A heart rate data analysis system and method

Abstract

This invention relates to the field of data processing technology and proposes a heart rate data analysis system and method, comprising: acquiring heart rate monitoring signals through a wearable heart rate monitoring device and recording motion acceleration data at various times; analyzing the dispersion and differences of heart rate monitoring data during the monitoring period and quantifying the intensity index of heart rate changes during the monitoring period; obtaining the motion transformation index of the monitoring period based on the dispersion of motion acceleration data during the monitoring period; obtaining the probability of interference sources for each independent component signal; filtering several interference source component signals based on the probability of interference sources; obtaining the frequency decoupling degree of each interference source component signal; obtaining the preprocessing requirement degree of each interference source component signal; smoothing based on the preprocessing requirement degree, reconstructing the processed heart rate monitoring signal, and performing data analysis. This invention aims to solve the problem of motion artifacts generated by the influence of exercise patterns in heart rate data monitoring.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a heart rate data analysis system and method. Background Technology

[0002] Heart rate (HR) is an important physiological parameter reflecting cardiac health, commonly used to monitor an individual's cardiovascular health, exercise status, and psychological stress. With increasing awareness of health management, heart rate data analysis systems are widely used in medicine, fitness, exercise analysis, and health monitoring. Traditional heart rate monitoring relies heavily on specialized equipment in medical institutions, but with the rapid development of wearable technology and the increasing popularity of devices such as smart bracelets and smartwatches, users can obtain their own heart rate data in real time. These devices not only monitor heart rate changes in real time but also analyze the user's exercise status and provide health advice, thereby helping users better manage their health.

[0003] Existing methods for analyzing heart rate data typically employ fixed templates or uniform algorithms, failing to adequately consider the impact of behavioral patterns on heart rate fluctuations and hindering the provision of refined, personalized risk assessments. Furthermore, current heart rate data analysis systems often rely on ECG (electrocardiogram) or PPG (photoplethysmography) signals collected by wearable devices. These signals are susceptible to various non-physiological factors during acquisition, such as motion artifacts, ambient light interference, changes in contact resistance, and sensor drift. This results in abnormal characteristics in the heart rate data from the initial generation stage, including noise superposition, baseline drift, peak distortion, or signal truncation, thereby affecting the accuracy of subsequent heart rate data analysis. Summary of the Invention

[0004] This invention provides a heart rate data analysis method and system to solve the problem of motion artifacts caused by the influence of exercise patterns in existing heart rate data monitoring. The specific technical solution adopted is as follows:

[0005] This invention proposes a method for analyzing heart rate data, which includes the following steps:

[0006] Heart rate monitoring signals are collected using wearable heart rate monitoring devices to obtain heart rate monitoring data for the monitoring period and several moments within it, and motion acceleration data at each moment are recorded.

[0007] The dispersion and differences in heart rate monitoring data during the monitoring period are analyzed to quantify the intensity index of heart rate changes during the monitoring period. Based on the dispersion of motion acceleration data during the monitoring period and the time difference of its dispersion relationship with heart rate monitoring data, combined with the intensity index of heart rate changes, the motion transformation index of the monitoring period is obtained. Based on the temporal distribution of extreme points in each independent component signal of heart rate monitoring data and the area difference between peak regions, the mixing degree of each independent component signal is obtained, and then combined with the motion transformation index, the probability of interference sources of each independent component signal is obtained.

[0008] Several interference source component signals are selected based on the probability of interference sources; several IMF component signals are obtained by EMD decomposition of each interference source component signal; based on the correlation between the instantaneous frequency performance of each IMF component signal and the order value of the IMF component signal, and combined with the quality difference index of the data points in each IMF component signal, the frequency decoupling degree of each interference source component signal is obtained; combined with the probability of interference sources of the interference source component signals, the preprocessing requirements of each interference source component signal are obtained.

[0009] The signals of each interference source are smoothed based on the preprocessing requirements, and the processed heart rate monitoring signal is obtained by reconstructing the signals from the independent component signals.

[0010] Optionally, the method for obtaining the heart rate variability index during the monitoring period is as follows:

[0011] All heart rate monitoring data during the monitoring period are arranged in ascending order to obtain an ascending heart rate sequence. The difference between any two adjacent heart rate monitoring data in the ascending heart rate sequence is obtained. The two adjacent heart rate monitoring data corresponding to the maximum difference between all two adjacent heart rate monitoring data are used as segmented heart rate data. The left segmented heart rate data in the two segmented heart rate data and all heart rate data in the left side of the ascending heart rate sequence constitute the low heart rate data set of the monitoring period. The right segmented heart rate data in the two segmented heart rate data and all heart rate data in the right side of the ascending heart rate sequence constitute the high heart rate data set of the monitoring period.

[0012] The difference between the mean of the heart rate monitoring data in the high heart rate dataset and the mean of the heart rate monitoring data in the low heart rate dataset is used as the heart rate dispersion coefficient for the monitoring period. The absolute value of the difference between the heart rate monitoring data at any two adjacent moments in the monitoring period is obtained, and the product of the mean of the absolute values ​​of the differences between the heart rate monitoring data at all two adjacent moments in the monitoring period and the heart rate dispersion coefficient is used as the heart rate variability index for the monitoring period.

[0013] Optionally, the specific method for obtaining the motion change index during the monitoring period includes:

[0014] All motion acceleration data during the monitoring period are arranged in ascending order to obtain an acceleration ascending sequence. Based on the difference between adjacent motion acceleration data in the acceleration ascending sequence, the low acceleration data set and the high acceleration data set for the monitoring period are obtained.

[0015] The absolute value of the difference between the minimum value of all motion acceleration data at the corresponding time in the high acceleration dataset and the minimum value of all heart rate monitoring data at the corresponding time in the high heart rate dataset is recorded as the high heart rate deviation time difference during the monitoring period; the absolute value of the difference between the number of motion acceleration data in the high acceleration dataset and the number of heart rate monitoring data in the high heart rate dataset is recorded as the high heart rate quantity deviation during the monitoring period.

[0016] Based on the intensity index of heart rate changes during the monitoring period, as well as the deviation of high heart rate from time and the deviation of high heart rate quantity, the exercise transformation index of the monitoring period is obtained; the exercise transformation index is positively correlated with the intensity index of heart rate changes, and negatively correlated with the deviation of high heart rate from time and the deviation of high heart rate quantity.

[0017] Optionally, the specific method for obtaining the mixing degree of each independent component signal includes:

[0018] The heart rate monitoring signal, composed of heart rate monitoring data during the monitoring period, was analyzed by ICA independent component analysis to obtain several independent component signals.

[0019] For any independent component signal, obtain several extreme points, including several maximum points and several minimum points, obtain the time interval between any two adjacent extreme points, and use the standard deviation of the time interval between all two adjacent extreme points as the extreme value distribution fluctuation coefficient of the independent component signal.

[0020] Based on the time corresponding to any maximum point and its two adjacent minimum points, the peak region of the maximum point is constructed, and the area of ​​the peak region of the maximum point is obtained by integration; the absolute value of the difference between the areas of the peak regions of any two maximum points is obtained, and the product of the mean of the absolute values ​​of the differences of all obtained areas and the extreme value distribution fluctuation coefficient is used as the mixing degree of the independent component signal.

[0021] Optionally, the specific method for obtaining the probability of interference sources for each independent component signal includes:

[0022] The product of the mixing degree of any independent component signal and the motion transformation index of the monitoring period is taken as the probability of interference source for that independent component signal.

[0023] Optionally, the specific method for filtering several interference source component signals based on the probability of the interference source includes:

[0024] Arrange all independent component signals in ascending order of their interference source probability to obtain a component signal sequence. Obtain the difference between the interference source probabilities of adjacent independent component signals in the component signal sequence. Take the rightmost independent component signal of the two independent component signals corresponding to the maximum value of the difference between all the interference source probabilities as the interference source component signal.

[0025] Other independent component signals to the right of the interference source component signal in the component signal sequence are taken as interference source component signals.

[0026] Optionally, the specific method for obtaining the frequency decoupling degree of each interference source component signal includes:

[0027] Based on the instantaneous frequency performance of each IMF component signal and the order value of the IMF component signals, the decomposition disorder of each interference source component signal is obtained.

[0028] For any IMF component signal of any interference source component signal, calculate the quality anomaly index for each data point, obtain the mean of the quality anomaly index of all data points in the IMF component signal, and use it as the periodic stability coefficient of the IMF component signal.

[0029] Based on the decomposition disorder of the interference source component signal and the mean periodic stability of all IMF component signals of the interference source component signal, the frequency decoupling degree of the interference source component signal is obtained. The frequency decoupling degree is negatively correlated with the decomposition disorder and positively correlated with the mean periodic stability.

[0030] Optionally, the specific method for obtaining the decomposition disorder of each interference source component signal includes:

[0031] Perform a Fourier transform on any IMF component signal of any interference source component signal to obtain several instantaneous frequencies of the IMF component signal, and take the average of all instantaneous frequencies as the frequency performance value of the IMF component signal.

[0032] The frequency performance values ​​of each IMF component signal in the interference source component signal are arranged according to the order of the IMF component signals to obtain the frequency performance sequence of the interference source component signal, and the order value sequence of each IMF component signal is obtained by arranging the order values ​​of the interference source component signal.

[0033] The Pearson correlation coefficient between the frequency performance sequence and the order value sequence is obtained, and normalized based on the range of the Pearson correlation coefficient. The normalization result is used as the decomposition disorder of the interference source component signal.

[0034] Optionally, the preprocessing requirements of each interference source component signal are obtained using the following method:

[0035] The Euclidean norm of the inversely proportional normalized result of the frequency decoupling degree of any interference source component signal and the normalized result of the interference source probability is used as the preprocessing requirement degree of that interference source component signal.

[0036] The present invention also proposes a heart rate data analysis system, which includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the above method.

[0037] The beneficial effects of this invention are as follows: This invention acquires heart rate monitoring signals in real time through a wearable heart rate monitoring device and records motion acceleration data. By analyzing the discrete distribution and fluctuation differences of heart rate monitoring data during the monitoring period, it quantifies the drastic changes in heart rate. Combined with the discrete performance of motion acceleration data and the time difference between the discrete distribution of heart rate monitoring data and the discrete distribution of heart rate monitoring data, it further obtains the motion transformation index to reflect the possibility that the heart rate monitoring data is affected by motion artifacts. Through independent component analysis, it quantifies the degree of mixing by analyzing the distribution of extreme points and the differences in the peak area of ​​independent component signals to reflect that the independent component signals do not conform to the independent decomposition process of independent components. Then, it combines the motion transformation index to present the possibility of interference sources. By performing local feature analysis of the quality index of the interfering source component signals based on EMD decomposition of the IMF component signals, and extracting frequency features to reflect whether the interfering source component signals are clearly decomposed and periodically stable, it obtains the frequency decoupling degree. The interfering source component signals with smaller frequency decoupling degrees and greater interference source possibility require greater filtering. Then, it reconstructs and obtains the processed heart rate monitoring signal to eliminate the influence of motion artifacts on heart rate monitoring data, thereby improving the accuracy of heart rate data analysis. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a schematic flowchart of a heart rate data analysis method provided in one embodiment of the present invention. Detailed Implementation

[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0041] Please see Figure 1 The diagram illustrates a flowchart of a heart rate data analysis method according to an embodiment of the present invention, which includes the following steps:

[0042] Step S001: Collect heart rate monitoring signals through a wearable heart rate monitoring device, obtain heart rate monitoring data for the monitoring period and several moments within it, and record motion acceleration data at each moment.

[0043] The purpose of this embodiment is to eliminate motion artifacts caused by factors such as exercise in the heart rate monitoring signals collected by wearable heart rate monitoring devices, thereby improving the accuracy of heart rate monitoring signals and thus improving the effect of data analysis.

[0044] Specifically, wearable heart rate monitoring devices (such as ECG wristbands and PPG smartwatches) collect users' heart rate data, obtaining raw physiological electrical signals or photovolume signals, which are then converted into the user's heart rate monitoring signal. The sampling time interval of the heart rate monitoring signal is set within the device, and acceleration data is recorded at the same sampling time interval. This data is obtained through the accelerometer built into the wristband or smartwatch, thus obtaining the motion acceleration data at each moment. The raw signal includes the timestamp, signal value, and channel number corresponding to the sampling point, which are then converted to obtain the heart rate signal.

[0045] Furthermore, the preset time period length is used. In this embodiment, the time period length is described as 10 minutes. The heart rate monitoring data corresponding to each moment of the collected heart rate monitoring signal is divided according to the time period length to obtain several monitoring time periods and the heart rate monitoring data of each moment in them. Then, a subsequent correction processing of the heart rate monitoring data is performed once for each monitoring time period. In this embodiment, the most recent 10-minute time period is used as the monitoring time period for subsequent processing.

[0046] Step S002: Analyze the dispersion and differences in heart rate monitoring data during the monitoring period, and quantify the intensity index of heart rate changes during the monitoring period; based on the dispersion of motion acceleration data during the monitoring period, and the time difference of its dispersion relationship with heart rate monitoring data, and combined with the intensity index of heart rate changes, obtain the motion transformation index of the monitoring period; based on the temporal distribution of extreme points in each independent component signal of heart rate monitoring data, and the area difference between peak regions, obtain the mixing degree of each independent component signal, and then combine with the motion transformation index to obtain the probability of interference sources for each independent component signal.

[0047] It should be noted that by analyzing the impact of exercise patterns on the fluctuation of heart rate monitoring data, we can provide data support for the identification of components in subsequent heart rate monitoring data. By introducing motion sensing features such as motion acceleration data, we can interpret the fluctuation patterns of heart rate signals behaviorally, distinguish non-physiological heart rate fluctuations caused by exercise from true heart rate changes, and significantly improve the interference targeting of subsequent decomposition processing.

[0048] Preferably, in one embodiment of the present invention, the specific method for analyzing the dispersion and differences in heart rate monitoring data during the monitoring period and quantifying the intensity index of heart rate changes during the monitoring period includes:

[0049] All heart rate monitoring data during the monitoring period are arranged in ascending order to obtain an ascending heart rate sequence. The difference between any two adjacent heart rate monitoring data in the ascending heart rate sequence is obtained (the difference between the previous and the next heart rate monitoring data). The two adjacent heart rate monitoring data corresponding to the maximum difference between all two adjacent heart rate monitoring data are used as segmented heart rate data. The left segmented heart rate data in the two segmented heart rate data and all heart rate data in the left side of the ascending heart rate sequence constitute the low heart rate data set of the monitoring period. The right segmented heart rate data in the two segmented heart rate data and all heart rate data in the right side of the ascending heart rate sequence constitute the high heart rate data set of the monitoring period.

[0050] Furthermore, the difference between the mean of the heart rate monitoring data in the high heart rate dataset and the mean of the heart rate monitoring data in the low heart rate dataset is used as the heart rate dispersion coefficient for the monitoring period. The absolute value of the difference between the heart rate monitoring data at any two adjacent moments in the monitoring period is obtained, and the product of the mean of the absolute values ​​of the differences between the heart rate monitoring data at all two adjacent moments in the monitoring period and the heart rate dispersion coefficient is used as the heart rate variability index for the monitoring period.

[0051] It should be noted that the greater the difference between the mean heart rate monitoring data of the high heart rate dataset and the low heart rate dataset, the more dispersed the heart rate monitoring data is during the monitoring period. At the same time, the greater the difference in the mean heart rate monitoring data between adjacent moments, the more frequent the drastic changes in heart rate monitoring data, and the greater the index of drastic heart rate changes.

[0052] It should be further noted that changes in a user's heart rate during the detection process are often caused by drastic changes in their own exercise state or changes in the heart rate data itself. Heart rate changes caused by drastic changes in their own exercise state are usually normal behavior and do not indicate abnormal behavior in the user's heart rate.

[0053] Preferably, in one embodiment of the present invention, based on the discrete characteristics of motion acceleration data during the monitoring period and the time difference of its discrete relationship with heart rate monitoring data, combined with the heart rate variability index, the motion transformation index for the monitoring period is obtained, including the following specific method:

[0054] All motion acceleration data during the monitoring period are arranged in ascending order to obtain an acceleration ascending sequence. Following the method used to obtain the low heart rate and high heart rate data sets for the monitoring period, the low acceleration and high acceleration data sets for the monitoring period are obtained based on the differences between adjacent motion acceleration data in the acceleration ascending sequence. The absolute value of the difference between the minimum value of all motion acceleration data at the corresponding time in the high acceleration data set and the minimum value of all heart rate monitoring data at the corresponding time in the high heart rate data set is recorded as the high heart rate deviation time difference for the monitoring period. The absolute value of the difference between the number of motion acceleration data in the high acceleration data set and the number of heart rate monitoring data in the high heart rate data set is recorded as the high heart rate quantity deviation for the monitoring period.

[0055] Furthermore, based on the intensity index of heart rate changes during the monitoring period, as well as the deviation of high heart rate from time and the deviation of high heart rate quantity, the exercise transformation index of the monitoring period is obtained; the exercise transformation index is positively correlated with the intensity index of heart rate changes, and negatively correlated with the deviation of high heart rate from time and the deviation of high heart rate quantity.

[0056] As an example, the product of the high heart rate deviation time difference and the high heart rate quantity deviation during the monitoring period is obtained, and the ratio of the heart rate change intensity index to the product is used as the motion change index during the monitoring period. It should be noted that, in order to avoid the denominator being 0 and the fraction being meaningless, a hyperparameter is added to the numerator and denominator respectively during the ratio calculation in this embodiment. The hyperparameter in this embodiment is described as 0.1.

[0057] It should be noted that the smaller the deviation of high heart rate from time difference and the smaller the deviation of high heart rate quantity, the more likely the drastic fluctuations in heart rate monitoring data are caused by changes in exercise patterns. At the same time, a larger heart rate change intensity index as a basis leads to a larger exercise change index, which means that drastic changes in the user's heart rate may originate from drastic changes in the user's exercise behavior, resulting in unexpected fluctuations in heart rate data and a larger exercise artifact.

[0058] It should be further noted that the true heart rate signal exhibits a certain rhythm and stability, while motion artifacts usually manifest as high-frequency oscillations (such as shaking during running). Disturbances do not reflect the true fluctuations in physiological state, but are caused by limb movements, environmental vibrations, etc. They are non-target source signals and belong to interference sources, which should be separated from the heart rate monitoring data. Independent component analysis (ICA) can decompose the heart rate monitoring signal into several original signal components, enhancing the ability to remove motion artifacts.

[0059] Preferably, in one embodiment of the present invention, based on the temporal distribution of extreme points in the independent component signals of the heart rate monitoring data and the area difference between peak regions, the mixing degree of each independent component signal is obtained, and then combined with the motion transformation index to obtain the probability of interference sources for each independent component signal. The specific method includes:

[0060] The heart rate monitoring signal, composed of heart rate monitoring data during the monitoring period, is analyzed using ICA (Independent Component Analysis) to obtain several independent component signals. In this embodiment, a total of 5 independent component signals are obtained. The independent component signals are set based on prior knowledge of ICA, which is a well-known technique and will not be described in detail here. For any independent component signal, several extreme points are obtained, including several maximum points and several minimum points. The time interval between any two adjacent extreme points is obtained, and the standard deviation of the time interval between all two adjacent extreme points is used as the extreme value distribution fluctuation coefficient of the independent component signal. The existing algorithm for extreme point extraction is used to obtain the extreme points of the independent component signals, and will not be described in detail here.

[0061] Furthermore, based on the times corresponding to any maximum point and its two adjacent minimum points, a peak region for that maximum point is constructed, i.e., the time interval between the times corresponding to the two minimum points, and the curve of the independent component signal constitutes the peak region. Specifically, if any maximum point is close to the boundary of the monitoring time interval, resulting in no minimum point on one side, then the peak region is constructed based on the portion from the corresponding boundary to the corresponding time of that maximum point. The area of ​​the peak region for that maximum point is obtained by integration. The absolute value of the difference between the areas of the peak regions of any two maximum points is obtained, and the product of the mean of the absolute values ​​of the differences in all obtained areas and the extreme value distribution fluctuation coefficient is used as the mixing degree of the independent component signal.

[0062] It should be noted that the greater the difference in the area of ​​the peak region in the independent component signal during the ICA decomposition process, and the more uneven the distribution of the extreme points, the less the decomposition process of the independent component signal conforms to the decomposition law of ICA. This indicates that the component signal is more chaotic and has poorer independence, that is, the waveform is unstable and there is high-frequency information in the component signal.

[0063] Furthermore, the product of the mixing degree of any independent component signal and the motion transformation index during the monitoring period is used as the probability of interference source for that independent component signal.

[0064] It should be noted that the motion transformation index is used as a weighting reference for component selection. That is, during periods with a high motion transformation index, components containing high-frequency components or those that change frequently are preferentially classified as interference sources, while the remaining components with stable waveforms and good rhythm are regarded as principal components of heart rate. In other words, in the process of obtaining independent component analysis of heart rate monitoring data, the probability that a component signal is an interference source is determined.

[0065] Step S003: Filter several interference source component signals based on the probability of interference sources; decompose each interference source component signal into several IMF component signals through EMD; based on the correlation between the instantaneous frequency performance of each IMF component signal and the order value of the IMF component signal, and combined with the quality difference index of the data points in each IMF component signal, obtain the frequency decoupling degree of each interference source component signal; and combine the interference source probability of the interference source component signal to obtain the preprocessing requirement of each interference source component signal.

[0066] Preferably, in one embodiment of the present invention, the specific method for filtering several interference source component signals based on the probability of interference sources is as follows:

[0067] Specifically, all independent component signals are arranged in ascending order of their interference source probability to obtain a component signal sequence. The difference between the interference source probabilities of adjacent independent component signals in the component signal sequence is obtained (by subtracting the interference source probability of the previous independent component signal from the subsequent independent component signal). The independent component signal to the right of the two independent component signals corresponding to the maximum value among all the differences between the interference source probabilities is taken as the interference source component signal. Other independent component signals to the right of the interference source component signal in the component signal sequence are also taken as interference source component signals.

[0068] It should be noted that the interference source component signals are divided according to the maximum difference to obtain the interference source component signals. At the same time, the screening process is more selective, avoiding the accidental deletion of the main heart rate signal or the retention of artifacts. After completing the independent component analysis (ICA) guided by the motion transformation index, the component signals that may belong to the interference source have been initially identified. However, in dynamic scenarios, some interference components in the heart rate data may have a certain degree of correlation or mixing with the target physiological signal statistically. This results in residual coupling between some components after ICA decomposition, which shows weak independence. Therefore, the interference source component signals need to be further decomposed and filtered.

[0069] Furthermore, to further improve the accuracy of signal purification, a local feature extraction mechanism is introduced for interference source components with insignificant independence, enabling deeper identification and qualitative analysis of potential interference components. By combining nonlinear analysis methods (such as Empirical Mode Decomposition (EMD)) with the retained signal, multimodal decomposition and spectral structure analysis are performed to identify high-frequency disturbances, spurious peaks, and short-time unstable waveforms mixed in with the principal components, thereby achieving deep signal purification.

[0070] Preferably, in one embodiment of the present invention, the interference source component signals are decomposed into several IMF component signals by EMD. Based on the correlation between the instantaneous frequency performance of each IMF component signal and the order value of the IMF component signals, and combined with the quality difference index of the data points in each IMF component signal, the frequency decoupling degree of each interference source component signal is obtained. The specific method includes:

[0071] EMD decomposition is performed on any interference source component signal to obtain several IMF component signals of the interference source component signal. The several IMF component signals have order values ​​and are arranged in ascending order. EMD decomposition is a well-known technique and will not be described in detail in this embodiment.

[0072] Furthermore, a Fourier transform is performed on any IMF component signal of the interference source signal to obtain several instantaneous frequencies of the IMF component signal (several frequencies in the spectrum obtained by the Fourier transform). The mean of all instantaneous frequencies is taken as the frequency performance value of the IMF component signal. The frequency performance values ​​of each IMF component signal in the interference source signal are arranged according to the order of the IMF component signals to obtain the frequency performance sequence of the interference source signal. The order values ​​of each IMF component signal are then arranged to obtain the order value sequence of the interference source signal. The Pearson correlation coefficient between the frequency performance sequence and the order value sequence is obtained, and normalization is performed based on the range of the Pearson correlation coefficient. The normalization result is taken as the decomposition disorder of the interference source signal. The range of the Pearson correlation coefficient is [missing information]. Normalizing its range is equivalent to linearly transforming the range to... In the linear transformation process, the sum of the Pearson correlation coefficients is increased by 1 and then divided by 2. The quotient is the normalized result.

[0073] Furthermore, for each data point in any IMF component signal of any interference source component signal, the quality anomaly index is calculated (the quality anomaly index calculation is an existing method, which will not be described in detail in this embodiment), and the mean of the quality anomaly index of all data points in the IMF component signal is obtained as the periodic stability coefficient of the IMF component signal.

[0074] Furthermore, based on the decomposition disorder of the interference source component signal and the mean periodic stability of all IMF component signals of the interference source component signal, the frequency decoupling degree of the interference source component signal is obtained. The frequency decoupling degree is negatively correlated with the decomposition disorder and positively correlated with the mean periodic stability.

[0075] As an example, the ratio of the mean periodic stability of all IMF component signals of any interference source component signal to the decomposition disorder is taken as the frequency decoupling degree of the interference source component signal. It should be noted that, in order to avoid the denominator being 0 and causing the fraction to be meaningless, a hyperparameter is added to the numerator and denominator respectively during the ratio calculation in this embodiment. The hyperparameter in this embodiment is described as 0.001.

[0076] It should be noted that the closer the correlation coefficient between the frequency performance sequence and the order value sequence is to -1 (i.e., negative correlation), the smaller the decomposition disorder, and the clearer the EMD decomposition of the interference source component signal. At the same time, the higher the periodic stability of the IMF component signal, the more uniform and stable the periodic change, thus obtaining a greater frequency decoupling degree, reflecting the clear decomposition process of the interference source component signal.

[0077] It should be further explained that for the collected heart rate monitoring data, Independent Component Analysis (ICA) is used to identify different components in the heart rate monitoring data. The capability of ICA is "global statistical independence decomposition," which mainly relies on the statistical independence between signal sources for demixing. However, the real heart rate plus motion interference signal may be a nonlinear superposition, which ICA cannot completely decouple and cannot retain the temporal and frequency hierarchical information of the components, making it unsuitable for instantaneous feature analysis. Therefore, ICA is used for preliminary identification, and then Empirical Mode Decomposition (EMD) is used for deeper component identification to achieve accurate identification and hierarchical processing of the main heart rate fluctuations, local interference artifacts, and signal abrupt changes, thereby improving the accuracy and stability of heart rate data processing.

[0078] Preferably, in one embodiment of the present invention, the preprocessing requirement of each interference source component signal is obtained by combining the interference source probability of the interference source component signal, including the following specific method:

[0079] The Euclidean norm of the inversely proportional normalized result of the frequency decoupling degree of any interference source component signal and the normalized result of the interference source probability is used as the preprocessing requirement degree of that interference source component signal.

[0080] It should be noted that this embodiment adopts... The model is used to represent the inverse proportional relationship and for normalization processing. As input to the model, This represents an exponential function with the natural constant as the base. The sigmoid function is used for separate normalization. Implementers can set the inverse proportional function and the normalization function according to the actual situation.

[0081] It should be noted that the Euclidean norm reflects the degree of interaction between the two factors, avoiding one-sided judgment and increasing the robustness of the preprocessing strategy. In terms of signal processing, combining the frequency decoupling degree and the probability of interference sources helps to accurately assess the degree of interference in the heart rate monitoring signal, thereby enabling accurate preprocessing of the heart rate signal.

[0082] Step S004: Smooth the signals of each interference source component based on the preprocessing requirement, and reconstruct the processed heart rate monitoring signal by combining the independent component signals. Perform data analysis on the processed heart rate monitoring signal.

[0083] Specifically, this embodiment uses a Gaussian filtering algorithm to smooth the interference source component signal. The preprocessing requirement of any interference source component signal is used as the filtering intensity value of the filter, and then several processed interference source component signals are obtained by smoothing. Combined with other independent component signals other than the interference source component signal, the reconstructed signal is used as the processed heart rate monitoring signal, that is, the motion artifacts of the heart rate monitoring signal are eliminated.

[0084] Furthermore, based on the processed heart rate monitoring signal, i.e. the "purified" heart rate monitoring signal, feature extraction is used to identify phenomena such as arrhythmia, paroxysmal tachycardia, and stress response, so as to realize the analysis of heart rate data and eliminate the problem of inaccurate heart rate monitoring data caused by motion artifacts in wearable heart rate monitoring devices.

[0085] This concludes the embodiment.

[0086] Another embodiment of the present invention provides a heart rate data analysis system, which includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the above-described method steps S001 to S004.

[0087] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for analyzing heart rate data, characterized in that, The method includes the following steps: Heart rate monitoring signals are collected using wearable heart rate monitoring devices to obtain heart rate monitoring data for the monitoring period and several moments within it, and motion acceleration data at each moment are recorded. The dispersion and differences in heart rate monitoring data during the monitoring period are analyzed to quantify the intensity index of heart rate changes during the monitoring period. Based on the dispersion of motion acceleration data during the monitoring period and the time difference of its dispersion relationship with heart rate monitoring data, combined with the intensity index of heart rate changes, the motion transformation index of the monitoring period is obtained. Based on the temporal distribution of extreme points in each independent component signal of heart rate monitoring data and the area difference between peak regions, the mixing degree of each independent component signal is obtained, and then combined with the motion transformation index, the probability of interference sources of each independent component signal is obtained. Several interference source component signals are selected based on the probability of interference sources; several IMF component signals are obtained by EMD decomposition of each interference source component signal; based on the correlation between the instantaneous frequency performance of each IMF component signal and the order value of the IMF component signal, and combined with the quality difference index of the data points in each IMF component signal, the frequency decoupling degree of each interference source component signal is obtained; combined with the probability of interference sources of the interference source component signals, the preprocessing requirements of each interference source component signal are obtained. The signals of each interference source are smoothed based on the preprocessing requirement, and the processed heart rate monitoring signal is obtained by reconstructing the signals from the independent component signals. The method for obtaining the heart rate variability index during the monitoring period is as follows: All heart rate monitoring data during the monitoring period are arranged in ascending order to obtain an ascending heart rate sequence. The difference between any two adjacent heart rate monitoring data in the ascending heart rate sequence is obtained. The two adjacent heart rate monitoring data corresponding to the maximum difference between all two adjacent heart rate monitoring data are used as segmented heart rate data. The left segmented heart rate data in the two segmented heart rate data and all heart rate data in the left side of the ascending heart rate sequence constitute the low heart rate data set of the monitoring period. The right segmented heart rate data in the two segmented heart rate data and all heart rate data in the right side of the ascending heart rate sequence constitute the high heart rate data set of the monitoring period. The difference between the mean of the heart rate monitoring data in the high heart rate dataset and the mean of the heart rate monitoring data in the low heart rate dataset is used as the heart rate dispersion coefficient for the monitoring period. The absolute value of the difference between the heart rate monitoring data at any two adjacent moments in the monitoring period is obtained, and the product of the mean of the absolute values ​​of the differences between the heart rate monitoring data at all two adjacent moments in the monitoring period and the heart rate dispersion coefficient is used as the heart rate variability index for the monitoring period. The specific method for obtaining the motion change index during the monitoring period is as follows: All motion acceleration data during the monitoring period are arranged in ascending order to obtain an acceleration ascending sequence. Based on the difference between adjacent motion acceleration data in the acceleration ascending sequence, the low acceleration data set and the high acceleration data set for the monitoring period are obtained. The absolute value of the difference between the minimum value of all motion acceleration data at the corresponding time in the high acceleration dataset and the minimum value of all heart rate monitoring data at the corresponding time in the high heart rate dataset is recorded as the high heart rate deviation time difference during the monitoring period; the absolute value of the difference between the number of motion acceleration data in the high acceleration dataset and the number of heart rate monitoring data in the high heart rate dataset is recorded as the high heart rate quantity deviation during the monitoring period. Based on the intensity index of heart rate changes during the monitoring period, as well as the deviation of high heart rate from time and the deviation of high heart rate quantity, the exercise transformation index of the monitoring period is obtained; the exercise transformation index is positively correlated with the intensity index of heart rate changes, and negatively correlated with the deviation of high heart rate from time and the deviation of high heart rate quantity.

2. The heart rate data analysis method according to claim 1, characterized in that, The specific methods for obtaining the degree of mixing of each independent component signal are as follows: The heart rate monitoring signal, composed of heart rate monitoring data during the monitoring period, was analyzed by ICA independent component analysis to obtain several independent component signals. For any independent component signal, obtain several extreme points, including several maximum points and several minimum points, obtain the time interval between any two adjacent extreme points, and use the standard deviation of the time interval between all two adjacent extreme points as the extreme value distribution fluctuation coefficient of the independent component signal. Based on the time corresponding to any maximum point and its two adjacent minimum points, the peak region of the maximum point is formed, and the area of ​​the peak region of the maximum point is obtained by integration. Obtain the absolute value of the difference between the areas of the peak regions of any two maximum points, and multiply the mean of the absolute values ​​of the differences in all obtained areas by the extreme value distribution fluctuation coefficient as the degree of mixing of the independent component signal.

3. The heart rate data analysis method according to claim 1, characterized in that, The specific methods for obtaining the probability of interference sources for each independent component signal are as follows: The product of the mixing degree of any independent component signal and the motion transformation index of the monitoring period is taken as the probability of interference source for that independent component signal.

4. The heart rate data analysis method according to claim 1, characterized in that, The specific method for filtering several interference source component signals based on the probability of interference sources is as follows: Arrange all independent component signals in ascending order of their interference source probability to obtain a component signal sequence. Obtain the difference between the interference source probabilities of adjacent independent component signals in the component signal sequence. Take the rightmost independent component signal of the two independent component signals corresponding to the maximum value of the difference between all the interference source probabilities as the interference source component signal. Other independent component signals to the right of the interference source component signal in the component signal sequence are taken as interference source component signals.

5. The heart rate data analysis method according to claim 1, characterized in that, The specific method for obtaining the frequency decoupling degree of each interference source component signal is as follows: Based on the instantaneous frequency performance of each IMF component signal and the order value of the IMF component signals, the decomposition disorder of each interference source component signal is obtained. For any IMF component signal of any interference source component signal, calculate the quality anomaly index for each data point, obtain the mean of the quality anomaly index of all data points in the IMF component signal, and use it as the periodic stability coefficient of the IMF component signal. Based on the decomposition disorder of the interference source component signal and the mean periodic stability of all IMF component signals of the interference source component signal, the frequency decoupling degree of the interference source component signal is obtained. The frequency decoupling degree is negatively correlated with the decomposition disorder and positively correlated with the mean periodic stability.

6. The heart rate data analysis method according to claim 5, characterized in that, The specific method for obtaining the decomposition disorder of each interference source component signal is as follows: Perform a Fourier transform on any IMF component signal of any interference source component signal to obtain several instantaneous frequencies of the IMF component signal, and take the average of all instantaneous frequencies as the frequency performance value of the IMF component signal. The frequency performance values ​​of each IMF component signal in the interference source component signal are arranged according to the order of the IMF component signals to obtain the frequency performance sequence of the interference source component signal, and the order value sequence of each IMF component signal is obtained by arranging the order values ​​of the interference source component signal. The Pearson correlation coefficient between the frequency performance sequence and the order value sequence is obtained, and normalized based on the range of the Pearson correlation coefficient. The normalization result is used as the decomposition disorder of the interference source component signal.

7. The heart rate data analysis method according to claim 1, characterized in that, The specific method for obtaining the preprocessing requirements of each interference source component signal is as follows: The Euclidean norm of the inversely proportional normalized result of the frequency decoupling degree of any interference source component signal and the normalized result of the interference source probability is used as the preprocessing requirement degree of that interference source component signal.

8. A heart rate data analysis system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the heart rate data analysis method as described in any one of claims 1-7.

Images