Blood oxygen detection and calculation method based on infrared light and visible light

Abstract

The invention discloses a blood oxygen detection and calculation method based on infrared light and visible light, and relates to the technical field of blood characteristic measurement. The method comprises the steps that physiological index information and pulse wave detection information of all bracelet wearers are obtained, the physiological index information comprises at least one of body fat information and age information, and the pulse wave detection information comprises infrared light pulse wave signals and visible light pulse wave signals; based on the physiological index information, clustering the pulse wave detection information to obtain a plurality of pulse wave detection groups; based on the infrared light pressure value and the visible light pressure value of each piece of pulse wave detection information in the pulse wave detection group, determining the light absorption abnormity degree of each piece of pulse wave detection information; based on the light absorption anomalies, screening out a target training data set from the pulse wave detection groups; and based on the target training data set, training the target model to obtain a blood oxygen detection model, and realizing blood oxygen detection calculation through the blood oxygen detection model. The accuracy of blood oxygen detection calculation can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of blood characteristic measurement, and particularly relates to a blood oxygen detection and calculation method based on infrared light and visible light. Background Art

[0002] Sports bracelets are generally equipped with a blood oxygen detection function, and the core of this function lies in the detection of the content of oxyhemoglobin and deoxyhemoglobin. Its working principle is that different types of hemoglobin have different absorption characteristics under the irradiation of infrared light and visible light, and the content of oxyhemoglobin and deoxyhemoglobin is estimated by measuring and analyzing these absorption differences.

[0003] In practical applications, a sports bracelet is used to collect pulse wave signals under infrared light and visible light from the wrist part of the human body, and the pulse wave signals and the corresponding analyzed blood oxygen saturation are used to form a training data set. Subsequently, the neural network model is trained using the training data set, so as to realize more efficient and accurate blood oxygen detection through the neural network model.

[0004] However, a major problem currently faced is that the overall accuracy of the pulse wave signals measured by sports bracelets is relatively poor, which directly leads to an increase in the analysis error of the trained blood oxygen detection model, and further affects the accuracy of blood oxygen detection and calculation. Summary of the Invention

[0005] An embodiment of the present invention provides a blood oxygen detection and calculation method based on infrared light and visible light, which can reduce the calculation error of blood oxygen detection and improve the accuracy of blood oxygen detection and calculation.

[0006] In a first aspect of an embodiment of the present invention, a blood oxygen detection and calculation method based on infrared light and visible light is provided, and the method includes: Obtain the physiological index information and pulse wave detection information of each bracelet wearer, where the physiological index information includes at least one of body fat information and age information, and the pulse wave detection information includes an infrared light pulse wave signal and a visible light pulse wave signal; Cluster each pulse wave detection information based on the physiological index information to obtain multiple pulse wave detection groups; Based on the infrared light pressure value and visible light pressure value of each pulse wave detection information in the pulse wave detection group, determine the light absorption abnormality degree of each pulse wave detection information, and the light absorption abnormality degree is used to characterize the deviation abnormality degree between the pulse wave detection information and the corresponding pulse wave detection group; Based on each light absorption abnormality degree, screen out the target training data set from each pulse wave detection group; Based on the target training data set, train the target model to obtain a blood oxygen detection model, so as to realize blood oxygen detection and calculation through the blood oxygen detection model.

[0007] Further, the present invention also proposes to determine the light absorption abnormality degree of each pulse wave detection information based on the infrared light pressure value and the visible light pressure value of each pulse wave detection information in the pulse wave detection group, which may specifically include: Perform mean processing on each infrared light pressure value and each visible light pressure value of each pulse wave detection information in the pulse wave detection group to obtain an infrared light pressure mean value and a visible light pressure mean value; Use the difference between the infrared light pressure value of the target pulse wave detection information and the infrared light pressure mean value, and the difference between the visible light pressure value and the visible light pressure mean value to determine the light absorption abnormality degree of the target pulse wave detection information, where the target pulse wave detection information is any one of the pulse wave detection information.

[0008] Further, the present invention also proposes that before screening out the target training data set from each pulse wave detection group based on each light absorption abnormality degree, the method may further include: For the infrared light pulse wave signal and the visible light pulse wave signal of the pulse wave detection information, determine a segmentation region between two adjacent valley points to obtain a plurality of first segmentation regions of the infrared light pulse wave signal and a plurality of second segmentation regions of the visible light pulse wave signal; For the target segmentation region, determine the motion interference parameter of the target segmentation region based on the time span and pulse pressure of the target segmentation region, where the target segmentation region is any one of the first segmentation region and the second segmentation region; Use each motion interference parameter of the pulse wave detection information and the light absorption abnormality degree of the pulse wave detection information to determine the pulse wave error degree of the pulse wave detection information; Use the pulse wave error degree of the pulse wave detection information to determine the training priority coefficient of the pulse wave detection information; Based on each light absorption abnormality degree, screening out the target training data set from each pulse wave detection group may specifically include: Based on each training priority coefficient, screen out the target training data set from each pulse wave detection group.

[0009] Further, the present invention also proposes that the pulse pressure includes a pulse DC component and a pulse AC component; For the target segmentation region, determining the motion interference parameter of the target segmentation region based on the time span and pulse pressure of the target segmentation region may specifically include: Based on the time span of the target segmentation region, determine the periodic law presentation degree of the target segmentation region; Based on the pulse DC component and the pulse AC component at each moment in the target segmentation region, determine the motion interference tendency degree of the target segmentation region; Determine the motion interference parameter of the target segmented area by using the periodic law presentation degree and the motion interference tendency degree of the target segmented area.

[0010] Furthermore, the present invention also proposes to determine the periodic law presentation degree of the target segmented area based on the time span of the target segmented area, which may specifically include: Obtain the first time span mean value corresponding to the segmented area in the target pulse wave signal and the second time span mean value corresponding to the segmented area in the reference pulse wave signal. The target pulse wave signal is the pulse wave signal to which the target segmented area belongs, and the reference pulse wave signal is the pulse wave signal corresponding to the pulse wave detection information except the target pulse wave signal; Use the time span of the target segmented area and the first time span mean value to determine the first period feature conformity degree of the target segmented area; Use the time span of the reference segmented area and the second time span mean value to determine the second period feature conformity degree of the reference segmented area. The reference segmented area is the segmented area corresponding to the target segmented area in the reference pulse wave signal; Determine the periodic law presentation degree of the target segmented area based on the first period feature conformity degree and the second period feature conformity degree.

[0011] Furthermore, the present invention also proposes to determine the periodic law presentation degree of the target segmented area based on the first period feature conformity degree and the second period feature conformity degree, which may specifically include: Take the absolute value after subtracting the second period feature conformity degree from the first period feature conformity degree to obtain the first calculated value; Perform a normalization process on the reciprocal of the first calculated value to obtain the second calculated value; Use the second calculated value and the first period feature conformity degree to determine the periodic law presentation degree of the target segmented area.

[0012] Furthermore, the present invention also proposes to determine the motion interference tendency degree of the target segmented area based on the pulse DC component and the pulse AC component at each moment in the target segmented area, which may specifically include: Based on the pulse DC component and the pulse AC component at each moment in the target segmented area, respectively determine the component transfer interference degree at each moment in the target segmented area; Obtain the Pearson correlation coefficient of the pulse pressure at each moment in the target segmented area and the reference segmented area. The reference segmented area is the segmented area corresponding to the target segmented area in the reference pulse wave signal, the reference pulse wave signal is the pulse wave signal corresponding to the pulse wave detection information except the target pulse wave signal, and the target pulse wave signal is the pulse wave signal to which the target segmented area belongs; Use each component transfer interference degree and each Pearson correlation coefficient to determine the motion interference tendency degree of the target segmented area.

[0013] Further, the present invention also proposes to determine the component transfer interference degree at each moment in the target segmented region based on the DC component and the AC component of the pulse at each moment in the target segmented region, which may specifically include: Obtain the first slope of the DC component of the pulse and the second slope of the AC component of the pulse at the target moment in the target segmented region, where the target moment is any moment in the target segmented region; Normalize the ratio of the first slope to the second slope to obtain the component transfer interference degree at the target moment in the target segmented region.

[0014] Further, the present invention also proposes to obtain the Pearson correlation coefficient of the pulse pressure at each moment in the target segmented region and the reference segmented region, which may specifically include: Taking the target moment in the target segmented region as the center, extend the target duration to both sides to obtain the first analysis period, where the target moment is any moment in the target segmented region; Taking the target moment in the reference segmented region as the center, extend the target duration to both sides to obtain the second analysis period; Compare the first analysis period with the second analysis period to obtain the Pearson correlation coefficient of the pulse pressure at the target moment in the target segmented region and the reference segmented region.

[0015] Further, the present invention also proposes to train the target model based on the target training dataset to obtain a blood oxygen detection model, which may specifically include: Divide the target training dataset into multiple training batches; Perform data augmentation processing and data preprocessing on the first training data samples in each training batch to obtain second training data samples; Adjust the model parameters of the target model in sequence based on the second training data samples in the training batches according to the priority order of the training batches until the preset training end condition is met, and obtain the blood oxygen detection model.

[0016] The present invention has the following beneficial effects: In the blood oxygen detection calculation method based on infrared light and visible light provided by the embodiments of the present invention, first, physiological index information (such as body fat, age) of the bracelet wearer and pulse wave detection information (infrared light pulse wave signal and visible light pulse wave signal) are obtained. The physiological indexes can reflect the influence of individual differences on the pulse wave signal. Then, based on the physiological indexes, the pulse wave detection information is clustered, and similar individual data is grouped into the same group to form multiple pulse wave detection groups, which helps to analyze the characteristics of different groups subsequently. Then, according to the infrared light pressure value and visible light pressure value of each pulse wave detection information in the group, the light absorption abnormality degree is determined. The light absorption abnormality degree can characterize the degree of deviation abnormality of the pulse wave detection information from the pulse wave detection group where it is located, and can effectively identify abnormal signals caused by factors such as bracelet measurement errors and individual physiological fluctuations. After that, the target training data set is screened according to the light absorption abnormality degree, and the abnormal signals are removed to ensure the quality of the training data. Finally, the blood oxygen detection model is trained with the high-quality target training data set. Since the training data is accurate and reliable, the trained model is more accurate, thereby being able to reduce the blood oxygen detection calculation error and improve the accuracy of blood oxygen detection calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0018] Figure 1 It is a schematic flowchart of the first blood oxygen detection calculation method based on infrared light and visible light provided by an embodiment of the present invention; Figure 2 It is a schematic flowchart of the second blood oxygen detection calculation method based on infrared light and visible light provided by an embodiment of the present invention; Figure 3 It is a schematic diagram of the division of the segmented area provided by an embodiment of the present invention; Figure 4 It is a schematic flowchart of S202 provided by an embodiment of the present invention; Figure 5 It is a schematic flowchart of S402 provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following, in combination with the accompanying drawings and preferred embodiments, details a blood oxygen detection and calculation method based on infrared light and visible light proposed according to the present invention, including its specific implementation manner, structure, features and effects. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0021] It should be noted that the acquisition, storage, use, processing, etc. of data in the technical solution of the present invention all comply with the relevant regulations of laws and regulations.

[0022] It should be noted that in the embodiments of the present invention, some existing industry solutions such as certain software, components, models, etc. may be mentioned. They should be regarded as exemplary, and their purpose is only to illustrate the feasibility in the implementation of the technical solution of the present invention, but it does not mean that the applicant has already or necessarily used this solution.

[0023] In traditional existing blood oxygen detection methods, when constructing a training data set relying on dual-channel pulse wave signals of infrared light and visible light, due to individual physiological differences and motion interference factors during the signal acquisition process, the sample distribution of the original data set shows non-uniform characteristics. Physiological indicators such as body fat percentage and age affect the optical properties of subcutaneous tissue, resulting in differences in the light signal absorption intensity among different individuals; at the same time, limb movement in a dynamic environment causes distortion of the photoplethysmogram signal morphology, reducing the signal periodicity and amplitude stability. These factors together make it difficult to distinguish effective physiological features from noise interference when screening training data by conventional methods, resulting in a non-ideal feature mapping relationship being learned during the model training process.

[0024] For example, when a user wears a bracelet in an indoor fitness scenario, the myoelectric interference and limb displacement generated by the exercise cause the signal baseline of the visible light channel to drift, and at the same time, the pulse wave peak of the infrared light channel shows abnormal attenuation. When directly using the pulse wave signal containing such interference to generate training samples, for individuals with a body fat percentage higher than 30%, due to the scattering effect of the subcutaneous fat layer, the linear relationship between their infrared light absorption intensity and the standard model deviates, while for elderly users, the slope of the rising edge of the pulse wave decreases due to changes in blood vessel elasticity. At this time, there are both signal distortion samples and normal samples in the training data set, but traditional clustering methods cannot effectively identify sample subsets with similar physiological features but different signal qualities, resulting in gradient direction conflicts during the model weight update process.

[0025] When facing the above problems, the present invention first considers optimizing the construction process of the training data set through data stratification and quality screening. When the traditional method directly uses the original signal to construct the training set, it does not fully analyze the influence of physiological differences on the absorption of optical signals, and at the same time ignores the signal form distortion caused by motion interference. The present invention realizes that there are systematic differences in the optical characteristics of the subcutaneous tissue of wearers with different body fat percentages and ages, and it is necessary to establish an association mechanism between physiological characteristics and signal quality in the data preprocessing stage. For this reason, an attempt is made to use individual physiological indicators as the basis for data grouping, and sample subsets with similar optical absorption characteristics are formed through clustering to reduce the interference of the signal intensity difference within the group on model training. At the same time, a dynamic quality assessment index is introduced for motion interference, and the signal offset degree is quantified by the optical absorption abnormality, and high-quality samples are selected by combining the pulse wave cycle stability analysis. Compared with the prior art that directly uses all data for training, the present invention selects to construct a hierarchical and multi-dimensional screening mechanism to improve the quality of the data set through the dual strategies of physiological feature clustering and abnormal signal elimination.

[0026] For this, as Figure 1 shown, there is provided a schematic flowchart of a blood oxygen detection calculation method based on infrared light and visible light. The blood oxygen detection calculation method based on infrared light and visible light can be applied to a server, and the blood oxygen detection calculation method based on infrared light and visible light can include the following S101 to S105: S101, obtaining the physiological index information and pulse wave detection information of each bracelet wearer, where the physiological index information includes at least one of body fat information and age information, and the pulse wave detection information includes an infrared light pulse wave signal and a visible light pulse wave signal; S102, clustering each pulse wave detection information based on the physiological index information to obtain a plurality of pulse wave detection groups; S103, determining the optical absorption abnormality of each pulse wave detection information based on the infrared light pressure value and the visible light pressure value of each pulse wave detection information in the pulse wave detection group, where the optical absorption abnormality is used to characterize the deviation abnormality degree of the pulse wave detection information from the corresponding pulse wave detection group; S104, screening out a target training data set from each pulse wave detection group based on each optical absorption abnormality; S105, training a target model based on the target training data set to obtain a blood oxygen detection model, so as to implement blood oxygen detection calculation through the blood oxygen detection model.

[0027] In this embodiment, the physiological index information refers to the body fat percentage or age data of the wearer obtained through the smart bracelet, which can be specifically implemented by using a bioelectrical impedance sensor or a user input interface. The physiological index information is used to reflect individual physiological differences, so as to more accurately divide similar groups in the subsequent clustering process. For example, people with higher body fat have a higher fat content in the wrist tissue. Excessive light absorption and consumption occur in the fat for infrared light and visible light, resulting in normal physiological differences in blood oxygen measurement data among different wearers. And the blood circulation rate of wearers with higher age is slower, which will also cause differences in blood oxygen measurement information. Therefore, in order to reduce the analysis error caused by normal physiological differences to blood oxygen detection, clustering is performed according to physiological index information such as the body fat and age of the wearer to obtain multiple pulse wave detection groups.

[0028] The pulse wave detection information refers to the periodic pulse fluctuation signal collected by an infrared light and visible light sensor, which can be specifically implemented by using the photoplethysmography technology. The pulse wave detection information is used to reflect the absorption characteristics of hemoglobin in the blood for light of different wavelengths.

[0029] Clustering refers to grouping and classifying pulse wave data according to physiological indexes, which can be specifically implemented by using the K-means algorithm or the hierarchical clustering algorithm. The role of clustering is to reduce the interference of individual physiological differences on pulse wave analysis and improve the data consistency within the group.

[0030] The light absorption abnormality degree refers to the deviation degree calculated by comparing the infrared light pressure value and the visible light pressure value within the same group, which can be specifically implemented by using the Euclidean distance or the standardized residual method. The light absorption abnormality degree is used to identify abnormal pulse wave signals and exclude low-quality data caused by measurement errors.

[0031] The target training data set refers to the set of high-quality pulse wave data retained after screening by the abnormality degree, which can be specifically implemented by setting an abnormality degree threshold or a sorting and selecting method. The role of the target training data set is to provide reliable samples for model training and reduce the impact of noise data on the model accuracy.

[0032] The blood oxygen detection model refers to the mapping model between the pulse wave signal and the blood oxygen saturation established by a machine learning algorithm, which can be specifically implemented by using a convolutional neural network or a support vector machine. The role of the blood oxygen detection model is to improve the accuracy of blood oxygen detection calculation through the optimized training data.

[0033] The working process and principle of the present invention are as follows: First, the physiological index information and the pulse wave detection information of the smart bracelet wearer are obtained. The physiological index information includes at least one of body fat information and age information, and the pulse wave detection information includes an infrared light pulse wave signal and a visible light pulse wave signal. These information serve as the basic data for subsequent processing.

[0034] Next, cluster each piece of pulse wave detection information based on the physiological index information to obtain multiple pulse wave detection groups. This step groups data with similar physiological characteristics, providing a more accurate basis for subsequent analysis.

[0035] Then, analyze the pulse wave detection information in each pulse wave detection group. Based on the infrared light pressure value and the visible light pressure value, determine the light absorption abnormality degree of each piece of pulse wave detection information. The light absorption abnormality degree is used to characterize the degree of deviation abnormality between the pulse wave detection information and the corresponding pulse wave detection group. This step can identify abnormal data in the pulse wave detection group, improving the accuracy of subsequent analysis.

[0036] Based on the calculated light absorption abnormality degree, screen out the target training data set from each pulse wave detection group. This step improves the quality of the training data set by removing abnormal data in the pulse wave detection group.

[0037] Finally, use the screened target training data set to train the target model to obtain a blood oxygen detection model. Through the trained blood oxygen detection model, more accurate blood oxygen detection calculations can be achieved.

[0038] As an example, first obtain the physiological index information and pulse wave detection information of the bracelet wearer. The physiological index information includes body fat percentage and age, and the pulse wave detection information includes infrared light pulse wave signals and visible light pulse wave signals. Use the K-means clustering algorithm to cluster the pulse wave detection information based on body fat percentage and age to obtain multiple pulse wave detection groups. For example, the data can be divided into a low body fat young group, a high body fat young group, a low body fat middle-aged group, a high body fat middle-aged group, etc.

[0039] Then, calculate the mean values of the infrared light pressure value and the visible light pressure value for the data in each pulse wave detection group. Then, calculate the difference between the pressure value of each piece of pulse wave detection information and the mean value to obtain the light absorption abnormality degree. Set a light absorption abnormality degree threshold, and thus screen out the data with a light absorption abnormality degree lower than the threshold to form the target training data set.

[0040] Then, use a deep learning model such as a long short-term memory network as the target model and train it using the target training data set. During the training process, the batch gradient descent method is used to optimize the model parameters until the model converges or reaches the preset number of training epochs. The finally obtained blood oxygen detection model can accept new pulse wave signals as input and output the corresponding estimated blood oxygen saturation value.

[0041] Through this embodiment, first, physiological index information (such as body fat and age) of the bracelet wearer and pulse wave detection information (infrared light pulse wave signal and visible light pulse wave signal) are obtained. The physiological index can reflect the influence of individual differences on the pulse wave signal. Then, based on the physiological index, the pulse wave detection information is clustered, and similar individual data are grouped into the same group to form multiple pulse wave detection groups, which helps subsequent analysis of different group characteristics. Then, according to the infrared light pressure value and visible light pressure value of each pulse wave detection information within the group, the light absorption abnormality degree is determined. The light absorption abnormality degree can characterize the deviation abnormality degree between the pulse wave detection information and the overall state within the pulse wave detection group, and can effectively identify abnormal signals caused by factors such as bracelet measurement errors and individual physiological fluctuations. After that, the target training data set is screened based on the light absorption abnormality degree, and abnormal signals are removed to ensure the quality of the training data. Finally, the target model is trained with the high-quality target training data set to obtain a blood oxygen detection model. Since the training data is accurate and reliable, the trained model is more accurate, thereby reducing the calculation error of blood oxygen detection and improving the accuracy of blood oxygen detection calculation.

[0042] In some of the above solutions of the present invention, after clustering the pulse wave detection information based on the physiological index information, it is necessary to determine the light absorption abnormality degree of each pulse wave detection information. However, when determining the light absorption abnormality degree, if a unified benchmark is not established for the pressure value distribution within the same detection group, the calculation result of the abnormality degree may be interfered by the discreteness within the group, and it cannot accurately reflect the deviation degree between the individual data and the overall state within the group.

[0043] For this reason, the present invention further proposes that S103 may specifically include: Perform mean processing on each infrared light pressure value and each visible light pressure value of each pulse wave detection information in the pulse wave detection group to obtain the infrared light pressure mean value and the visible light pressure mean value; Use the difference between the infrared light pressure value of the target pulse wave detection information and the infrared light pressure mean value, and the difference between the visible light pressure value and the visible light pressure mean value to determine the light absorption abnormality degree of the target pulse wave detection information, where the target pulse wave detection information is any one of the pulse wave detection information.

[0044] In this embodiment, the mean processing can adopt the arithmetic mean or weighted mean method, where the weighting coefficient can be dynamically adjusted according to the physiological index differences corresponding to each pulse wave detection information. The difference calculation can choose the absolute value difference or square difference form. When using the square difference, normalization processing is required to eliminate the dimension difference. The determination of the light absorption abnormality degree can combine the linear combination of the infrared light pressure difference and the visible light pressure difference, where the combination weight is set according to the measurement stability of the two light sources. For example, in a motion interference environment, the visible light signal fluctuates greatly. At this time, the weight ratio of the visible light pressure difference can be reduced to improve the robustness of the abnormality degree evaluation.

[0045] Specifically, first, the average of all infrared light pressure values within the same pulse wave detection group is taken to obtain the infrared light pressure mean value reflecting the typical state within the group. Similarly, the visible light pressure mean value is calculated in the same way. Subsequently, for the target pulse wave detection information, the absolute deviation between its infrared light pressure value and the group mean value is calculated respectively, and the total infrared light deviation amount is obtained after summation. The visible light deviation amount is obtained by the same method. Finally, the deviation amounts of the two light sources are weighted and summed or multiplied according to a preset ratio to obtain the light absorption abnormality degree comprehensively reflecting the signal deviation degree.

[0046] As an example, the light absorption abnormality degree of the target pulse wave detection information can be specifically determined by the following formula 1: Formula 1 In formula 1, is used to characterize the light absorption abnormality degree of the i-th pulse wave detection information in the pulse wave detection group, is used to characterize the infrared light pressure value of the j-th data in the i-th pulse wave detection information, is used to characterize the visible light pressure value of the j-th data in the i-th pulse wave detection information. is used to characterize the infrared light pressure mean value of all pulse wave detection information at the j-th data in the pulse wave detection group corresponding to the i-th pulse wave detection information, is used to characterize the visible light pressure mean value of all pulse wave detection information at the j-th data in the pulse wave detection group corresponding to the i-th pulse wave detection information, and M is used to characterize the amount of data in the i-th pulse wave detection information, represents taking the absolute value.

[0047] Among them, The larger it is, the greater the deviation between the infrared light pressure value in the i-th pulse wave detection information and the overall data in the similar pulse wave detection group; The larger it is, the greater the deviation between the visible light pressure value in the i-th pulse wave detection information and the overall data in the similar pulse wave detection group.

[0048] Through this embodiment, abnormal data in the pulse wave detection information can be effectively identified and quantified. By calculating the light absorption abnormality degree, the deviation degree between each pulse wave detection information and the overall data can be accurately evaluated. This method can specifically screen out high-quality training data, thereby improving the training effect and accuracy of the subsequent blood oxygen detection model. At the same time, this method also provides a quantitative basis for the identification and processing of abnormal data, which helps to improve the overall accuracy and reliability of blood oxygen detection.

[0049] In some of the above solutions of the present invention, the pulse wave detection information is clustered based on physiological index information, and the target training data set is screened based on the light absorption abnormality. However, the influence of motion interference on the pulse wave signal is not considered during the screening process, resulting in some pulse wave detection information containing motion noise being selected into the training data set, which affects the training effect of the blood oxygen detection model.

[0050] In response to this, as Figure 2 shown, before S104, the blood oxygen detection calculation method based on infrared light and visible light may further include the following S201 to S204: S201, for the infrared light pulse wave signal and the visible light pulse wave signal of the pulse wave detection information, a segmented area is determined between two adjacent valley points, and a plurality of first segmented areas of the infrared light pulse wave signal and a plurality of second segmented areas of the visible light pulse wave signal are obtained; S202, for the target segmented area, based on the time span and pulse pressure of the target segmented area, determine the motion interference parameter of the target segmented area, and the target segmented area is any one of the first segmented area and the second segmented area; S203, use each motion interference parameter of the pulse wave detection information and the light absorption abnormality of the pulse wave detection information to determine the pulse wave error degree of the pulse wave detection information; S204, use the pulse wave error degree of the pulse wave detection information to determine the training priority coefficient of the pulse wave detection information; S104 may specifically include: Based on each training priority coefficient, the target training data set is screened from each pulse wave detection group.

[0051] In this embodiment, the division of the segmented area is realized by detecting the valley point. As Figure 3 shown, a schematic diagram of the division of the segmented area is provided. Among them, the maximum pulse pressure within one pulse period is the peak point 310, the minimum pulse pressure within one pulse period is the valley point 320, and a single segmented area can be formed between two adjacent valley points 320.

[0052] The motion interference parameter is a parameter used to measure the degree of motion interference on the pulse wave signal within the target segmented area. Since human body movement will interfere with the pulse wave signal and cause signal distortion, the determination of the motion interference parameter helps to evaluate the signal quality.

[0053] The pulse wave error degree is an index obtained by comprehensively considering each motion interference parameter and the light absorption abnormality, and is used to measure the overall error degree of the pulse wave detection information. The larger the pulse wave error degree, the greater the deviation between the detection information and the true physiological condition, and the worse the signal quality.

[0054] The training priority coefficient is a parameter determined according to the pulse wave error degree, and is used to evaluate the priority of each pulse wave detection group in subsequent model training. The higher the training priority coefficient, the relatively better the data quality of the pulse wave detection group, and higher weight or priority should be given when training the model.

[0055] As an example, first use a signal processing algorithm (such as a derivative-based extreme point detection algorithm) to detect the valley points of the infrared light pulse wave signal and the visible light pulse wave signal. Specifically, calculate the first derivative of the signal, and the point where the derivative is zero and the second derivative is greater than zero is the valley point. Then, according to the order of the detected valley points, the signal segment between two adjacent valley points is determined as a segmented region.

[0056] Then, record the start time and end time of the target segmented region to determine the time span of the target segmented region; at the same time, measure the pulse pressure value corresponding to the target segmented region in real time. The motion interference parameter D can be calculated by a preset function or model, which comprehensively considers the influence of the time span T and the pulse pressure P. For example, D = f(T, P), where f can be a linear function, a non-linear function, or a model trained based on a machine learning algorithm. In practical applications, a large amount of experimental data is required to determine the specific form of the function f so that the motion interference parameter can accurately reflect the degree of motion interference on the pulse wave signal in the target segmented region.

[0057] Then, using each motion interference parameter of the pulse wave detection information and the light absorption abnormality degree of the pulse wave detection information, the pulse wave error degree of the pulse wave detection information is determined by the following formula 2: Formula 2 In formula 2, represents the pulse wave error degree of the i-th pulse wave detection information, represents the light absorption abnormality degree of the i-th pulse wave detection information, represents the average value of each motion interference parameter of the i-th pulse wave detection information.

[0058] Then, using the pulse wave error degree of the pulse wave detection information, the training priority coefficient of the pulse wave detection information is determined by the following formula 3: Formula 3 In formula 3, represents the training priority coefficient of the i-th pulse wave detection information, The pulse wave error degree is used to characterize the i-th pulse wave detection information, and norm is used to characterize the normalization process. It should be noted that to ensure the significance of the calculation results, in the fractional operation of the embodiments of the present invention, when the denominator is 0, a tuning factor greater than 0 needs to be added to the denominator for addition to prevent the denominator from being 0. The value of the tuning factor is set by the implementer according to the actual situation, and the present invention does not make special restrictions.

[0059] Through this embodiment, it is possible to effectively screen out pulse wave detection data with higher quality, reducing the noise and interference in the training dataset. By considering the motion interference factors, the robustness of the blood oxygen detection model is improved. At the same time, by introducing the training priority coefficient, it is ensured that high-quality data is fully utilized in model training, thereby improving the overall performance and accuracy of the blood oxygen detection model.

[0060] In some of the above solutions of the present invention, it is proposed to screen training data by dividing the segmented region of the pulse wave signal and calculating the motion interference parameter. However, in this process, due to the physiological differences or motion interference of different wearers, the time span of the segmented region may show irregular periodic changes, and the DC component and AC component in the pulse wave may have abnormal fluctuations due to motion, thus affecting the accuracy of the motion interference parameter.

[0061] In response to this, as Figure 4 shown, the present invention further proposes that the pulse pressure includes a pulse DC component and a pulse AC component; S202 may specifically include the following S401 to S403: S401, based on the time span of the target segmented region, determine the periodicity regularity degree of the target segmented region; S402, based on the pulse DC component and pulse AC component at each moment in the target segmented region, determine the motion interference tendency degree of the target segmented region; S403, using the periodicity regularity degree and motion interference tendency degree of the target segmented region, determine the motion interference parameter of the target segmented region.

[0062] In this embodiment, the periodicity regularity degree is used to measure the obviousness of the periodic law of the pulse wave signal within the target segmented region. Under normal circumstances, the pulse wave signal has a certain periodicity. The higher the periodicity regularity degree, the more the signal conforms to the normal periodic law and the relatively smaller the interference it receives; conversely, the lower the periodicity regularity degree, it may mean that the signal is more interfered and the periodicity is disrupted.

[0063] The DC component of the pulse refers to the relatively stable part of the pulse wave signal that does not change rapidly with the pulse beat, and usually reflects the average level or background signal intensity of the pulse wave signal. The AC component of the pulse refers to the part of the pulse wave signal that changes rapidly with the pulse beat, directly reflecting the pulse beat situation, and contains important information such as the frequency and amplitude of the pulse.

[0064] The motion interference tendency is used to describe the change trend or degree of the motion interference on the pulse wave signal in the target segmented area. By analyzing the relationship between the DC component and the AC component of the pulse at each moment, it can be judged whether the influence of the motion interference on the pulse wave signal is gradually increasing, decreasing or remaining relatively stable, etc.

[0065] As an example, first, calculate the relative deviation between the time span of the target segmented area and the average value of the time spans of each segmented area of the pulse wave signal to which the target segmented area belongs, that is, the coefficient of variation of the period. The coefficient of variation of the period reflects the degree of dispersion of the period values. The smaller the coefficient of variation, the more stable the period and the higher the degree of presentation of the period law. Then, determine the degree of presentation of the period law according to the size of the coefficient of variation of the period of the target segmented area. For example, a mapping relationship can be preset, and different values of the degree of presentation of the period law are assigned according to different coefficient of variation intervals of the period.

[0066] Then, for each moment in the target segmented area, the DC component and the AC component of the pulse can be extracted by methods such as filtering. For example, a low-pass filter is used to extract the DC component, and a high-pass filter is used to extract the AC component. The cut-off frequency of the low-pass filter is usually set in a lower frequency band to filter out the high-frequency interference in the pulse wave signal and retain the relatively stable DC part; the cut-off frequency of the high-pass filter is determined according to the main frequency range of the pulse wave signal to extract the AC part reflecting the pulse beat.

[0067] Then, calculate the DC-AC ratio at each moment, and calculate the change rate of this ratio at adjacent moments. Then analyze the change rate to judge its change trend. If the change rate shows an obvious increasing or decreasing trend within a period of time, it means that the motion interference has an impact on the proportional relationship between the DC component and the AC component of the pulse, and the motion interference tendency is relatively high; if the change rate changes little and is relatively stable, it means that the motion interference tendency is relatively low. The change trend can be quantified by statistical indicators such as the variance and slope of the change rate.

[0068] Finally, the motion interference parameter is determined by the following formula 4: Formula 4 In formula 4, is used to characterize the motion interference parameter of the rth segmented area in the ith pulse wave detection information, Used to characterize the motion interference tendency of the r-th segmented area in the i-th pulse wave detection information, Used to characterize the motion interference tendency of the r-th segmented area in the i-th pulse wave detection information.

[0069] Wherein, The larger it is, the greater the impact of motion interference on the ratio relationship between the DC component and the AC component of the pulse in the r-th segmented area of the i-th pulse wave detection information, that is, the larger the motion interference parameter; The larger it is, the more stable the period of the r-th segmented area in the i-th pulse wave detection information, that is, the smaller the motion interference parameter.

[0070] Through this embodiment, the motion interference in the pulse wave signal can be effectively identified and quantified. Thereby improving the accuracy and reliability of subsequent blood oxygen detection calculations. Specifically, by analyzing the periodic regularity of the pulse wave signal and the change characteristics of the DC / AC components, it is possible to more accurately determine whether the abnormal fluctuations in the signal are caused by physiological changes or motion interference. By introducing the motion interference parameter, a more reliable basis can be provided for subsequent data screening and model training, thereby constructing a more robust blood oxygen detection model.

[0071] In some of the above solutions of the present invention, it is proposed to evaluate the motion interference parameter by determining the periodic law presentation degree of the segmented area. However, in this process, it is difficult to accurately reflect the true physiological cycle characteristics only relying on the time span information of a single pulse wave signal, and it is easy to cause deviations in the judgment of the periodic law due to individual differences or instantaneous interference, affecting the reliability of the subsequent motion interference parameter.

[0072] In response to this, the present invention further proposes that S401 may specifically include: Obtain the first time span mean value corresponding to the segmented area in the target pulse wave signal and the second time span mean value corresponding to the segmented area in the reference pulse wave signal. The target pulse wave signal is the pulse wave signal to which the target segmented area belongs, and the reference pulse wave signal is the pulse wave signal corresponding to the pulse wave detection information except the target pulse wave signal; Use the time span of the target segmented area and the first time span mean value to determine the first cycle feature conformity degree of the target segmented area; Use the time span of the reference segmented area and the second time span mean value to determine the second cycle feature conformity degree of the reference segmented area. The reference segmented area is the segmented area corresponding to the target segmented area in the reference pulse wave signal; Based on the first cycle feature conformity degree and the second cycle feature conformity degree, determine the periodic law presentation degree of the target segmented area.

[0073] In this embodiment, it should be noted that whether it is visible light or infrared light, the periodic law of the corresponding pulse wave signal is caused by the contraction and relaxation of arterial blood vessels, which is reflected as the regular change of the pressure on the corresponding hemoglobin. That is, although visible light is more sensitive to oxyhemoglobin and infrared light is more sensitive to deoxyhemoglobin, the contraction and relaxation of arterial blood vessels on both of them are the same. That is, for a single blood oxygen measurement process, the periodic performance of visible light and infrared light should be the same. Therefore, the periodic law presentation degree of the target segmented area can be determined based on the first periodic feature conformity degree of the target segmented area and the second periodic feature conformity degree of the reference segmented area.

[0074] The target pulse wave signal and the reference pulse wave signal can respectively refer to the infrared light pulse wave signal and the visible light pulse wave signal, and both have the characteristic of synchronous acquisition. The first time span mean value is the average value of the time spans of all segmented areas in the target pulse wave signal, and the second time span mean value is the average value of the time spans of the corresponding segmented areas in the reference pulse wave signal.

[0075] The first periodic feature conformity degree is obtained by calculating the deviation ratio of the time span of the target segmented area to the first time span mean value. The second periodic feature conformity degree is calculated by the same method to obtain the deviation ratio of the time span of the reference segmented area to the second time span mean value. The periodic law presentation degree is comprehensively judged by comparing the differences between the two periodic feature conformity degrees.

[0076] As an example, first obtain the first time span mean value corresponding to the segmented area in the target pulse wave signal and the second time span mean value corresponding to the segmented area in the reference pulse wave signal. The target pulse wave signal is the pulse wave signal to which the target segmented area belongs, and the reference pulse wave signal is the corresponding pulse wave signal other than the target pulse wave signal in the pulse wave detection information.

[0077] Then, use the time span of the target segmented area and the first time span mean value to determine the first periodic feature conformity degree of the target segmented area. Specifically, the first periodic feature conformity degree can be determined by the following formula 5: Formula 5 In formula 5, is used to represent the periodic feature conformity degree of the r-th segmented area in the i-th pulse wave detection information, is used to represent the time span of the r-th segmented area in the i-th pulse wave detection information, is used to represent the time span mean value of the pulse wave signal to which the r-th segmented area in the i-th pulse wave detection information belongs.

[0078] Referring to the above formula 5 again, the second cycle feature compliance of the reference segmented region is determined by using the time span of the reference segmented region and the mean value of the second time span.

[0079] Finally, based on the first cycle feature compliance and the second cycle feature compliance, the periodic law presentation degree of the target segmented region is determined. For example, the first cycle feature compliance and the second cycle feature compliance can be weighted and averaged to obtain the periodic law presentation degree of the target segmented region.

[0080] Through this embodiment, the periodicity of the target segmented region can be effectively evaluated, thereby providing a more accurate data basis for subsequent blood oxygen detection calculations. Since the time span characteristics of the target pulse wave signal and the reference pulse wave signal are considered, this method can more comprehensively reflect the periodic changes of the pulse wave signal, improving the accuracy and reliability of blood oxygen detection.

[0081] In some of the above solutions of the present invention, it is proposed to evaluate the periodic characteristics of the target segmented region through the periodic law presentation degree. However, directly comparing the first cycle feature compliance with the second cycle feature compliance to determine the accuracy of the periodic law presentation degree of the target segmented region is relatively low.

[0082] In response to this, the present invention further proposes to determine the periodic law presentation degree of the target segmented region based on the first cycle feature compliance and the second cycle feature compliance, which specifically may include: Taking the absolute value after subtracting the first cycle feature compliance from the second cycle feature compliance to obtain a first calculated value; Performing a normalization process on the reciprocal of the first calculated value to obtain a second calculated value; Using the second calculated value and the first cycle feature compliance to determine the periodic law presentation degree of the target segmented region.

[0083] In this embodiment, the first calculated value is obtained by taking the absolute value after subtracting the first cycle feature compliance from the second cycle feature compliance, the second calculated value is obtained by performing a normalization process on the reciprocal of the first calculated value, and the periodic law presentation degree is further determined by combining the second calculated value with the first cycle feature compliance.

[0084] As an example, the periodic law presentation degree of the target segmented region can be specifically determined by the following formula 6: Formula 6 In formula 6, is used to represent the periodic law presentation degree of the r-th segmented region in the i-th pulse wave detection information, Used to represent the absolute value of the difference between the periodic feature conformity degree of the r-th segmented region in the i-th pulse wave detection information and the periodic feature conformity degree of the corresponding reference segmented region Used to represent the periodic feature conformity degree of the r-th segmented region in the i-th pulse wave detection information, and norm is used to represent the normalization process

[0085] Among them, if The smaller it is, the better the periodic consistency of this segmented region in the infrared light pulse wave signal and the visible light pulse wave signal, indicating that the degree of periodic law presentation is higher

[0086] Through this embodiment, the degree of periodic law presentation of the target segmented region can be evaluated more accurately. Thus, the pulse wave signal segment with good periodicity can be effectively identified, thereby improving the accuracy of subsequent blood oxygen detection calculation. Further, by introducing the second calculated value, this solution can consider the periodic features of the target pulse wave signal and the reference pulse wave signal at the same time, making the evaluation of the degree of periodic law presentation more comprehensive and objective

[0087] In some of the above solutions of the present invention, based on the method for determining the motion interference tendency degree of the pulse DC component and the pulse AC component, although it can reflect the influence of motion interference to a certain extent, in practical applications, due to the instantaneous and random nature of the signal interference caused by wrist movement, it is difficult to accurately quantify the tendency degree of motion interference only relying on the component changes of a single pulse wave signal. Especially when the synchronization difference between different optical signals is not fully considered, it may lead to deviations in the evaluation of the motion interference tendency degree

[0088] In this regard, as Figure 5 shown, S402 may specifically include the following S501 to S503 S501, based on the pulse DC component and the pulse AC component at each moment in the target segmented region, respectively determine the component transfer interference degree at each moment in the target segmented region S502, obtain the Pearson correlation coefficient of the pulse pressure at each moment in the target segmented region and the reference segmented region. The reference segmented region is the segmented region corresponding to the target segmented region in the reference pulse wave signal, the reference pulse wave signal is the pulse wave signal corresponding to the pulse wave detection information except the target pulse wave signal, and the target pulse wave signal is the pulse wave signal to which the target segmented region belongs S503, use each component transfer interference degree and each Pearson correlation coefficient to determine the motion interference tendency degree of the target segmented region

[0089] In this embodiment, the component transfer interference degree is obtained through normalization of the ratio of the first slope of the DC component of the pulse to the second slope of the AC component of the pulse at the target moment. The Pearson correlation coefficient is obtained by comparison after forming a first analysis period and a second analysis period by extending the target duration to both sides centered on the target moment. For example, the first slope is the linear regression coefficient of the DC component within 0.5 seconds before and after the target moment, the second slope is the change rate of the AC component within the same time window, and the ratio is limited to the interval [0, 1] after normalization processing; the target duration can be set to 0.3 seconds. At this time, the first analysis period covers three sampling points before and after the target moment respectively, and the second analysis period corresponds to the same time span of the reference pulse wave signal.

[0090] Specifically, when the wrist moves, the DC components and AC components in the infrared light and visible light signals will produce asynchronous fluctuations. By calculating the ratio of the slope of the DC component to the slope of the AC component at the target moment, the coordination of the component changes at this moment can be quantified. The higher the degree of deviation of the ratio from 1 indicates the greater the signal distortion caused by motion interference. At the same time, adjacent periods are intercepted centered on the target moment, and the Pearson correlation coefficients of the two optical signals in the corresponding periods are calculated respectively. When the correlation coefficient is lower than 0.8, it indicates that there is a significant desynchronization phenomenon between the signals. After weighted summation of the component transfer interference degree and the Pearson correlation coefficient, when the weighted value exceeds the threshold of 0.6, it is determined that there is high-intensity motion interference at this moment.

[0091] As an example, first, based on the DC component and AC component of the pulse at each moment in the target segmented region, the component transfer interference degree at each moment in the target segmented region is determined respectively. Specifically, first, the DC component data and AC component data of the pulse at each moment in the target segmented region are obtained. Then, the data at each moment is processed according to the pre-determined component transfer interference degree algorithm to calculate the component transfer interference degree at this moment.

[0092] Furthermore, the Pearson correlation coefficient of the pulse pressure at each moment in the target segmented region and the reference segmented region is obtained. Among them, the reference segmented region is the segmented region corresponding to the target segmented region in the reference pulse wave signal, and the reference pulse wave signal is the pulse wave signal corresponding to the pulse wave detection information except the target pulse wave signal. In specific implementation, the infrared light pulse wave signal can be selected as the target pulse wave signal, and the visible light pulse wave signal can be selected as the reference pulse wave signal. For each moment, the Pearson correlation coefficient of the pulse pressure data of the target segmented region and the reference segmented region at this moment is calculated.

[0093] Thus, using each component transfer interference degree and each Pearson correlation coefficient, the motion interference tendency degree of the target segmented region is determined. Specifically, the motion interference tendency degree of the target segmented region can be determined by the following formula 7: Formula 7 In Formula 7, is used to characterize the motion interference trend degree of the r-th segmented region in the i-th pulse wave detection information, is used to characterize the average value of the Pearson correlation coefficients in the r-th segmented region in the i-th pulse wave detection information, is used to characterize the average value of the component transfer interference degrees in the r-th segmented region in the i-th pulse wave detection information.

[0094] Through this embodiment, the motion interference degree in the pulse wave signal can be effectively evaluated. By analyzing the change characteristics of the pulse DC component and the AC component, and comparing the correlation of the pulse wave signals under different light sources, the signal abnormalities caused by motion can be accurately identified. This method not only considers the internal characteristics of a single signal, but also utilizes the relationship between multi-light source signals, thereby improving the accuracy and reliability of motion interference recognition.

[0095] In some of the above solutions of the present invention, in the process of calculating the motion interference trend degree for the target segmented region, there is a problem of insufficient accuracy in the determination method of the component transfer interference degree. Specifically, the traditional method does not fully consider the dynamic change relationship between the pulse DC component and the AC component, resulting in a deviation in the sensitivity evaluation of the motion interference in the signal.

[0096] In response to this, the present invention further proposes that S501 may specifically include: Obtain the first slope of the pulse DC component and the second slope of the pulse AC component at the target moment in the target segmented region, where the target moment is any moment in the target segmented region; Normalize the ratio of the first slope to the second slope to obtain the component transfer interference degree at the target moment in the target segmented region.

[0097] In this embodiment, the first slope is calculated using the linear regression method within a sliding time window, and the window length is set as an adjustable parameter within the range of 0.2 seconds to 0.5 seconds. The second slope is calculated using the synchronous difference method based on the change rate between adjacent sampling points. The normalization process is implemented through the z-score method. Specifically, the ratio is divided by the standard deviation after subtracting the within-group mean to ensure the comparability of data in different signal channels. The configuration of the time window is associated with the sampling frequency of the pulse wave signal. When the sampling frequency is 100Hz, the window length preferably includes 20 to 50 data points.

[0098] Specifically, a time window of 0.1 seconds before and after the target moment is selected, and the DC component data within this window is linearly fitted to obtain the first slope to characterize the steady-state change trend of the DC component. The difference between adjacent sampling points of the AC component within this window is synchronously extracted, and the average change rate is calculated as the second slope to reflect the instantaneous fluctuation characteristics of the pulse wave. The dynamic relationship coefficient between the two components is obtained through ratio calculation, and then the influence of individual differences is eliminated through normalization. This method effectively captures the signal distortion characteristics caused by motion interference. When the motion interference causes a sudden change in the DC component, the first slope increases significantly, while the regular fluctuation of the AC component is disrupted, resulting in an abnormal decrease in the second slope. At this time, the abnormal change in the ratio can accurately reflect the degree of component transfer interference. By quantifying the co-variation mode of the two components, the sensitivity of motion interference detection can be improved, providing a more reliable evaluation basis for subsequent training data screening.

[0099] As an example, when a wearable device worn on the wrist collects visible light pulse wave signals, for the time series data of a certain target segmented area in the signal, the DC component and AC component of the pulse at each sampling moment within this area are differentiated. Taking the DC component data point at a certain moment as an example, a DC component data segment of 50 milliseconds before and after this moment is taken, and the slope of the DC component corresponding to this moment is obtained through least squares fitting; at the same time, an AC component data segment with the same time span is taken, and the slope of the AC component is calculated. The absolute ratio of the DC component slope to the AC component slope is input into the sigmoid function for normalization processing, and the component transfer interference degree value at this moment is output, and this numerical range is constrained between 0 and 1.

[0100] Through this embodiment, the problem of difficult interference recognition caused by signal component offset in the motion state is effectively solved. By quantitatively evaluating the dynamic relationship change characteristics of the DC component and the AC component at each moment, the signal distortion period caused by limb movement can be accurately captured, thereby providing a reliable interference evaluation basis for subsequent screening of high-quality training data, and ultimately improving the anti-motion interference ability of the blood oxygen detection model.

[0101] In some of the above solutions of the present invention, when determining the Pearson correlation coefficient, only the pulse pressures at a single moment in the target segmented area and the reference segmented area are compared, which is likely to result in a low accuracy of the Pearson correlation coefficient.

[0102] In this regard, the present invention further proposes that S502 may specifically include: Centering on the target moment in the target segmented area, extend the target duration to both sides to obtain the first analysis period, and the target moment is any moment in the target segmented area; Centering on the target moment in the reference segmented area, extend the target duration to both sides to obtain the second analysis period; Compare the first analysis period with the second analysis period to obtain the Pearson correlation coefficient of the pulse pressure at the target moment in the target segmented area and the reference segmented area.

[0103] In this embodiment, the target moment is a time point arbitrarily selected in the target segmented area, and subsequent analysis will be carried out around this moment. By constructing an analysis period centered on this moment, the changes in characteristics such as pulse pressure are studied.

[0104] The target duration is the time length used to determine the analysis period and is a preset fixed value. Centered on the target moment, extend this target duration to both sides to construct a time window for analysis.

[0105] The first analysis period is a time period obtained by extending the target duration to both sides centered on the target moment in the target segmented area. The data such as pulse pressure within this period will be used for subsequent comparative analysis with the data in the corresponding period of the reference segmented area.

[0106] The second analysis period is a time period obtained by extending the target duration to both sides centered on the target moment in the reference segmented area, and has the same time span as the first analysis period, and is used for data comparison with the first analysis period.

[0107] Pearson correlation coefficient: A statistical indicator used to measure the degree of linear correlation between two variables, with a value range between -1 and 1. In this embodiment, it is used to measure the degree of linear correlation between the pulse pressures in the target segmented area and the reference segmented area near the target moment (i.e., within the first analysis period and the second analysis period). When the Pearson correlation coefficient is close to 1, it indicates a high degree of positive correlation between the two variables; when it is close to -1, it indicates a high degree of negative correlation; when it is close to 0, it indicates a weak linear correlation.

[0108] As an example, select any moment in the target segmented area as the target moment. Taking this target moment as the reference point, extend the preset duration in both the forward and backward directions of the time axis to form the first analysis period. Select the time point corresponding to the target moment in the reference segmented area and extend the preset duration in the same way to form the second analysis period. By calculating the data sequences of the pulse pressure waveforms within the first analysis period and the second analysis period, and using the Pearson correlation coefficient formula for similarity comparison, the Pearson correlation coefficient value at the target moment is finally obtained. This value is used to quantify the consistency of the pressure changes in the two segmented areas in the time dimension.

[0109] Through this embodiment, the correlation differences in pressure changes of different optical bands in the same pulse wave signal can be effectively identified, thereby eliminating the interference of waveform distortion in segmented regions caused by limb movement. By dynamically analyzing the pressure correlation in a local time period, the recognition accuracy of noise signals is improved, providing a reliable basis for subsequent screening of high-quality training data, and ultimately reducing the error introduced by motion artifacts in the blood oxygen detection model.

[0110] In some of the above solutions of the present invention, there may be problems of quality differences or uneven distribution in the samples of the training dataset. Directly using the untreated training data may reduce the model training efficiency, resulting in a slow convergence speed or getting stuck in a local optimum during the model parameter adjustment process, affecting the final performance of the blood oxygen detection model.

[0111] In response to this, the present invention further proposes that S105 may specifically include: Dividing the target training dataset into multiple training batches; Performing data augmentation processing and data preprocessing on the first training data samples in each training batch to obtain second training data samples; According to the priority order of the training batches, sequentially adjusting the model parameters of the target model based on the second training data samples in the training batches until the preset training end condition is met, obtaining the blood oxygen detection model.

[0112] In this embodiment, the division of the training dataset can be based on the number of samples or the time sequence, and each training batch contains the same or different numbers of samples. The data augmentation processing may include amplitude scaling, time stretching or adding noise to the pulse wave signal, and the data preprocessing may include normalization, filtering or outlier removal. The determination of the priority order can be based on the average value of the light absorption abnormality, pulse wave error degree or training priority coefficient of the samples in the training batch, and the batches with higher priority participate in the training first. The preset training end condition can be set as the loss function dropping to a threshold or the number of training rounds reaching the upper limit. Modularizing the training process by dividing batches, data augmentation and preprocessing improve the data quality, and adjusting the priority order optimizes the parameter update path.

[0113] Specifically, after the training dataset is divided into multiple batches, each batch is independently subjected to data augmentation and preprocessing. For example, random amplitude changes are applied to the infrared and visible light pulse wave signals respectively to simulate the effects of different wearing tightness, or high-frequency motion interference is eliminated through sliding window filtering. The preprocessed second training data samples are input into the model in the order of batch priority. Batches with higher priority may contain samples with lower light absorption abnormality, and prior training can accelerate the model's capture of core features. The model parameters are updated after each batch of training and gradually optimized through multiple rounds of iteration until the loss function stabilizes or reaches the preset number of rounds. The finally obtained blood oxygen detection model can more accurately process pulse wave signals of different qualities, improving the robustness of blood oxygen detection calculation.

[0114] As an example, the training dataset is divided into six training batches containing 300 groups of samples. The original samples in each training batch are first subjected to data augmentation by adding Gaussian noise, with the noise amplitude controlled within 5% of the peak-to-peak value of the pulse wave signal. At the same time, normalization processing is used to map the signal amplitude to the range [-1, 1]. In the preprocessing stage, the enhanced data is subjected to Butterworth band-pass filtering, with the passband range set to 0.5 Hz - 8 Hz, and the baseline drift elimination operation is performed. The priority of the training batches is sorted by calculating the average pulse wave signal-to-noise ratio of each batch of samples. Batches with a signal-to-noise ratio higher than 35 dB are marked as high priority. During the model training process, high-priority batch data is preferentially loaded in each training cycle. An adaptive learning rate adjustment strategy is adopted, with the initial learning rate set to 0.001. When the validation set loss does not decrease for three consecutive cycles, the learning rate decay coefficient is adjusted to 0.5. The model parameters are subjected to gradient clipping once every three training cycles of iteration update, with the threshold set to 1.0, until the validation set accuracy remains stable for ten consecutive training cycles, at which point the training process is terminated.

[0115] Through this embodiment, the problem of slow model convergence speed caused by unbalanced training data distribution is effectively solved, and the model parameter update path is optimized by dynamically adjusting the training order. The data augmentation and preprocessing operations significantly reduce the interference of motion artifacts on pulse wave feature extraction, and the adaptive learning mechanism enables the model to automatically adapt to the feature distribution differences of different quality data during the training process. The introduction of the gradient clipping strategy effectively avoids the phenomenon of gradient explosion that may occur during the training process, thus ensuring that the blood oxygen detection model finally has stable calculation accuracy and generalization performance.

[0116] It should be noted that the above sequence of embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or consecutive order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0117] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other, and the differences between each embodiment and other embodiments are emphasized.

Claims

1. A blood oxygen detection and calculation method based on infrared light and visible light, characterized in that, The method includes: Obtaining physiological index information and pulse wave detection information of each bracelet wearer, where the physiological index information includes at least one of body fat information and age information, and the pulse wave detection information includes an infrared light pulse wave signal and a visible light pulse wave signal; Based on the physiological index information, clustering each of the pulse wave detection information to obtain a plurality of pulse wave detection groups; Based on the infrared light pressure value and the visible light pressure value of each of the pulse wave detection information in the pulse wave detection group, determining the light absorption abnormality degree of each of the pulse wave detection information, where the light absorption abnormality degree is used to characterize the deviation abnormality degree between the pulse wave detection information and the corresponding pulse wave detection group; Based on each of the light absorption abnormality degrees, screening out a target training data set from each of the pulse wave detection groups; Based on the target training data set, training a target model to obtain a blood oxygen detection model, so as to implement blood oxygen detection calculation through the blood oxygen detection model.

2. The blood oxygen detection and calculation method based on infrared light and visible light according to claim 1, wherein The determining the light absorption abnormality degree of each of the pulse wave detection information based on the infrared light pressure value and the visible light pressure value of each of the pulse wave detection information in the pulse wave detection group includes: Performing mean processing on each of the infrared light pressure values and each of the visible light pressure values of each of the pulse wave detection information in the pulse wave detection group to obtain an infrared light pressure mean value and a visible light pressure mean value; Using the difference between the infrared light pressure value of the target pulse wave detection information and the infrared light pressure mean value, and the difference between the visible light pressure value and the visible light pressure mean value, to determine the light absorption abnormality degree of the target pulse wave detection information, where the target pulse wave detection information is any one of the pulse wave detection information.

3. The blood oxygen detection and calculation method based on infrared light and visible light according to claim 1, wherein Before the screening out a target training data set from each of the pulse wave detection groups based on each of the light absorption abnormality degrees, the method further includes: For the infrared light pulse wave signal and the visible light pulse wave signal of the pulse wave detection information, determining a section area between two adjacent valley points as a section area, to obtain a plurality of first section areas of the infrared light pulse wave signal and a plurality of second section areas of the visible light pulse wave signal; For a target section area, determining a motion interference parameter of the target section area based on the time span and pulse pressure of the target section area, where the target section area is any one of the first section areas and the second section areas; Using each of the motion interference parameters of the pulse wave detection information and the light absorption abnormality degree of the pulse wave detection information to determine the pulse wave error degree of the pulse wave detection information; Using the pulse wave error degree of the pulse wave detection information to determine the training priority coefficient of the pulse wave detection information; The screening out a target training data set from each of the pulse wave detection groups based on each of the light absorption abnormality degrees includes: Based on each of the training priority coefficients, screening out a target training data set from each of the pulse wave detection groups.

4. The blood oxygen detection and calculation method based on infrared light and visible light according to claim 3, characterized in that The pulse pressure includes a pulse DC component and a pulse AC component; Determining the motion interference parameter of the target segmented region based on the time span and pulse pressure of the target segmented region includes: Determining the periodic law presentation degree of the target segmented region based on the time span of the target segmented region; Determining the motion interference tendency degree of the target segmented region based on the DC component and AC component of the pulse at each moment in the target segmented region; Determining the motion interference parameter of the target segmented region by using the periodic law presentation degree and the motion interference tendency degree of the target segmented region.

5. The blood oxygen detection and calculation method based on infrared light and visible light according to claim 4, characterized in that, The determining the periodic law presentation degree of the target segmented region based on the time span of the target segmented region includes: Obtaining a first time span mean value corresponding to the segmented region in the target pulse wave signal and a second time span mean value corresponding to the segmented region in the reference pulse wave signal, where the target pulse wave signal is the pulse wave signal to which the target segmented region belongs, and the reference pulse wave signal is the pulse wave signal corresponding to the pulse wave detection information except the target pulse wave signal; Determining the first periodic feature conformity degree of the target segmented region by using the time span of the target segmented region and the first time span mean value; Determining the second periodic feature conformity degree of the reference segmented region by using the time span of the reference segmented region and the second time span mean value, where the reference segmented region is the segmented region corresponding to the target segmented region in the reference pulse wave signal; Determining the periodic law presentation degree of the target segmented region based on the first periodic feature conformity degree and the second periodic feature conformity degree.

6. The blood oxygen detection and calculation method based on infrared light and visible light according to claim 5, wherein The determining the periodic law presentation degree of the target segmented region based on the first periodic feature conformity degree and the second periodic feature conformity degree includes: Taking the absolute value after subtracting the second periodic feature conformity degree from the first periodic feature conformity degree to obtain a first calculated value; Performing a normalization process on the reciprocal of the first calculated value to obtain a second calculated value; Determining the periodic law presentation degree of the target segmented region by using the second calculated value and the first periodic feature conformity degree.

7. The blood oxygen detection and calculation method based on infrared light and visible light according to claim 4, characterized in that The determining the motion interference tendency degree of the target segmented region based on the DC component and AC component of the pulse at each moment in the target segmented region includes: Respectively determining the component transfer interference degree at each moment in the target segmented region based on the DC component and AC component of the pulse at each moment in the target segmented region; Obtaining the Pearson correlation coefficient of the pulse pressure at each moment in the target segmented region and the reference segmented region, where the reference segmented region is the segmented region corresponding to the target segmented region in the reference pulse wave signal, the reference pulse wave signal is the pulse wave signal corresponding to the pulse wave detection information except the target pulse wave signal, and the target pulse wave signal is the pulse wave signal to which the target segmented region belongs; Determining the motion interference tendency degree of the target segmented region by using each component transfer interference degree and each Pearson correlation coefficient.

8. The blood oxygen detection calculation method based on infrared light and visible light according to claim 7, wherein Determining the component transfer interference degree at each moment in the target segmented region based on the DC component and the AC component of the pulse at each moment in the target segmented region respectively, includes: Obtaining a first slope of the DC component of the pulse and a second slope of the AC component of the pulse at a target moment in the target segmented region, where the target moment is any moment in the target segmented region; Performing normalization processing on the ratio of the first slope to the second slope to obtain the component transfer interference degree at the target moment in the target segmented region.

9. The blood oxygen detection and calculation method based on infrared light and visible light according to claim 7, characterized in that, The obtaining of the Pearson correlation coefficient of the pulse pressure at each moment in the target segmented region and the reference segmented region includes: Taking the target moment in the target segmented region as the center and extending a target duration to both sides respectively to obtain a first analysis period, where the target moment is any moment in the target segmented region; Taking the target moment in the reference segmented region as the center and extending a target duration to both sides respectively to obtain a second analysis period; Comparing the first analysis period with the second analysis period to obtain the Pearson correlation coefficient of the pulse pressure in the target segmented region and the reference segmented region at the target moment.

10. The blood oxygen detection and calculation method based on infrared light and visible light according to any one of claims 1-9, characterized in that, Training a target model based on the target training dataset to obtain a blood oxygen detection model, includes: Dividing the target training dataset into multiple training batches; Performing data augmentation processing and data preprocessing on the first training data samples in each of the training batches to obtain second training data samples; Adjusting the model parameters of the target model successively based on the second training data samples in the training batches according to the priority order of the training batches until a preset training end condition is met to obtain the blood oxygen detection model.

Images