User data analysis method based on wearable devices

Abstract

The present invention relates to the technical field of auxiliary training equipment and proposes a user data analysis method based on a wearable device, comprising: collecting electrocardiogram (ECG) data of an exercise user during training to obtain an ECG sequence, an ECG waveform sequence, and a quasi-period; determining a frequency curve, obtaining an extreme value pair of the frequency curve, and then obtaining a frequency extreme value disorder coefficient; obtaining a fundamental frequency shift coefficient of the ECG data, and then obtaining an intra-cluster fundamental drift deviation coefficient, obtaining a first correlation coefficient, obtaining an ECG waveform quasi-period deviation coefficient, and then obtaining an ECG waveform disturbance distortion index; obtaining a first particle number of the ECG data, denoising the ECG sequence based on the first particle number, obtaining a denoised ECG sequence, and implementing intelligent analysis of exercise activity data based on the wearable device based on the denoised ECG sequence. The present invention aims to solve the problem of being unable to dynamically adjust the denoising of training data based on the exercise state of the exercise user, resulting in poor denoising effect.

Description

Technical Field

[0001] The present invention relates to the technical field of auxiliary training equipment, and in particular to a user data analysis method based on wearable devices. Background Art

[0002] Wearable devices can provide real-time data and feedback to athletes during exercise. By analyzing the data collected by wearable devices, we can better understand the user's training results and physical condition, helping to develop more scientific training plans and adjust training intensity. Wearable devices can also provide real-time feedback to athletes, helping them adjust the intensity and pace of their training, thereby improving training results and reducing the risk of injury. Common wearable devices include fitness tracking bracelets, ECG monitoring chest straps, smart insoles, smart gloves, and sports wristbands.

[0003] Typically, existing wearable devices can be categorized into various types based on their working principle, such as electromagnetic and optical. Due to the flexible structure of human joints and the complex nature of movement, miniaturized sensors with expansion and compression characteristics are often used to better adapt to changes in human motion. However, wearable devices can be sensitive to temperature fluctuations and electronic interference, resulting in reduced monitoring sensitivity. Therefore, analysis of the data collected by wearable devices is necessary to meet high-precision monitoring requirements in a variety of fast-paced and complex motion scenarios. Summary of the Invention

[0004] The present invention provides a user data analysis method based on a wearable device to solve the problem that the denoising effect is poor due to the inability to dynamically adjust the training data according to the exercise state of the user. The technical solution adopted is as follows:

[0005] An embodiment of the present invention provides a user data analysis method based on a wearable device, the method comprising the following steps:

[0006] Collect ECG data of users during training, fill missing values ​​and cluster the ECG data, obtain ECG sequences and ECG waveform sequences, and obtain the quasi-periodicity of the ECG waveform sequences;

[0007] According to the ECG waveform sequence, a time-frequency signal of the ECG waveform sequence is obtained; according to the time-frequency signal, a frequency curve of the acquisition time to be analyzed is determined; an extreme value pair of the frequency curve is obtained; according to the extreme value pair of the frequency curve, an extreme value fluctuation coefficient of the ECG data is obtained; according to the extreme value fluctuation coefficient of the ECG data and the frequency curve, a frequency extreme value disorder coefficient of the ECG data waveform moment corresponding to the frequency curve is obtained;

[0008] Obtaining the fundamental frequency shift coefficient of the ECG data based on the frequency curve of the ECG data and the acquisition time of the ECG data contained in the ECG waveform sequence, thereby obtaining the intra-cluster fundamental drift deviation coefficient of the ECG waveform sequence, obtaining the first correlation coefficient of two adjacent ECG waveform sequences, obtaining the ECG waveform quasi-periodic deviation coefficient of the ECG waveform sequence based on the quasi-periodicity of the adjacent ECG waveform sequences, and obtaining the ECG waveform disturbance distortion index of the ECG waveform sequence in combination with the first correlation coefficients of the adjacent ECG waveform sequences;

[0009] The first particle number of the ECG data is obtained based on the frequency extreme value disorder coefficient of the ECG data waveform moment and the ECG waveform disturbance distortion index of the ECG waveform sequence. Based on the first particle number, the ECG sequence is denoised to obtain a denoised ECG sequence. Based on the denoised ECG sequence, intelligent analysis of sports activity data based on wearable devices is realized.

[0010] Furthermore, the specific method of obtaining the extreme value pair of the frequency curve includes:

[0011] Get the maximum and minimum values ​​of the frequency curve. Starting from the first maximum value of the frequency curve, each maximum value and the adjacent minimum value whose acquisition time is later than the maximum value are regarded as a set of extreme value pairs. Discard the maximum and minimum values ​​that do not form an extreme value pair.

[0012] Furthermore, the specific method for obtaining the extreme value fluctuation coefficient of the electrocardiogram data is as follows:

[0013] The difference between the maximum value and the minimum value of the same extreme value pair contained in the frequency curve of the acquisition time of the ECG data is taken as the extreme value difference of the extreme value pair;

[0014] The frequency interval between the maximum and minimum values ​​of the same extreme value pair contained in the frequency curve of the ECG data acquisition time is taken as the frequency difference of the extreme value pair;

[0015] The ratio of the extreme value difference to the frequency difference of the extreme value pair is taken as the first ratio of the extreme value pair;

[0016] The sum of the first ratios of all the extreme value pairs included in the frequency curve of the acquisition time of the electrocardiogram data is used as the extreme value fluctuation coefficient of the electrocardiogram data.

[0017] Furthermore, the frequency extreme value disorder coefficient of the waveform moment of the electrocardiogram data corresponding to the frequency curve is obtained by:

[0018] The ratio of the frequency bandwidth corresponding to the time when the maximum ECG energy intensity on the frequency curve of the ECG data acquisition time decays by half to the maximum value of all ECG energy intensities on the frequency curve is recorded as the attenuation value, and the product of the attenuation value and the extreme value fluctuation coefficient of the ECG data is recorded as the frequency extreme value disorder coefficient of the ECG data waveform moment.

[0019] Furthermore, the fundamental frequency shift coefficient of the electrocardiogram data is obtained by:

[0020] Record the ratio of the maximum ECG energy intensity on the frequency curve of the kth and k-1th ECG data acquisition times as a first ratio, input the sum of the first ratio and the number 1 into a logarithmic function with the natural constant e as the base, and record the result as the first function value, and record the first function value as the fundamental wave frequency shift coefficient;

[0021] Where k is a positive integer.

[0022] Furthermore, the specific method for obtaining the intra-cluster fundamental wave drift deviation coefficient of the ECG waveform sequence is as follows:

[0023] The DTW distance between the acquisition time frequency curve of the ECG data and the previous adjacent ECG data is recorded as the first distance of the ECG data;

[0024] The product of the fundamental frequency shift coefficient of the electrocardiogram data and the first distance of the electrocardiogram data is recorded as the first product of the electrocardiogram data;

[0025] The sum of the first products of all the electrocardiographic data included in the electrocardiographic waveform sequence is recorded as the intra-cluster fundamental wave drift deviation coefficient of the electrocardiographic waveform sequence.

[0026] Furthermore, the specific method of obtaining the first correlation coefficient of two adjacent ECG waveform sequences includes:

[0027] The minimum value of the number of ECG data contained in two adjacent ECG waveform sequences is recorded as the length threshold of the two ECG waveform sequences, the two adjacent ECG waveform sequences are truncated according to the length threshold, and the Pearson correlation coefficient of the two truncated sequences is used as the first correlation coefficient of the two adjacent ECG waveform sequences.

[0028] Furthermore, the method of obtaining the ECG waveform quasi-periodic deviation coefficient of the ECG waveform sequence based on the quasi-periodicity of the adjacent ECG waveform sequences and obtaining the ECG waveform disturbance distortion index of the ECG waveform sequence in combination with the first correlation coefficient of the adjacent ECG waveform sequences includes the following specific methods:

[0029]

[0030] Where, Represents the ECG waveform disturbance distortion index of the mth ECG waveform sequence; θ m represents the quasi-periodic deviation coefficient of the mth ECG waveform sequence; log2() represents the logarithmic function with base 2; T m represents the quasi-period of the mth ECG waveform sequence; T m-1represents the quasi-period of the m-1th ECG waveform sequence; T m+1 represents the quasi-period of the m+1th ECG waveform sequence; b1 represents the first adjustment parameter; b2 represents the second adjustment parameter; δ m represents the drift deviation coefficient of the fundamental wave in the cluster of the mth ECG waveform sequence; exp() represents the exponential function with a natural constant as the base; min(,) represents the minimum value of the values ​​separated by commas in the brackets; ρ(A m ,A m-1 ) represents the first correlation coefficient between the mth ECG waveform sequence and the m-1th ECG waveform sequence; ρ(A m ,A m+1 ) represents the first correlation coefficient between the mth ECG waveform sequence and the m+1th ECG waveform sequence.

[0031] Furthermore, the specific method for obtaining the first particle number of the electrocardiogram data is as follows:

[0032] The product of a power with a natural constant as the base and an electrocardiogram waveform disturbance distortion index of the electrocardiogram waveform sequence as the exponent and the frequency extreme value disorder coefficient of the waveform moment of the electrocardiogram data in the electrocardiogram waveform sequence is recorded as the second product, and the maximum value of the rounded value of the second product and the particle number adjustment factor is used as the first particle number of the electrocardiogram data.

[0033] Furthermore, the method of denoising the ECG sequence according to the first particle number to obtain the denoised ECG sequence and implementing intelligent analysis of exercise activity data based on the wearable device according to the denoised ECG sequence includes the following specific methods:

[0034] The first particle number of the ECG data is used as the adaptive particle number of the ECG data, and the ECG sequence is used as the input of the particle filter algorithm to obtain the denoised ECG sequence;

[0035] Professional medical staff perform ECG data analysis on the denoised ECG sequence, obtain ECG data analysis results, and realize intelligent analysis of sports activity data based on wearable devices.

[0036] The beneficial effects of the present invention are:

[0037] The present invention divides the collected ECG data of the exercise user during training into ECG waveform sequences for analysis based on the periodic characteristics of the ECG waveform, analyzes each complete ECG waveform separately, and improves the accuracy of the analysis. According to the characteristics that the fundamental signal of the ECG data carries greater energy and the number of extreme value pairs contained in the frequency curve is small and the intervals between the extreme value pairs are far when the ECG data is not interfered with or is slightly interfered with, the frequency extreme value disorder coefficient of the ECG data waveform moment is obtained according to the frequency extreme value; according to the characteristics that when the exercise user is more intense, the ECG data is easily subject to greater myoelectric interference, resulting in the occurrence of fundamental wave drift, the fundamental wave drift deviation coefficient within the cluster of the ECG waveform sequence is obtained, and then the ECG waveform sequence is obtained. The waveform disturbance distortion index reflects the degree of ECG data distortion in the ECG waveform sequence. The first particle number of the ECG data is obtained based on the frequency extreme value disorder coefficient of the ECG data waveform moment and the ECG waveform disturbance distortion index of the ECG waveform sequence. The first particle number of the ECG data is used as the adaptive particle number of the ECG data to denoise the ECG sequence, solving the problem of poor denoising effect caused by the inability to dynamically adjust the training data according to the exercise state of the exercise user. The number of particles in the denoising process can be adaptively adjusted according to the degree to which the ECG data is affected by the exercise user's movement, thereby reducing the degree of interference to the ECG data, making the ECG data more accurate, and accelerating the convergence speed of the particle filter algorithm and reducing errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0039] Figure 1 A flowchart of a user data analysis method based on a wearable device provided by one embodiment of the present invention;

[0040] Figure 2 Obtain the flow chart for the frequency extreme value disorder coefficient. DETAILED DESCRIPTION

[0041] 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.

[0042] See also Figure 1 , which shows a flow chart of a user data analysis method based on a wearable device provided by an embodiment of the present invention, the method comprising the following steps:

[0043] Step S1: Collect the ECG data of the exercise user during training, perform missing value filling and clustering on the ECG data, obtain the ECG sequence and ECG waveform sequence, and obtain the quasi-period of the ECG waveform sequence.

[0044] In order to perform intelligent analysis on the training data of sports users, sports users who need intelligent analysis are required to wear an ECG monitoring chest strap, so that the ECG monitoring chest strap is tightly placed below the chest of the sports user and slightly to the left of the heart position, so that the ECG monitor is in the position closest to the heart, thereby improving the accuracy of ECG data collected by the ECG monitoring chest strap.

[0045] The wearable device ECG monitoring chest strap is used to obtain the ECG data of the athlete during training in real time. In this embodiment, the sampling interval of the ECG data is set to 20ms, and the ECG data of the athlete is continuously obtained during training.

[0046] During ECG data collection, the user's movements can vary significantly, potentially leading to missing values. To minimize the impact of missing values ​​on training data analysis, the ECG data is arranged chronologically to create an ECG sequence. Lagrangian interpolation is then used to fill missing values ​​in the sequence. This method is well known and will not be further described.

[0047] To facilitate analysis of ECG data in an ECG sequence, the DTC time series clustering model is used as input. The model clusters the ECG data based on the similarity of its fluctuations, outputting multiple time clusters. DTC is an unsupervised algorithm that automatically determines the number of clusters based on the ECG data, eliminating the need for manual clustering. The DTC time series clustering model is well-known and will not be further described.

[0048] ECG data has a certain periodicity. Therefore, each acquired time cluster corresponds to a complete cycle of ECG waveform. All ECG data contained in the time cluster are arranged in the order of acquisition time to obtain an ECG waveform sequence.

[0049] The time difference between the earliest acquisition time and the latest acquisition time corresponding to all ECG data in the ECG waveform sequence is taken as the quasi-period of the ECG waveform sequence.

[0050] Ideally, the ECG data of an athlete presents a pure and periodically stable waveform. However, during training, the ECG data collected by the chest strap is subject to myoelectric interference, resulting in differences in different time clusters that require further analysis.

[0051] At this point, the ECG waveform sequence and the quasi-period of the ECG waveform sequence are obtained.

[0052] Step S2: Determine the frequency curve of the acquisition time to be analyzed based on the ECG waveform sequence, obtain the extreme value pair of the frequency curve, obtain the extreme value fluctuation coefficient of the ECG data, and obtain the frequency extreme value disorder coefficient based on the extreme value fluctuation coefficient of the ECG data and the frequency curve.

[0053] All ECG data contained in the ECG waveform sequence are used as the input of the least squares method, and the ECG data is nonlinearly fitted, and the output is the ECG waveform obtained by fitting. In order to better analyze the frequency component of the ECG waveform, the ECG waveform is used as the input of the continuous wavelet transform, and the time-frequency signal of the ECG waveform sequence is output. In this embodiment, Morlet wavelet is selected as the wavelet basis function. The time-frequency signal of the ECG waveform sequence is a three-dimensional surface. The horizontal coordinate of the three-dimensional surface represents the acquisition time of the ECG data contained in the ECG waveform sequence, the vertical coordinate represents the frequency, and the vertical coordinate represents the ECG energy intensity. In this embodiment, the frequency range is set to 0Hz-1000Hz. Among them, the use of the least squares method for nonlinear fitting and continuous wavelet transform is a well-known technology and will not be repeated here.

[0054] The acquisition time of each ECG data contained in the ECG waveform sequence is used as the acquisition time to be analyzed. The moment corresponding to the acquisition time to be analyzed and the curve corresponding to the frequency and ECG energy intensity are obtained according to the time-frequency signal of the ECG waveform sequence. The curve corresponding to the frequency and ECG energy intensity is recorded as the frequency curve of the acquisition time to be analyzed.

[0055] The extreme points of the frequency curve are obtained through extreme point detection. Due to the periodicity of the electrocardiographic waves, the maximum and minimum values ​​in the extreme points appear alternately. Each maximum value and the adjacent minimum value whose acquisition time is later than the maximum value are regarded as a set of extreme value pairs, and the extreme points without corresponding extreme value pairs are discarded.

[0056] According to the extreme value pairs of the frequency curve, the extreme value fluctuation coefficient of the ECG data is obtained. According to the extreme value fluctuation coefficient of the ECG data and the frequency curve, the frequency extreme value disorder coefficient of the ECG data waveform moment corresponding to the frequency curve is obtained.

[0057]

[0058] Where, The frequency extreme value disorder coefficient of the waveform moment of the kth ECG data contained in the mth ECG waveform sequence; The maximum value of all ECG energy intensities on the frequency curve representing the acquisition time of the kth ECG data; The frequency bandwidth corresponding to the time when the maximum ECG energy intensity on the frequency curve of the k-th ECG data acquisition time is attenuated by half; shows the extreme value fluctuation coefficient of the kth ECG data contained in the mth ECG waveform sequence; The number of extreme value pairs contained in the frequency curve representing the acquisition time of the kth ECG data contained in the mth ECG waveform sequence; The i-th maximum value contained in the frequency curve representing the acquisition time of the k-th ECG data contained in the m-th ECG waveform sequence; The i-th minimum value contained in the frequency curve representing the acquisition time of the k-th ECG data contained in the m-th ECG waveform sequence; The frequency interval of the i-th group of extreme value pairs contained in the frequency curve representing the acquisition time of the k-th ECG data contained in the m-th ECG waveform sequence.

[0059] Under normal circumstances, ECG data contains fundamental wave signals, which are the frequency values ​​when ECG energy intensity is the highest. Energy is mainly concentrated in the fundamental wave signals. Therefore, when the ECG data collected during the analysis period is not disturbed or is slightly disturbed, the fundamental wave carries a large amount of energy, i.e. The frequency curve contains fewer extreme value pairs, and the intervals between extreme value pairs are larger. This results in a smaller extreme value fluctuation coefficient and, in turn, a smaller frequency extreme value disorder coefficient at each waveform moment of the ECG data. The frequency extreme value disorder coefficient at each waveform moment of the ECG data can reflect the interference level of the ECG data.

[0060] At this point, the frequency extreme value disorder coefficient of the ECG data waveform moment is obtained. The frequency extreme value disorder coefficient acquisition flow chart is as follows: Figure 2 shown.

[0061] Step S3: Obtain the fundamental frequency shift coefficient of the ECG data, and then obtain the intra-cluster fundamental drift deviation coefficient of the ECG waveform sequence, obtain the first correlation coefficient of two adjacent ECG waveform sequences, obtain the ECG waveform quasi-periodic deviation coefficient of the ECG waveform sequence based on the quasi-periodicity, and obtain the ECG waveform disturbance distortion index of the ECG waveform sequence in combination with the first correlation coefficient.

[0062] At different collection times of ECG data contained in an ECG waveform sequence, the degree of interference of the ECG data is different. In order to measure the degree of interference of the ECG data contained in each ECG waveform sequence, the ECG waveform sequence is further analyzed.

[0063] According to the frequency curve of the ECG data contained in the ECG waveform sequence and the acquisition time of the ECG data, the fundamental frequency shift coefficient of the ECG data is obtained, and then the intra-cluster fundamental drift deviation coefficient of the ECG waveform sequence is obtained.

[0064]

[0065] Where, δ m N represents the drift deviation coefficient of the mth ECG waveform sequence within the cluster; m represents the number of ECG data in the mth ECG waveform sequence; γ k A represents the fundamental frequency shift coefficient of the kth ECG data; k,f A represents the frequency curve of the acquisition time of the kth ECG data; k-1,f The frequency curve representing the acquisition time of the k-1th ECG data; Dist(A k,f ,A k-1,f ) represents the distance between the frequency curves of the acquisition time of the kth and k-1th ECG data; log2() represents a logarithmic function with base 2; The maximum value of all ECG energy intensities on the frequency curve representing the acquisition time of the kth ECG data; Indicates the maximum value of all ECG energy intensities on the frequency curve at the acquisition time of the k-1th ECG data.

[0066] When a user exercises vigorously, the ECG data is susceptible to greater myoelectric interference, resulting in fundamental wave drift. At the same time, different degrees of exercise will lead to occasional myoelectric interference, that is, the muscle state is different when the exercise state is different. Therefore, the ECG energy intensity corresponding to the fundamental wave signal at adjacent sampling moments is quite different. At this time, the fundamental wave frequency shift coefficient of the ECG data is larger, and the fundamental wave drift deviation coefficient within the cluster of the ECG waveform sequence is larger.

[0067] Two adjacent ECG waveform sequences are truncated according to the minimum value of the number of ECG data contained in the two ECG waveform sequences, and the Pearson correlation coefficient of the two truncated sequences is used as the first correlation coefficient of the two adjacent ECG waveform sequences.

[0068] A complete ECG wave consists of a P wave, QRS complex, and T wave. Different wave shapes represent different working states of the heart. During training, the degree of interference with ECG data varies with the user's exercise state. Therefore, we analyze adjacent ECG waveform sequences and, based on their quasi-periodicity, obtain the ECG waveform quasi-periodicity deviation coefficient. Combined with the first correlation coefficient of adjacent ECG waveform sequences, we obtain the ECG waveform interference distortion index.

[0069]

[0070] Where, Represents the ECG waveform disturbance distortion index of the mth ECG waveform sequence; θ m represents the quasi-periodic deviation coefficient of the mth ECG waveform sequence; log2() represents the logarithmic function with base 2; T m represents the quasi-period of the mth ECG waveform sequence; T m-1 represents the quasi-period of the m-1th ECG waveform sequence; T m+1 represents the quasi-period of the m+1th ECG waveform sequence; b1 represents the first adjustment parameter, and the value of this embodiment is 1; b2 represents the second adjustment parameter, and the value of this embodiment is 2; δ m represents the drift deviation coefficient of the fundamental wave in the cluster of the mth ECG waveform sequence; exp() represents the exponential function with a natural constant as the base; min(,) represents the minimum value of the values ​​separated by commas in the brackets; ρ(A m ,A m-1 ) represents the first correlation coefficient between the mth ECG waveform sequence and the m-1th ECG waveform sequence; ρ(A m ,A m+1 ) represents the first correlation coefficient between the mth ECG waveform sequence and the m+1th ECG waveform sequence.

[0071] The ECG waveform distortion index (EDDI) of an ECG waveform sequence reflects the degree of ECG data distortion in the sequence. The more severe the ECG data distortion, the smaller the quasi-period of the ECG waveform sequence, and the greater the difference in quasi-periods between adjacent ECG waveform sequences, the smaller the ECG waveform quasi-period deviation coefficient of the ECG waveform sequence. At the same time, the smaller the first correlation coefficient of adjacent ECG waveform sequences, the greater the ECG waveform distortion index of the ECG waveform sequence.

[0072] At this point, the ECG waveform disturbance distortion index of the ECG waveform sequence is obtained.

[0073] Step S4: obtaining the first particle number of the ECG data based on the frequency extreme value disorder coefficient at the waveform moment of the ECG data and the ECG waveform disturbance distortion index of the ECG waveform sequence; denoising the ECG sequence based on the first particle number to obtain a denoised ECG sequence; and implementing intelligent analysis of sports activity data based on the wearable device based on the denoised ECG sequence.

[0074] In the particle filter algorithm, a fixed number of particles must be set for each sampling point, the particles with the fixed number of particles must be initialized, and then iterative optimization must be performed. The choice of particle number has a significant impact on the effectiveness of the particle filter. When the number of particles is set too large, the computational complexity increases, resulting in a slower iteration speed at each time step. It also increases the risk of overfitting, making the filtered data overly sensitive and the filtering results unstable. When the number of particles is set too small, there is a risk of particle degradation, that is, some particles are eliminated during the iterative optimization process. Since traditional particle filter algorithms use a fixed number of particles, but sports user training is a dynamic process, the ECG data collected under different training states is subject to different levels of interference. Therefore, it is necessary to dynamically adjust the number of particles set at each sampling point during the iterative process based on the degree of interference with the collected ECG data.

[0075] The first particle number of the ECG data is obtained according to the frequency extreme value disorder coefficient of the ECG data waveform moment and the ECG waveform disturbance distortion index of the ECG waveform sequence.

[0076]

[0077] Where N m,k The number of first particles of the kth ECG data contained in the mth ECG waveform sequence; The frequency extreme value disorder coefficient of the waveform moment of the kth ECG data contained in the mth ECG waveform sequence; represents the ECG waveform disturbance distortion index of the mth ECG waveform sequence; exp() represents an exponential function with a natural constant as the base; ε represents a particle number adjustment factor, which is 50 in this embodiment; int() represents a rounding-up function, which is used to round up the value in the brackets; max(,) represents the maximum value of the comma-separated values ​​in the brackets.

[0078] When the degree of interference to the ECG data is greater, the ECG waveform distortion index of the ECG waveform sequence in which the ECG data is located is greater, the frequency extreme value disorder coefficient of the ECG data waveform moment is greater, and at this time, the number of first particles of the ECG data is greater.

[0079] Assigning a larger number of first particles to ECG data brings the estimated value closer to the true value, reducing the degree of interference with the ECG data and making it more accurate. Assigning a smaller number of first particles to ECG data reduces the degree of interference with the ECG data, which can accelerate the convergence of the particle filter algorithm and reduce errors while ensuring the accuracy of the ECG data.

[0080] The first particle number of the ECG data is used as the adaptive particle number of the ECG data, and the ECG sequence is used as the input of the particle filter algorithm to obtain a denoised ECG sequence. Using the particle filter algorithm to denoise data is a well-known technology and will not be described in detail.

[0081] Professional medical staff analyze the denoised ECG data, including P-wave, QRS complex, and T-wave characteristics, to obtain ECG data analysis results. This analysis helps understand the user's heart's physiological activity and myocardial blood supply. Cardiac activity includes atrial contraction, ventricular contraction, and ventricular repolarization, allowing real-time monitoring of the user's training intensity tolerance and potential for improvement.

[0082] At this point, intelligent analysis of sports activity data based on wearable devices is achieved.

[0083] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A user data analysis method based on wearable devices, characterized in that: The method comprises the following steps: Collect ECG data of users during training, fill missing values ​​and cluster the ECG data, obtain ECG sequences and ECG waveform sequences, and obtain the quasi-periodicity of the ECG waveform sequences; According to the ECG waveform sequence, a time-frequency signal of the ECG waveform sequence is obtained; according to the time-frequency signal, a frequency curve of the acquisition time to be analyzed is determined; an extreme value pair of the frequency curve is obtained; according to the extreme value pair of the frequency curve, an extreme value fluctuation coefficient of the ECG data is obtained; according to the extreme value fluctuation coefficient of the ECG data and the frequency curve, a frequency extreme value disorder coefficient of the ECG data waveform moment corresponding to the frequency curve is obtained; Obtaining the fundamental frequency shift coefficient of the ECG data based on the frequency curve of the ECG data and the acquisition time of the ECG data contained in the ECG waveform sequence, thereby obtaining the intra-cluster fundamental drift deviation coefficient of the ECG waveform sequence, obtaining the first correlation coefficient of two adjacent ECG waveform sequences, obtaining the ECG waveform quasi-periodic deviation coefficient of the ECG waveform sequence based on the quasi-periodicity of the adjacent ECG waveform sequences, and obtaining the ECG waveform disturbance distortion index of the ECG waveform sequence in combination with the first correlation coefficients of the adjacent ECG waveform sequences; The first particle number of the ECG data is obtained based on the frequency extreme value disorder coefficient of the ECG data waveform moment and the ECG waveform disturbance distortion index of the ECG waveform sequence. Based on the first particle number, the ECG sequence is denoised to obtain a denoised ECG sequence. The denoised ECG sequence is used to implement intelligent analysis of exercise activity data based on wearable devices. The specific method for obtaining the frequency extreme value disorder coefficient of the waveform moment of the electrocardiogram data corresponding to the frequency curve is as follows: The ratio of the frequency bandwidth corresponding to the time when the maximum ECG energy intensity on the frequency curve at the time of ECG data acquisition is attenuated by half to the maximum value of all ECG energy intensities on the frequency curve is recorded as the attenuation value, and the product of the attenuation value and the extreme value fluctuation coefficient of the ECG data is recorded as the frequency extreme value disorder coefficient of the ECG data waveform at the time of acquisition; The method of obtaining the ECG waveform quasi-periodic deviation coefficient of the ECG waveform sequence based on the quasi-periodicity of the adjacent ECG waveform sequence and obtaining the ECG waveform disturbance distortion index of the ECG waveform sequence in combination with the first correlation coefficient of the adjacent ECG waveform sequence includes the following specific methods: Where, Represents the ECG waveform disturbance distortion index of the mth ECG waveform sequence; represents the quasi-periodic deviation coefficient of the mth ECG waveform sequence; represents the logarithmic function with base 2; represents the quasi-period of the mth ECG waveform sequence; represents the quasi-period of the m-1th ECG waveform sequence; represents the quasi-period of the m+1th ECG waveform sequence; represents the first adjustment parameter; represents the second adjustment parameter; Represents the intra-cluster fundamental wave drift deviation coefficient of the mth ECG waveform sequence; represents an exponential function with a natural constant as its base; Indicates taking the minimum value of the values ​​separated by commas in the brackets; represents the first correlation coefficient between the mth ECG waveform sequence and the m-1th ECG waveform sequence; represents the first correlation coefficient between the mth ECG waveform sequence and the m+1th ECG waveform sequence; The specific method for obtaining the first particle number of the electrocardiogram data is as follows: The product of a power with a natural constant as the base and an ECG waveform disturbance distortion index of the ECG waveform sequence as the exponent and a frequency extreme value disorder coefficient of the waveform moment of the ECG data in the ECG waveform sequence is recorded as a second product, and the maximum value of the rounded value of the second product and the particle number adjustment factor is taken as the first particle number of the ECG data; The method of denoising the ECG sequence according to the first particle number to obtain the denoised ECG sequence and implementing intelligent analysis of the exercise activity data based on the wearable device according to the denoised ECG sequence includes the following specific methods: The first particle number of the ECG data is used as the adaptive particle number of the ECG data, and the ECG sequence is used as the input of the particle filter algorithm to obtain the denoised ECG sequence; Professional medical staff perform ECG data analysis on the denoised ECG sequence, obtain ECG data analysis results, and realize intelligent analysis of sports activity data based on wearable devices.

2. The user data analysis method based on wearable devices according to claim 1, characterized in that: The specific method of obtaining the extreme value pair of the frequency curve includes: Get the maximum and minimum values ​​of the frequency curve. Starting from the first maximum value of the frequency curve, each maximum value and the adjacent minimum value whose acquisition time is later than the maximum value are regarded as a set of extreme value pairs. Discard the maximum and minimum values ​​that do not form an extreme value pair.

3. The user data analysis method based on wearable devices according to claim 1, characterized in that: The specific method for obtaining the extreme value fluctuation coefficient of the electrocardiogram data is as follows: The difference between the maximum value and the minimum value of the same extreme value pair contained in the frequency curve of the acquisition time of the ECG data is taken as the extreme value difference of the extreme value pair; The frequency interval between the maximum and minimum values ​​of the same extreme value pair contained in the frequency curve of the ECG data acquisition time is taken as the frequency difference of the extreme value pair; The ratio of the extreme value difference to the frequency difference of the extreme value pair is taken as the first ratio of the extreme value pair; The sum of the first ratios of all the extreme value pairs included in the frequency curve of the acquisition time of the electrocardiogram data is used as the extreme value fluctuation coefficient of the electrocardiogram data.

4. The user data analysis method based on wearable devices according to claim 1, characterized in that: The specific method for obtaining the fundamental frequency shift coefficient of the electrocardiogram data is as follows: Record the ratio of the maximum ECG energy intensity on the frequency curve of the kth and k-1th ECG data acquisition times as a first ratio, input the sum of the first ratio and the number 1 into a logarithmic function with the natural constant e as the base, and record the result as the first function value, and record the first function value as the fundamental wave frequency shift coefficient; Where k is a positive integer.

5. The user data analysis method based on wearable devices according to claim 1, characterized in that: The specific method for obtaining the intra-cluster fundamental wave drift deviation coefficient of the ECG waveform sequence is as follows: The DTW distance between the acquisition time frequency curve of the ECG data and the previous adjacent ECG data is recorded as the first distance of the ECG data; The product of the fundamental frequency shift coefficient of the electrocardiogram data and the first distance of the electrocardiogram data is recorded as the first product of the electrocardiogram data; The sum of the first products of all the electrocardiographic data included in the electrocardiographic waveform sequence is recorded as the intra-cluster fundamental wave drift deviation coefficient of the electrocardiographic waveform sequence.

6. The user data analysis method based on wearable devices according to claim 1, characterized in that: The specific method of obtaining the first correlation coefficient of two adjacent electrocardiogram waveform sequences includes: The minimum value of the number of ECG data contained in two adjacent ECG waveform sequences is recorded as the length threshold of the two ECG waveform sequences, the two adjacent ECG waveform sequences are truncated according to the length threshold, and the Pearson correlation coefficient of the two truncated sequences is used as the first correlation coefficient of the two adjacent ECG waveform sequences.

Images