0055] Figure 1 is a schematic flowchart of a method for improving the accuracy of heart rate acquisition provided by the first embodiment of the present invention;

[0056] Figure 2 is a schematic structural diagram of a system for improving the accuracy of heart rate acquisition provided by the second embodiment of the present invention. Detailed Embodiments

[0057] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0058] Heart rate can be an important indicator reflecting the health status of the human body. In the past, due to technical limitations, it could only be obtained through professional electrocardiogram acquisition equipment.

[0059] Heart rate data. With technological advancements, wearable devices such as smart bracelets and smart watches have become mainstream electronic products due to their heart rate monitoring function and good user experience. Most of the smart watch and bracelet products sold on the market today have both an ECG electrocardiogram sensor and a PPG photoplethysmogram sensor, which measure heart rate based on different theoretical bases, namely, measuring ECG heart rate based on the electrocardiogram sensor and measuring PPG heart rate based on photoplethysmography. However, the accuracy of both is greatly affected by multiple factors such as individual heart rate, exercise intensity, and heart rate variability, resulting in significant differences in heart rate data collected by the same instrument for different individuals or in different scenarios. Therefore, how to comprehensively utilize data and improve the reliability of heart rate data, and enhancing the algorithm model is one of the keys.

[0060] Currently, relevant research results have been accumulated in terms of the accuracy of heart rate acquisition. For example, algorithms based on the joint processing of ECG and PPG can effectively improve the detection results of heart rate under the cooperation of ECG and PPG, such as improving the anti-noise ability and heart rate detection range. However, the existing technology does not consider the impact of heart rate differences between individuals and situational factors on the accuracy of heart rate acquisition, and does not consider the influence of the physiological parameter of heart rate variability (Autonomic Nervous System, ANS) in heart rate estimation. Moreover, its effect on the heart rate estimation deviation caused by individual differences is poor, and it does not fully consider the impact of heart rate caused by exercise or different intensity actions, resulting in low accuracy of heart rate acquisition.

[0061] Refer to Figure 1 , the first embodiment of the present invention provides a method for improving the accuracy of heart rate acquisition, including the following steps:

[0062] S11, obtaining an electrocardiogram signal and a photoplethysmogram signal;

[0063] S12, calculating the pulse rate based on the photoplethysmogram signal, and obtaining a heart rate variability parameter through heart rate variability analysis;

[0064] S13, calculating the heart rate based on the electrocardiogram signal;

[0065] S14, obtaining a heart rate category classification through a linear discriminant analysis algorithm based on the heart rate variability parameter;

[0066] S15, obtaining a heart rate prediction value through a heart rate model based on the heart rate;

[0067] S16, obtaining a final heart rate output value through an error backpropagation neural network based on the heart rate prediction value, the heart rate category classification, the photoplethysmogram signal, and the heart rate.

[0068] In step S11, an electrocardiogram signal and a photoplethysmogram signal are acquired.

[0069] In one implementation, an intelligent wearable device such as a smart bracelet, a smart watch, or a professional electrocardiograph is used to acquire the electrocardiogram signal. One or more electrocardiogram electrodes are placed on the user's chest, usually beside the sternum or on the left chest; a reference electrode is attached to a position far from the heart, such as the arm or leg; a ground electrode is placed on the abdomen or other appropriate positions to provide circuit grounding. The electrical signals of the heart are collected through the electrodes and digitally sampled. The sampling frequency is 250 Hz to ensure that sufficient detailed electrocardiogram information is captured. Further, an intelligent wearable device such as a smart bracelet or a smart watch is used to acquire the photoplethysmogram signal, and its sensor is a light-emitting diode sensor or a photodiode. The intelligent wearable device is worn on the user's arm, usually at the wrist, ensuring that the device is worn firmly to avoid the absorption of light by the back of the hand or the bracelet affecting the light. At the same time, the light intensity emitted by the light-emitting diode is about 60 mW. The light-emitting diode emits a light signal to a certain fixed position point under the skin. Part of the light signal is absorbed by the blood vessels in the skin, resulting in a reduction in the light flux. The photodiode receives the light signal transmitted through the skin and converts it into an electrical signal. The signal received by the photodiode is digitally sampled, and the common sampling frequency is 100 Hz.

[0070] In step S12, the pulse rate is calculated based on the photoplethysmogram signal, and the heart rate variability parameters are obtained through heart rate variability analysis.

[0071] In one implementation, the photoplethysmogram signal is subjected to frequency domain filtering to obtain a filtered photoplethysmogram signal;

[0072] The pulse rate is calculated by the following formula:

[0073]

[0074] where R bpm is the pulse rate, and T rp is the time interval between two adjacent peaks in the filtered photoplethysmogram signal;

[0075] Based on the filtered photoplethysmogram signal, the heartbeat peak interval of each heartbeat cycle is extracted;

[0076] The heart rate variability parameters include respiratory sinus arrhythmia, the root mean square of the difference in adjacent heartbeat peak intervals, and the standard deviation of normal heartbeat peak intervals;

[0077] Respiratory sinus arrhythmia is calculated by the following formula:

[0078]

[0079] Among them, RSA is respiratory sinus arrhythmia, and F H is the high-frequency proportion of the filtered photoplethysmogram signal, and F L is the low-frequency proportion of the filtered photoplethysmogram signal;

[0080] Calculate the root mean square of the difference in adjacent heartbeat peak intervals through the root mean square formula;

[0081] Calculate the standard deviation of the normal heartbeat peak interval through the standard deviation formula.

[0082] It should be noted that the frequency-domain filtering process uses the fast Fourier transform method to remove noises such as baseline drift, electromyogram interference, motion artifacts, and power frequency interference. Respiratory sinus arrhythmia reflects the influence of respiration on heart rate. For the filtered photoplethysmogram signal, according to the standard frequency band division, the high-frequency and low-frequency parts are determined, where the high-frequency part corresponds to the respiratory frequency and the low-frequency part corresponds to the sympathetic and parasympathetic nerve activities. The standard deviation of the normal heartbeat peak interval reflects the overall heart rate fluctuation degree. The root mean square of the difference in adjacent heartbeat peak intervals reflects the short-term heart rate fluctuation characteristics. These parameters can comprehensively reflect the user's heart rate condition and the regulation ability of the autonomic nervous system, thereby improving the accuracy and reliability of heart rate monitoring.

[0083] In step S13, the heart rate is calculated according to the electrocardiogram signal.

[0084] In one implementation, perform frequency-domain filtering processing on the electrocardiogram signal to obtain a filtered electrocardiogram signal;

[0085] The heart rate is calculated through the following formula:

[0086]

[0087] Among them, H r is the heart rate, and T r is the time interval between two adjacent peaks in the filtered electrocardiogram signal.

[0088] It should be noted that a band-pass filter is used to filter the electrocardiogram signal in the range of 0.5 Hz - 40 Hz to remove noises such as baseline drift, electromyogram interference, motion artifacts, and power frequency interference. At the same time, methods such as the threshold method, derivative method, adaptive threshold method, or template matching method can be used to determine the peaks of the electrocardiogram signal.

[0089] In step S14, according to the heart rate variability parameters, a heart rate category classification is obtained through a linear discriminant analysis algorithm.

[0090] In one implementation, the heart rate category classification includes: resting heart rate, active heart rate, and exercise heart rate;

[0091] When the value of respiratory sinus arrhythmia and the root mean square of the difference in adjacent heartbeat peak intervals in the heart rate variability parameters are higher than a preset threshold, and the standard deviation of the normal heartbeat peak intervals is less than the preset threshold, it is determined as exercise heart rate; when the value of respiratory sinus arrhythmia and the root mean square of the difference in adjacent heartbeat peak intervals in the heart rate variability parameters are lower than the preset threshold, and the standard deviation of the normal heartbeat peak intervals is greater than the preset threshold, it is determined as resting heart rate.

[0092] It should be noted that respiratory sinus arrhythmia reflects the influence of respiration on heart rate. Usually during respiration, the heart rate changes with the respiration. A high value of respiratory sinus arrhythmia usually indicates that the heart is strongly affected by respiration, which is common in the exercise state; a low value of respiratory sinus arrhythmia indicates that the heart is less affected by respiration, which is common in the resting state. The root mean square of the difference in adjacent heartbeat peak intervals reflects the short-term heart rate fluctuation characteristics. A high root mean square value of the difference in adjacent heartbeat peak intervals usually indicates large heart rate fluctuations, which is common in the exercise state; a low root mean square value of the difference in adjacent heartbeat peak intervals indicates small heart rate fluctuations, which is common in the resting state. The standard deviation of normal heartbeat peak intervals reflects the overall heart rate fluctuation degree. A high standard deviation value of normal heartbeat peak intervals usually indicates large heart rate fluctuations, which is common in the resting state; a low standard deviation value of normal heartbeat peak intervals indicates small heart rate fluctuations, which is common in the exercise state.

[0093] In one embodiment, the respiratory sinus rhythm, the root mean square of the difference in adjacent heartbeat peak intervals, and the standard deviation of the normal heartbeat peak intervals are calculated respectively, and by comparing with the preset threshold, these thresholds can be determined according to experimental data. Since different individuals have different physiques, the setting of the threshold is related to the population, age, and race, and can be adjusted stage by stage as the usage time of the product increases and the amount of actual usage data of the user increases, so as to achieve the goal of high accuracy in threshold setting. When the value of respiratory sinus arrhythmia and the root mean square of the difference in adjacent heartbeat peak intervals in the heart rate variability parameters are higher than the preset threshold, and the standard deviation of the normal heartbeat peak intervals is less than the preset threshold, it is determined as exercise heart rate; when the value of respiratory sinus arrhythmia and the root mean square of the difference in adjacent heartbeat peak intervals in the heart rate variability parameters are lower than the preset threshold, and the standard deviation of the normal heartbeat peak intervals is greater than the preset threshold, it is determined as resting heart rate. When the above two conditions are not met, it is determined as active heart rate.

[0094] In step S15, according to the heart rate, a heart rate prediction value is obtained through a heart rate model.

[0095] In one embodiment, the heart rate model formula is:

[0096] PPG = c 1 ·HR + c2 ·HR 2 -c 3 ·HR 3 +c 4 ·HR 4 -c 5 ·HR 5

[0097] where HR is the heart rate, PPG is the predicted heart rate, and c 1 、c 2 、c 3 、c 4 and c 5 are set coefficients;

[0098] Among them, the heart rate model is obtained by training with historical big data, and the heart rate model can obtain a heart rate prediction value according to the input heart rate. This model is constructed based on the Kubicek model, uses the electrocardiogram signal to calculate the mean heart rate, calculates the average value of HR as the input, and calculates the heart rate variance as an influencing factor of the heart rate prediction value. The mean heart rate reflects the average level of the user's heart rate over a period of time. The heart rate variance reflects the degree of heart rate fluctuation. A larger variance indicates a larger change in heart rate, which is affected by factors such as activity status and emotion. The mean heart rate and heart rate variance are used as input features to train the heart rate prediction model. A machine learning algorithm is used to construct the prediction model.

[0099] In one embodiment, c 1 = 1.41, c 2 = 0.013, c 3 = 0.007, c 4 = 0.0001, and c 5 = 0.000001. The parameters are obtained by collecting a large amount of experimental data, including the actually measured heart rate and the corresponding photoplethysmogram signal. These data cover different heart rate ranges and different activity states including quiet, active, and exercise. The robustness of the model is improved by collecting more data. Further, regularization techniques are used to prevent overfitting, and finally cross-validation is used to select the best model parameters.

[0100] In step S16, according to the heart rate prediction value, the heart rate category classification, the photoplethysmogram signal, and the heart rate, the final heart rate output value is obtained through an error backpropagation neural network.

[0101] In one implementation, the heart rate prediction value, the heart rate category classification, the photoplethysmogram signal, and the heart rate are used as the neurons in the input layer of the error backpropagation neural network; through the error backpropagation neural network, the neural network predicted heart rate is obtained; when the heart rate category classification is the active heart rate, the neural network predicted heart rate is used as the final heart rate output value.

[0102] In one implementation, a BP neural network model is constructed, including an input layer, a hidden layer, and an output layer, where the number of neurons in the input layer is 3, the number of neurons in the output layer is 1, and the activation function is selected as Sigmoid; the data is divided into a training set and a test set in a ratio of 3:1, and the BP neural network is trained; the trained BP neural network model is used to output the final heart rate prediction value, and in this step, only when the heart rate data is the quiet value, the heart rate calculated value is taken as the output value.

[0103] In one implementation, when the heart rate class

[0104] When the heart rate category is classified as a resting heart rate, the heart rate prediction value is used as the final heart rate output value; when the heart rate category is classified as an active heart rate, the heart rate prediction value of the neural network is used as the final heart rate output value. In a resting state, the heart rate is relatively stable, and the heart rate variability parameters can better reflect the state of the heart. Therefore, the heart rate prediction value calculated based on these parameters can provide higher accuracy. Directly using the formula to calculate the heart rate prediction value avoids the complex machine learning model reasoning process and improves the real-time performance and efficiency of the system. In an active state, the heart rate changes greatly and is affected by many factors. The neural network model can capture these complex change patterns and provide more accurate heart rate prediction. The neural network model has been trained with a large amount of data and has strong learning and generalization capabilities, and can provide high-precision heart rate prediction under different activity states. In an active state, the user's heart rate changes greatly and is affected by many factors, such as exercise intensity, emotional fluctuations, etc. In this case, relying solely on the heart rate variability parameters cannot accurately reflect the current heart rate state. Therefore, introducing a neural network model for heart rate prediction can better capture the dynamic changes of heart rate. Furthermore, under the condition of exercise heart rate, when the final heart rate output value is also selected, the predicted heart rate of the neural network model is used as the final output. In the state of exercise, the heart rate changes greatly and frequently, and is affected by many factors, such as exercise intensity, breathing rate, body position changes, etc., with high variability. At the same time, the relationship between heart rate and these factors is often nonlinear, and traditional linear models are difficult to accurately capture this complex relationship. Heart rate monitoring in exercise requires high real-time performance to provide timely feedback on the user's heart condition. Neural networks (especially deep learning models) can process highly nonlinear data and learn complex patterns from them. They can also consider multiple input features (such as ECG signals, PPG signals, accelerometer data, etc.) at the same time, and automatically extract the most relevant feature combination, thereby improving prediction accuracy.

[0105] In summary, the present invention aims to solve the problem of low accuracy of heart rate collection by integrating electrocardiogram (ECG) and photoplethysmography (PPG) technology, introducing machine learning models, identifying activity states of measured values, and realizing heart rate prediction and acquisition in different situations, so as to make the heart rate value more accurate. The system includes three main modules: ECG and PPG signal acquisition module, data processing module, and machine learning model training and implementation module. When faced with situations where the heart rate changes greatly, such as exercise, emergencies, or emotional fluctuations, the heart rate calculated by traditional physiological models often has large errors. In order to solve this problem, this method innovatively introduces machine learning

[0106] It combines physical models with machine learning algorithms to accurately identify activity states, thereby effectively improving the accuracy of heart rate data.

[0107] Specifically, the present invention enhances the comprehensiveness and real-time nature of heart rate data acquisition by synchronously collecting ECG and PPG signals on a smart wearable device. These two signals respectively reflect the changes in cardiac electrical signals and blood volume changes, and their combination can provide richer physiological information. In addition, the introduction of machine learning models, especially the BP neural network model, can not only deeply process and analyze the data, but also improve the accuracy of heart rate calculation. As a powerful non-linear model, the BP neural network can find hidden patterns in complex data, thus better predicting the heart rate.

[0108] To further improve the ability to contextually interpret heart rate data, the present invention also uses the Linear Discriminant Analysis (LDA) algorithm to classify heart rate variability parameters. The LDA algorithm can classify the user's state into three categories: quiet, active, and exercising, according to different characteristics of heart rate variability parameters. This classification method helps to understand the changing patterns of heart rate in different activity states of the user, and thus provides more personalized health advice.

[0109] By integrating ECG and PPG technologies, the present invention achieves the comprehensiveness and real-time nature of heart rate data acquisition; by introducing machine learning models, especially the BP neural network model, it improves the accuracy of heart rate calculation; by classifying heart rate variability parameters using the LDA algorithm, it realizes the contextual interpretation of heart rate data. The combination of these technologies not only solves the inaccuracy problems existing in traditional heart rate monitoring, but also provides a more reliable and intelligent heart rate monitoring solution for users. In addition, the present invention particularly focuses on the parameter analysis of heart rate variability. Through the comprehensive analysis of parameters such as the RR interval, respiratory sinus arrhythmia (RSA), root mean square of the differences between adjacent heartbeat peak intervals (RMSSD), and standard deviation of normal heartbeat peak intervals (SDNN), the user's cardiac health status can be more comprehensively evaluated. These parameters not only reflect the immediate state of the heart, but also reveal the regulatory effect of the autonomic nervous system on the heart, thus providing an important reference basis for medical diagnosis and health management.

[0110] In summary, the present invention significantly improves the accuracy and reliability of heart rate monitoring through multi-modal signal fusion, the application of machine learning models, and contextual data analysis. This method is not only applicable to daily health monitoring, but also plays an important role in the clinical environment, providing more refined management and treatment plans for heart disease patients. In the future, with the continuous progress of technology and the expansion of application scenarios, this heart rate monitoring method based on the integration of ECG and PPG is expected to become one of the standards in the field of heart rate monitoring, safeguarding people's health.

[0111] Refer to Figure 2 , the second embodiment of the present invention provides a system for improving the accuracy of heart rate acquisition, including:

[0112] A data acquisition module, configured to acquire electrocardiogram signals and photoplethysmogram signals;

[0113] A heart rate parameter module, configured to calculate a pulse rate based on the photoplethysmogram signal and obtain heart rate variability parameters through heart rate variability analysis;

[0114] A heart rate calculation module, configured to calculate a heart rate based on the electrocardiogram signal;

[0115] A heart rate classification module, configured to obtain a heart rate category classification through a linear discriminant analysis algorithm based on the heart rate variability parameters;

[0116] A heart rate prediction module, configured to obtain a heart rate prediction value through a heart rate model based on the heart rate;

[0117] A heart rate output module, configured to obtain a final heart rate output value through an error backpropagation neural network based on the heart rate prediction value, the heart rate category classification, the photoplethysmogram signal, and the heart rate.

[0118] Preferably, the data acquisition module is configured to:

[0119] Acquire electrocardiogram signals and photoplethysmogram signals.

[0120] In one implementation, an electrocardiogram (ECG) signal is acquired using a smart wearable device such as a smart bracelet, a smart watch, or a professional electrocardiograph. The specific operations are as follows: One or more electrocardiogram electrodes are placed on the user's chest, usually at the parasternal or left chest position, to capture the electrical signals of the heart; at the same time, a reference electrode is attached to a position far from the heart, such as the arm or leg, to provide a stable reference point; and the ground electrode is placed on the abdomen or other appropriate position to provide circuit grounding. Through these electrodes, the electrical signals of the heart can be collected and digitally sampled. To ensure that sufficient detailed electrocardiogram information is captured, the sampling frequency is set to 250 Hz.

[0121] In one implementation, an intelligent wearable device such as a smart bracelet or a smart watch is used to acquire photoplethysmogram (PPG) signals. These devices are usually equipped with light-emitting diode sensors and photodiodes. The user needs to wear the intelligent wearable device on the arm, usually at the wrist, ensuring that the device is stable and not affected by the absorption of light by the back of the hand or the bracelet, which may affect the measurement results. The light-emitting diode emits a light signal with an intensity of about 60 mW, irradiating a fixed position point under the skin. Part of the light signal is absorbed by the blood vessels in the skin, resulting in a reduction in light flux. The photodiode receives the light signal passing through the skin and converts it into an electrical signal. Subsequently, the signal received by the photodiode is digitally sampled, and a common sampling frequency is 100 Hz. This dual-modal signal acquisition method combines the advantages of ECG and PPG, and can provide more comprehensive and accurate heart rate monitoring data. The ECG signal provides detailed information about the cardiac electrical signal, and the R wave can be accurately detected to calculate the heart rate. The PPG signal reflects the change in blood volume, can continuously monitor the heart rate change for a long time, and is suitable for non-invasive daily wear. Through the fusion of these two signals, not only can the accuracy of heart rate monitoring be improved, but also the performance of the heart in different activity states can be better understood. In practical applications, after these signals are preprocessed (such as filtering and smoothing), they can be further used for heart rate variability analysis, activity state recognition, and heart rate prediction. By calculating and analyzing heart rate variability parameters (such as RSA, RMSSD, and SDNN), a deeper understanding of the user's cardiac health status can be obtained

[0122] It is worth noting that, combined with a machine learning model such as a neural network, more accurate heart rate prediction can be provided in complex situations, thus realizing personalized health management. In short, this implementation not only improves the accuracy and reliability of heart rate monitoring, but also enhances the practicality and user experience of the system, and is applicable to various application scenarios from daily health monitoring to clinical diagnosis.

[0123] Preferably, the heart rate parameter module is used for:

[0124] Calculating the pulse rate according to the photoplethysmogram signal, and obtaining heart rate variability parameters through heart rate variability analysis, including:

[0125] Performing frequency-domain filtering on the photoplethysmogram signal to obtain a filtered photoplethysmogram signal;

[0126] Calculating the pulse rate through the following formula:

[0127]

[0128] where R bpm is the pulse rate, and T rpAccording to the time interval between two adjacent peaks in the filtered photoplethysmogram signal;

[0129] Extracting the heartbeat peak interval of each heartbeat cycle according to the filtered photoplethysmogram signal;

[0130] The heart rate variability parameters include respiratory sinus arrhythmia, the root mean square of the difference between the peak intervals of adjacent heartbeats, and the standard deviation of the normal peak intervals of heartbeats;

[0131] Respiratory sinus arrhythmia was calculated by the following formula:

[0132]

[0133] Among them, RSA is respiratory sinus arrhythmia, F H is the high frequency proportion of the filtered photoplethysmogram signal, F L is the low frequency proportion of the filtered photoplethysmogram signal;

[0134] The root mean square of the difference between the intervals between adjacent heartbeat peaks is calculated by the root mean square formula;

[0135] The standard deviation of the normal heartbeat peak interval is calculated using the standard deviation formula.

[0136] It is worth noting that respiratory sinus arrhythmia (RSA) can reflect the impact of breathing on heart rate. During breathing, heart rate often changes with the rhythm of breathing. Specifically, a higher RSA value usually means that the heart is strongly affected by breathing, which is common in exercise; on the contrary, a lower RSA value indicates that the heart is less affected by breathing, which is usually a characteristic of a quiet state. The root mean square difference (RMSSD) of the intervals between adjacent heartbeat peaks is an important indicator for measuring the characteristics of short-term heart rate fluctuations. A high RMSSD value usually indicates a large heart rate fluctuation, which is particularly obvious in exercise; while a low RMSSD value indicates a small heart rate fluctuation, which is common in a quiet state. The standard deviation of the normal heartbeat peak interval (SDNN) reflects the overall degree of heart rate fluctuation. A high SDNN value usually indicates a large heart rate fluctuation, which is more common in a quiet state; while a low SDNN value indicates a small heart rate fluctuation, which is common in an exercise state. Through the comprehensive analysis of these parameters, we can have a more comprehensive understanding of the characteristics of heart rate changes under different activity states, thereby providing an important reference for health monitoring and medical diagnosis. Whether it is a steady heart rate at rest or drastic fluctuations during exercise, these parameters can reveal the dynamic performance of the heart in different situations.

[0137] Preferably, the heart rate calculation module is used to:

[0138] Calculating the heart rate according to the electrocardiogram signal includes:

[0139] Perform frequency-domain filtering on the electrocardiogram signal to obtain a filtered electrocardiogram signal;

[0140] Calculate the heart rate through the following formula:

[0141]

[0142] where H r is the heart rate, and T r is the time interval between two adjacent peaks in the filtered electrocardiogram signal.

[0143] Preferably, the heart rate classification module is used for:

[0144] Obtain a heart rate category classification through a linear discriminant analysis algorithm according to the heart rate variability parameter, including:

[0145] The heart rate category classification includes: resting heart rate, active heart rate, and exercise heart rate;

[0146] When the value of respiratory sinus arrhythmia and the root mean square of the difference in adjacent heartbeat peak intervals in the heart rate variability parameter are higher than a preset threshold, and the standard deviation of the normal heartbeat peak interval is less than the preset threshold, it is determined as the exercise heart rate;

[0147] When the value of respiratory sinus arrhythmia and the root mean square of the difference in adjacent heartbeat peak intervals in the heart rate variability parameter are lower than the preset threshold, and the standard deviation of the normal heartbeat peak interval is greater than the preset threshold, it is determined as the resting heart rate.

[0148] Preferably, the heart rate prediction module is used for:

[0149] Obtain a heart rate prediction value through a heart rate model according to the heart rate, including:

[0150] The heart rate model formula is:

[0151] PPG = c 1 ·HR + c 2 ·HR 2 - c 3 ·HR 3 + c 4 ·HR 4 - c 5 ·HR 5

[0152] where HR is the heart rate, PPG is the predicted heart rate, and c 1 , c 2 , c 3 , c 4 and c 5 are set coefficients;

[0153] Among them, the heart rate model is trained through historical big data, and the heart rate model can obtain a heart rate prediction value according to the input heart rate.

[0154] Preferably, the heart rate output module is used for:

[0155] According to the heart rate prediction value, the heart rate category classification, the photoplethysmogram signal, and the heart rate, through an error backpropagation neural network, obtain a final heart rate output value, including:

[0156] Take the heart rate prediction value, the heart rate category classification, the photoplethysmogram signal, and the heart rate as the neurons of the input layer of the error backpropagation neural network;

[0157] Through the error backpropagation neural network, obtain the neural network predicted heart rate;

[0158] When the heart rate category classification is the active heart rate, take the neural network predicted heart rate as the final heart rate output value.

[0159] Preferably, when the heart rate category classification is the resting heart rate, take the heart rate prediction value as the final heart rate output value; when the heart rate category classification is the active heart rate, take the neural network predicted heart rate as the final heart rate output value.

[0160] It should be noted that heart rate, as an important health indicator, its accurate measurement is crucial for health management. Although existing smart wearable devices have made significant progress in heart rate monitoring, there are still challenges in the face of individual differences, situational factors, and heart rate variability. By introducing personalized calibration, considering situational factors, heart rate variability analysis, improving algorithm models, and data augmentation and preprocessing, etc., the reliability and accuracy of heart rate data can be significantly improved. The development of these technologies not only helps to improve the level of personal health management, but also provides strong support for clinical diagnosis and medical research.

[0161] It should be noted that a system for improving the accuracy of heart rate acquisition provided by an embodiment of the present invention is used to execute all the process steps of a method for improving the accuracy of heart rate acquisition in the above embodiment, and the working principles and beneficial effects of the two correspond one by one, so they will not be elaborated here.

[0162] An embodiment of the present invention also provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a heart rate prediction program. When the processor executes the computer program, it implements the steps in the above embodiments of the method for improving the accuracy of heart rate acquisition, such as Figure 1The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as the data acquisition module.

[0163] Exemplarily, the computer program may be divided into one or more modules / units, and the one or more modules / units are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and these instruction segments are used to describe the execution process of the computer program in the electronic device.

[0164] The electronic device may be a desktop computer

[0165] such as a notebook, a palm computer, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the above components are only examples of the electronic device and do not constitute a limitation on the electronic device. It may include more or fewer components than the above, or combine certain components, or different components. For example, the electronic device may further include input / output devices, network access devices, a bus, etc.

[0166] The so-called processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor is the control center of the electronic device and connects various parts of the entire electronic device through various interfaces and lines.

[0167] The memory can be used to store the computer programs and / or modules. By running or executing the computer programs and / or modules stored in the memory, and invoking the data stored in the memory, the processor realizes various functions of the electronic device. The memory mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function (such as a sound playback function, an image playback function, etc.); the data storage area can store data created according to the use of the mobile phone (such as audio data, phone book, etc.). In addition, the memory can include high-speed random access memory, and can also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, at least one magnetic disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0168] Among them, if the modules / units integrated in the electronic device are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above-mentioned embodiment methods of the present invention, it can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-mentioned various method embodiments can be realized. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, an executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disc, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0169] It should be noted that the device embodiments described above are merely illustrative. 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 to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the attached drawings of the device embodiments provided by the present invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art can understand and implement without creative efforts.

[0170] The specific embodiments described above further elaborate on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. In particular, for those skilled in the art, any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for improving the accuracy of heart rate collection, characterized in that: include: Acquire electrocardiogram signals and photoplethysmogram signals; Calculating the pulse rate according to the photoplethysmographic signal, and obtaining the heart rate variability parameter through heart rate variability analysis; Calculating the heart rate according to the electrocardiogram signal; According to the heart rate variability parameters, a heart rate category classification is obtained by using a linear discriminant analysis algorithm; According to the heart rate, a predicted heart rate value is obtained through a heart rate model; Obtaining a final heart rate output value through an error back propagation neural network according to the heart rate prediction value, the heart rate category classification, the photoplethysmography signal and the heart rate; Wherein, the heart rate category classification is obtained according to the heart rate variability parameter by using a linear discriminant analysis algorithm, including: The heart rate categories include: resting heart rate, active heart rate and exercise heart rate; When the value of respiratory sinus arrhythmia and the root mean square of the difference between the peak intervals of adjacent heartbeats in the heart rate variability parameter are higher than the preset threshold, and the standard deviation of the normal heartbeat peak interval is less than the preset threshold, it is determined to be an exercise heart rate; When the value of respiratory sinus arrhythmia and the root mean square of the difference between the peak intervals of adjacent heartbeats in the heart rate variability parameter are lower than the preset threshold, and the standard deviation of the normal heartbeat peak interval is greater than the preset threshold, it is determined to be a resting heart rate; When the above two conditions are not met, it is determined to be active heart rate; Wherein, the heart rate model formula is: PPG=c1·HR+c2·HR 2 -c3·HR 3 +c4·HR 4 -c5·HR 5 Among them, HR is the heart rate, PPG is the predicted heart rate, and c1, c2, c3, c4 and c5 are setting coefficients.

2. The method for improving the accuracy of heart rate acquisition according to claim 1, characterized in that: The pulse rate is calculated according to the photoplethysmogram signal, and the heart rate variability parameter is obtained by heart rate variability analysis, including: Performing frequency domain filtering on the photoplethysmogram signal to obtain a filtered photoplethysmogram signal; The pulse rate is calculated using the following formula: Among them, R bpm is the pulse rate, T rp According to the time interval between two adjacent peaks in the filtered photoplethysmogram signal; Extracting the heartbeat peak interval of each heartbeat cycle according to the filtered photoplethysmogram signal; The heart rate variability parameters include respiratory sinus arrhythmia, the root mean square of the difference between the peak intervals of adjacent heartbeats, and the standard deviation of the normal peak intervals of heartbeats; Respiratory sinus arrhythmia was calculated by the following formula: Among them, RSA is respiratory sinus arrhythmia, F H is the high frequency proportion of the filtered photoplethysmogram signal, F L is the low frequency proportion of the filtered photoplethysmogram signal; The root mean square of the difference between the peak intervals of adjacent heartbeats is calculated by the root mean square formula; The standard deviation of the normal heartbeat peak interval is calculated using the standard deviation formula.

3. The method for improving the accuracy of heart rate acquisition according to claim 1, characterized in that: The step of obtaining a predicted heart rate value based on the heart rate through a heart rate model includes: The heart rate model is obtained through historical big data training, and the heart rate model can obtain a heart rate prediction value according to an input heart rate.

4. The method for improving the accuracy of heart rate acquisition according to claim 1, characterized in that: The method of obtaining a final heart rate output value according to the heart rate prediction value, the heart rate category classification, the photoplethysmogram signal and the heart rate through an error back propagation neural network includes: Using the heart rate prediction value, the heart rate category classification, the photoplethysmography signal and the heart rate as neurons of the input layer of the error back propagation neural network; Obtaining a neural network predicted heart rate through the error back propagation neural network; When the heart rate category is classified as active heart rate, the heart rate predicted by the neural network is used as the final heart rate output value.

5. The method for improving the accuracy of heart rate acquisition according to claim 1, characterized in that: The step of calculating the heart rate according to the electrocardiogram signal comprises: Performing frequency domain filtering on the electrocardiogram signal to obtain a filtered electrocardiogram signal; The heart rate is calculated using the following formula: Among them, H r is the heart rate, T r According to the time interval between two adjacent peaks in the filtered electrocardiogram signal.

6. The method for improving the accuracy of heart rate acquisition according to claim 4, characterized in that: The method further comprises: When the heart rate category is classified as a resting heart rate, the heart rate prediction value is used as the final heart rate output value.

7. A system for improving the accuracy of heart rate collection, used to implement the method for improving the accuracy of heart rate collection as described in any one of claims 1 to 6, characterized in that: include: A data acquisition module, used for acquiring electrocardiogram signals and photoelectric volume pulse wave signals; A heart rate parameter module, used to calculate the pulse rate according to the photoplethysmography signal, and obtain the heart rate variability parameter through heart rate variability analysis; A heart rate calculation module, used to calculate the heart rate according to the electrocardiogram signal; A heart rate classification module, used to obtain a heart rate category classification according to the heart rate variability parameter by a linear discriminant analysis algorithm; A heart rate prediction module, used to obtain a heart rate prediction value according to the heart rate through a heart rate model; The heart rate output module is used to obtain a final heart rate output value through an error back propagation neural network according to the heart rate prediction value, the heart rate category classification, the photoplethysmogram signal and the heart rate.

8. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for improving the accuracy of heart rate collection as claimed in any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the method for improving the accuracy of heart rate collection as described in any one of claims 1 to 6.

Images