Biometric value prediction method

Abstract

The present invention relates to a method for predicting a biometric value in a blood glucose measurement system and, more particularly, to a biometric value prediction method capable of predicting a future biometric value of a user by generating a predictive model through a communication terminal having a small memory and amount of calculations, such as a smartphone that the user always carries to manage a biometric value, and applying the biometric value of the user to the generated predictive model, and capable of predicting a future biometric value of the user without requiring biometric information of other nearby users and without access to a server, by generating a predictive model personalized for the user on the basis of biometric history information of the user.

Description

Methods for predicting biometric values

[0001] The present invention relates to a method for predicting a biometric value in a blood glucose measurement system, and more specifically, to a method for predicting a biometric value, which generates a prediction model through a communication terminal with a small memory and computational capacity, such as a smartphone, which the user always carries with him or her to manage his or her biometric value, and applies the user's biometric information to the generated prediction model to predict the user's future biometric value, and generates a prediction model personalized for the user based on the user's biometric history information, thereby predicting the user's future biometric value without needing biometric information of other surrounding users and without accessing a server.

[0002] Diabetes is a chronic disease that is common among modern people, affecting more than 2 million people in Korea, or 5% of the total population.

[0003] Diabetes occurs when the insulin produced by the pancreas is absolutely or relatively insufficient due to various causes such as obesity, stress, poor eating habits, and congenital genetics, and thus the blood sugar level becomes extremely high due to the inability to correct the blood sugar balance.

[0004] Blood normally contains a certain concentration of glucose, and tissue cells obtain energy from it.

[0005] However, if glucose increases more than necessary, it cannot be properly stored in the liver, muscles, or fat cells, but accumulates in the blood, and as a result, diabetic patients maintain blood sugar levels that are much higher than those of normal people, and as the excessive blood sugar passes through the tissues and is excreted in the urine, the sugar absolutely necessary for each tissue of the body becomes insufficient, causing abnormalities in each tissue of the body.

[0006] Diabetes is characterized by having almost no symptoms in the early stages, but as the disease progresses, it causes specific symptoms such as diabetes-specific polydipsia, polyuria, weight loss, general malaise, itchy skin, and long-lasting wounds on the hands and feet that do not heal. As the disease progresses further, complications such as vision impairment, high blood pressure, kidney disease, stroke, periodontal disease, muscle cramps and neuralgia, and gangrene appear.

[0007] To diagnose this type of diabetes and manage it so that it does not progress to complications, systematic blood sugar measurement and treatment must be performed in parallel.

[0008] Diabetes requires regular blood sugar monitoring for management, and demand for blood sugar monitoring devices is steadily increasing. Various studies have shown that when diabetic patients strictly control their blood sugar, the incidence of diabetic complications is significantly reduced. Therefore, it is crucial for diabetic patients to regularly monitor their blood sugar levels to maintain good blood sugar control.

[0009] Finger prick glucose meters are commonly used to manage blood sugar levels in diabetic patients. While these devices can help manage blood sugar levels in diabetic patients, they only display results at the time of measurement, making it difficult to accurately assess frequently fluctuating blood sugar levels. Furthermore, finger prick glucose meters require frequent blood draws to measure blood sugar levels throughout the day, placing a significant burden on patients.

[0010] Diabetics typically experience fluctuations between hyperglycemia and hypoglycemia, with hypoglycemia being a major emergency. Hypoglycemia occurs when blood sugar levels are depleted over a prolonged period and can lead to loss of consciousness or, in the worst case, death. Therefore, promptly detecting hypoglycemia is crucial for diabetics. However, blood glucose meters, which measure blood sugar levels intermittently, have clear limitations.

[0011] To overcome the limitations of these blood glucose meters, a continuous glucose monitoring system (CGMS) was developed that is inserted into the body and measures blood glucose levels at intervals of several minutes, making it easier to manage diabetes patients and respond to emergencies.

[0012] A continuous blood glucose monitoring system includes a body attachment unit that is inserted into a human body and measures blood glucose using a body fluid such as a user's blood, and a communication terminal that communicates with the body attachment unit and displays the blood glucose level measured by the body attachment unit.

[0013] The body attachment unit is composed of a sensor that is inserted into the body for a certain period of time, for example, within 3 months (7 weeks, 15 days, 1 month, etc.) and generates a biosignal representing the user's blood sugar level from body fluid, and a transmitter that transmits the biosignal received from the sensor to a communication terminal in real time, periodically, or upon request from the communication terminal.

[0014] The body-attached unit generates bio-signals and transmits them to a communication terminal. The communication terminal has a blood sugar management application installed, receives bio-signals from the body-attached unit, and goes through preprocessing such as noise removal from the received bio-signals, unit correction of the bio-signals of the received current values ​​into blood sugar values, and correction using the reference blood sugar value, and outputs the bio-signals so that the user can check them.

[0015] In this way, the continuous blood sugar measurement system not only measures the user's bio-signals in real time and outputs the user's blood sugar level, but also predicts the user's future blood sugar level based on the user's blood sugar level history, activity history, meal history, and history of medication administration such as insulin.

[0016] Predicting a user's future blood sugar level can be done in various ways. A representative example is to create a prediction model based on the user's blood sugar level history and the blood sugar level history of other users around them, and then apply the user's current blood sugar level to the created prediction model to predict the user's future blood sugar level.

[0017] This technology for predicting future blood sugar levels of users has the effect of predicting the possibility of the user falling into hypoglycemia or hyperglycemia in advance, allowing the user to take action in advance before the user actually reaches a hypoglycemia or hyperglycemia situation.

[0018] However, since the prediction model conventionally used to predict future blood sugar levels is created using a large amount of blood sugar history information, meal history information, activity history, etc. of not only the user but also other surrounding users, a very large memory space is required to store the information required to create the prediction model, and a large amount of computation is required to predict future blood sugar levels by applying the user's bio-signals to the prediction model.

[0019] Therefore, the creation and modification of these prediction models and the prediction of future blood sugar levels using the prediction models must be performed primarily through a separate server, and there is a problem in that the communication terminal must always communicate with the server to predict future blood sugar levels.

[0020] The present invention is intended to solve the problems of the conventional future blood sugar value prediction method mentioned above, and the purpose of the present invention is to provide a bio-value prediction method that can predict the user's future blood sugar value with a small amount of memory and computation through a communication terminal such as a smartphone that the user always carries to manage his / her blood sugar level.

[0021] Another object of the present invention is to provide a method for generating a personalized prediction model for a user based on the user's biometric information and predicting the user's future biometric values ​​without accessing a server.

[0022] Another object of the present invention is to provide a method for predicting biometric values, which can accurately predict a user's future biometric values ​​by calculating a prediction error from a user's predicted biometric values ​​and actual biometric values, determining the expression characteristics of the prediction error, and retraining or regenerating a prediction model according to the expression characteristics of the prediction error.

[0023] Another object of the present invention is to provide a method for generating a personalized prediction model for a user by using a first feature value extracted from preprocessed biometric information and a second feature value extracted from biometric information that has been time-corrected and unit-corrected from the preprocessed biometric information, and for accurately predicting the biometric value of the user through the generated prediction model.

[0024] In order to achieve the object of the present invention, a method for predicting a user's biometric value using biometric information measured from a sensor of a body-attached unit is characterized by including the steps of: extracting a first feature value from the measured user's biometric information; correcting the measured user's biometric information and extracting a second feature value from the corrected biometric information; reducing and combining the first and second feature values ​​to generate a feature vector value; and applying the generated feature vector value to a prediction model to predict the user's biometric value.

[0025] Here, the sensor is characterized as a sensor that is inserted into the user's body for a certain period of time and measures the user's biometric information.

[0026] Preferably, the method for predicting biometric information according to the present invention further includes a step of preprocessing the measured biometric information by removing noise from the measured biometric information, and the first feature value and the second feature value are extracted from the preprocessed biometric information.

[0027] Here, the first feature value is extracted directly from the preprocessed biometric information, and the second feature value is extracted from the corrected biometric information generated by correcting the preprocessed biometric information for time delay and unit mismatch.

[0028] Here, the unit mismatch is characterized by being corrected based on preprocessed biometric information or reference biometric values.

[0029] In one embodiment of the present invention, unit mismatch is characterized in that the preprocessed biometric information is corrected by giving weight when it rises or falls.

[0030] In another embodiment of the present invention, the unit mismatch is characterized in that it is corrected by a weight assigned according to the difference between the biometric value determined from the measured biometric information and the reference biometric value.

[0031] Preferably, the method for predicting a biometric value according to the present invention is characterized by further comprising a step of calculating a prediction error from the difference between the predicted biometric value at a first prediction time point and the biometric value actually measured at the first prediction time point, and a step of determining whether to retrain the prediction model based on the prediction error.

[0032] Here, it is characterized by deciding to retrain the prediction model if the prediction error is greater than a threshold or a critical ratio.

[0033] In the method for predicting biometric values ​​according to the present invention, when it is decided to retrain the prediction model, the prediction model is characterized in that it is retrained using a subsequent data set generated from the user's biometric information measured up to the present point in time in addition to the previous data set used to generate the prediction model.

[0034] Preferably, the method for predicting a biometric value according to the present invention is characterized by further including a step of determining whether to regenerate a prediction model based on the expression characteristics of a prediction error over a unit time.

[0035] Here, the expression characteristic is characterized by at least one of the number of times the prediction error continuously exceeds a threshold or a critical ratio during a unit time and the total number of times the prediction error exceeds a threshold or a critical ratio during a unit time.

[0036] The method for predicting biometric values ​​according to the present invention has the following effects.

[0037] First, the method for predicting biometric values ​​according to the present invention can predict the user's future biometric values ​​by creating a prediction model through a communication terminal with a small memory and computational capacity, such as a smartphone, which the user always carries with him or her to manage the biometric values, and applying the user's biometric information to the created prediction model.

[0038] Second, the method for predicting biometric values ​​according to the present invention generates a prediction model personalized for the user based on the user's biometric history information, thereby predicting the user's future biometric values ​​without requiring biometric information from other surrounding users and without accessing a server.

[0039] Third, the method for predicting biometric values ​​according to the present invention calculates a prediction error from the user's predicted biometric value and the actual biometric value, determines the expression characteristics of the prediction error, and retrains or regenerates the prediction model according to the expression characteristics of the prediction error, thereby accurately predicting the user's future biometric value.

[0040] Fourth, the method for predicting biometric values ​​according to the present invention generates a personalized prediction model for a user by using a first feature value extracted from preprocessed biometric information and a second feature value extracted from corrected biometric information obtained by time-correcting and unit-correcting the preprocessed biometric information, thereby enabling accurate prediction of the user's biometric value through the generated prediction model.

[0041] FIG. 1 is a schematic diagram illustrating a blood sugar measurement system according to one embodiment of the present invention.

[0042] Figure 2 is a functional block diagram for explaining a biometric value prediction device according to the present invention.

[0043] Figure 3 is a functional block diagram illustrating an example of a feature value generation unit according to the present invention.

[0044] Figure 4 is a functional block diagram illustrating an example of a learning unit according to the present invention.

[0045] Figure 5 is a flowchart for explaining a method for predicting biometric values ​​according to the present invention.

[0046] Figure 6 is a flowchart illustrating an example of a step of retraining a prediction model in the present invention.

[0047] Figure 7 is a flowchart illustrating an example of a step of regenerating a prediction model in the present invention.

[0048] Figure 8 is a drawing for explaining an example of time delay compensation.

[0049] Figure 9 is a drawing for explaining an example of a method for correcting unit mismatch.

[0050] Figure 10 is a drawing illustrating another example of a method for correcting unit mismatch.

[0051] Figures 11 and 12 are diagrams for explaining an example of determining whether to retrain a prediction model.

[0052] Figure 13 is a diagram illustrating an example of regenerating a prediction model.

[0053] It should be noted that the technical terms used in the present invention are used merely to describe specific embodiments and are not intended to limit the present invention. Furthermore, unless specifically defined otherwise herein, the technical terms used herein should be interpreted in the sense generally understood by those skilled in the art to which the present invention pertains, and should not be interpreted in an overly comprehensive or overly narrow sense. Furthermore, if the technical terms used herein are incorrect and do not accurately express the spirit of the present invention, they should be replaced with technical terms that can be correctly understood by those skilled in the art.

[0054] Additionally, singular expressions used in the present invention include plural expressions unless the context clearly dictates otherwise. In the present invention, terms such as "consist of" or "include" should not be construed to necessarily include all of the components or steps described in the invention, and should be construed to mean that some of the components or steps may not be included, or that additional components or steps may be included.

[0055] In addition, it should be noted that the attached drawings are only intended to facilitate easy understanding of the spirit of the present invention, and should not be construed as limiting the spirit of the present invention by the attached drawings.

[0056]

[0057] The method for predicting biometric values ​​according to the present invention will be described in more detail with reference to the attached drawings below.

[0058] Hereinafter, a continuous blood sugar measurement system is described that continuously measures biometric information indicating blood sugar levels through a body-attached unit attached to a user's body for a certain period of time and transmits the measured biometric information to a communication terminal. However, depending on the field to which the present invention is applied, the body-attached unit can measure various types of biometric information and transmit the measured biometric information to a communication terminal, and this falls within the scope of the present invention.

[0059]

[0060] FIG. 1 is a schematic diagram illustrating a blood sugar measurement system according to one embodiment of the present invention.

[0061] Referring to FIG. 1, a blood glucose measurement system (1) according to one embodiment of the present invention includes a body attachment unit (10) and a communication terminal (30).

[0062] The body attachment unit (10) is attached to the body, and when the body attachment unit (10) is attached to the body, one end of the sensor of the body attachment unit (10) is inserted into the skin and continuously measures biometric information indicating the user's blood sugar level using body fluids, etc. during the period of use of the sensor.

[0063] The communication terminal (30) is a terminal that can receive biometric information from the body attachment unit (10) and display the received biometric information to the user. For example, a mobile terminal capable of communicating with the body attachment unit (10), such as a smartphone, tablet PC, or laptop, can be used. Of course, the communication terminal (13) is not limited thereto, and any type of terminal can be used as long as it includes a communication function and can have a program or application installed.

[0064] The body attachment unit (10) transmits measured biometric information to the communication terminal (30) at the request of the communication terminal (30) or at set times. For data communication between the body attachment unit (10) and the communication terminal (30), the body attachment unit (10) and the communication terminal (30) may be connected to each other by wire using a USB cable or the like, or may be connected to each other by wireless communication methods such as infrared communication, NFC communication, or Bluetooth.

[0065] The communication terminal (30) predicts the user's future biometric values ​​based on the received biometric information and provides the predicted biometric values ​​to the user. Preferably, the communication terminal (30) may provide the user with a hyperglycemic or hypoglycemic alarm based on the predicted biometric values, or may provide the user with a necessary prescription along with a hypoglycemic or hypoglycemic alarm.

[0066] Here, the communication terminal (30) can store the received biometric information for a certain period of time, and the communication terminal (30) can create a prediction model using the stored biometric information.

[0067] In addition, the communication terminal (30) monitors and compares predicted biometric values ​​for a certain time in the future with actual biometric values ​​measured after a certain time has elapsed, and can relearn or regenerate a prediction model created based on a prediction error between the predicted biometric values ​​and the actual biometric values ​​or based on the expression characteristics of the prediction error.

[0068]

[0069] Figure 2 is a functional block diagram for explaining a biometric value prediction device according to the present invention.

[0070] The biometric value prediction device according to the present invention can be implemented in a communication terminal. Referring to FIG. 2, a transceiver (110) receives the user's biometric information from a body-attached unit, and a storage unit (130) stores the received biometric information. Here, the storage unit (130) can store the received biometric information by mapping it to the time of reception.

[0071] The biometric value determination unit (180) determines the user's biometric value from the received biometric information and outputs the determined biometric value to the user or stores it in the storage unit (130) by mapping it with the reception time.

[0072] Here, various types of biometric information can be measured through sensors inserted into the user's body for a certain period of time. Below, an example of biometric information may be biometric information indicating the user's blood sugar information, and the biometric value may be a blood sugar value determined from the biometric information.

[0073] The storage unit (130) can store biometric values ​​determined from biometric information together with biometric information received from a body attachment unit, and the feature value generation unit (150) generates a feature value vector used to determine a predicted biometric value at a future point in time using the biometric information.

[0074] The prediction unit (170) uses the prediction model stored in the storage unit (130) to determine the predicted biometric value after a certain period of time has elapsed based on the current point in time. The prediction unit (170) applies the generated feature value vector to the prediction model to determine the predicted biometric value. Here, the predicted biometric value is a predicted biometric value that the user will have after a certain period of time has elapsed based on the current point in time. Using the predicted biometric value, the user's biometric value can be predicted in advance after a certain period of time has elapsed, and an alarm can be provided to the user or a necessary prescription can be provided based on the predicted biometric value.

[0075] For example, if a user's predicted blood sugar level is determined to be 60 mg / dL or lower 3 hours from the current time, a low blood sugar alarm can be provided to the user or a prescription can be provided to instruct the user to eat food.

[0076] Meanwhile, the storage unit (130) stores the predicted biometric value determined by the prediction unit (170) and the actual biometric value determined by the biometric value determination unit (180), and the learning unit (190) monitors and compares the actual biometric value determined after a certain period of time with the predicted biometric value of the user predicted before a certain period of time.

[0077] The learning unit (190) can relearn the prediction model stored in the storage unit (130) based on the size or ratio of the prediction error between the predicted biometric value and the actual biometric value, or can regenerate a new prediction model by relearning the prediction model from the beginning based on the expression characteristics of the prediction error, and can update the prediction model stored in the storage unit (130) with the newly generated prediction model.

[0078]

[0079] Figure 3 is a functional block diagram illustrating an example of a feature value generation unit according to the present invention.

[0080] Looking more specifically at Fig. 3, the preprocessing unit (151) removes or reduces noise from biometric information received through the transceiver unit.

[0081] Biometric data measured through or received from a body-attached unit may contain noise. For example, because the sensors of a body-attached unit are partially implanted into the human body, they may move as the person moves. This movement of these sensors may introduce noise into the biometric data measured by the body-attached unit.

[0082] Or, when biometric information is transmitted from a body attachment unit to a communication terminal, it may be affected by surrounding electromagnetic waves, etc., and as a result, noise may be included in the biometric information received by the communication terminal.

[0083] The preprocessing unit (151) performs an outlier processing filtering process on the received biometric information or performs a low-pass filtering process on the biometric information for which outlier processing has been completed.

[0084] The first feature value extraction unit (153) extracts the first feature value from the preprocessed biometric information, and the second feature value extraction unit (157) extracts the second feature value from the corrected biometric information corrected by the correction unit (155). The preprocessed biometric information received from the body attachment unit is current value information that changes differently depending on the biometric value of the user, and the corrected biometric information means biometric information that has been time-delay corrected or unit-mismatch corrected for the biometric information preprocessed by the correction unit (155).

[0085] Preferably, the correction unit (155) corrects the preprocessed biometric information and provides the corrected biometric information to the second feature value extraction unit (157). The correction unit corrects the time delay of the preprocessed biometric information or corrects the biometric information of the current value into a unit of the biometric value, for example, a unit of the blood sugar value.

[0086] The feature vector value generation unit (159) generates a feature vector value by simply combining the first feature value and the second feature value or by reducing and combining the first feature value and the second feature value through a method such as resource reduction.

[0087]

[0088] Figure 4 is a functional block diagram illustrating an example of a learning unit according to the present invention.

[0089] Looking more specifically with reference to Fig. 4, the prediction error calculation unit (191) monitors the actual biometric value determined after a certain period of time has elapsed and the predicted biometric value of the user predicted in advance a certain period of time, compares them with each other, and calculates the prediction error between the actual biometric value and the predicted biometric value. For example, the predicted biometric value after t time has elapsed based on the current time is determined from the feature vector value, and the prediction error between the actual biometric value determined using the biometric information after t time has elapsed is calculated.

[0090] The prediction error characteristic judgment unit (195) judges the occurrence characteristics of the prediction error, such as the number of times the prediction error exceeds a threshold value or a critical ratio during a unit time, or the number of times the prediction error continuously exceeds a threshold value or a critical ratio during a unit time.

[0091] The condition judgment unit (193) determines the relearning condition based on whether the prediction error exceeds a set threshold or whether the prediction error ratio compared to the actual biometric value or predicted biometric value exceeds a set threshold ratio, or determines whether the regeneration condition of the prediction model is satisfied based on the expression characteristics of the prediction error.

[0092] The relearning unit (197) relearns the prediction model stored in the storage unit when the relearning condition is satisfied and updates the prediction model with the relearned prediction model, and the generation unit (199) creates a new prediction model stored in the storage unit when the regeneration condition is satisfied and updates the prediction model with the newly created prediction model.

[0093]

[0094] Figure 5 is a flowchart for explaining a method for predicting biometric values ​​according to the present invention.

[0095] Referring to Figure 5, a more detailed description is given of receiving the user's biometric information (S110). Here, the user's biometric information can be received from a body-attached unit that is attached to the user's body for a certain period of time and continuously measures the user's biometric information.

[0096] The communication terminal preprocesses the received biometric information (S130). More specifically, the preprocessing process involves searching for data outside a predetermined range among the received biometric information for singular value processing filtering, and processing the corresponding biometric information. If the biometric information is determined to contain singular values, the biometric information may be removed and processed. However, this is not limited thereto, and the biometric information with singular values ​​may be corrected and utilized as needed. Low-pass filtering may be performed on the singular value-processed biometric information. Low-pass filtering can remove components corresponding to high bandwidths and retain only biometric information corresponding to low bandwidths. The low-pass filtered biometric information can be used to calculate and process the average value of the low-pass filtered biometric information, and a truncated average value may be utilized. Various preprocessing methods are possible depending on the field to which the present invention is applicable, and such methods fall within the scope of the present invention.

[0097] A first feature value is extracted from the preprocessed biometric information (S140). The preprocessed biometric information may be a user's biometric value measured by a body-attached unit, for example, a current value representing blood sugar levels. Here, the first feature value is a feature value such as difference, slope, deviation, mean, root mean square, sharpness, etc. of the biometric information extracted using statistical techniques, or a feature value extracted by analyzing the biometric information in the frequency domain, such as Fourier transform or wavelet transform.

[0098] Meanwhile, preprocessed biometric information is corrected to generate corrected biometric information (S150), and a second feature value is extracted from the corrected biometric information (S150). Here, the preprocessed biometric information can correct for time delays or unit mismatches contained in the preprocessed biometric information before extracting the second feature value.

[0099] There is a time delay between the time when the user's biometric information is actually measured by the body attachment unit and the time when the user's biometric information is received by the communication terminal. This time delay may occur depending on the physical structure of the body attachment unit's sensor until it measures and generates biometric information from the user's body fluids, or may occur due to the computational time required to generate biometric information.

[0100] In addition, since the preprocessed biometric information is current value information representing the user's biometric value and is not a biometric value itself, the unit discrepancy must be corrected from the current value to the actual user's biometric value. The unit discrepancy can be corrected using the reference biometric value measured through a blood collection type biometric value measuring device using a separate sensor strip (blood test strip) and collected blood fluid. For example, if the current value of the biometric information is 10 nA and the reference biometric value measured at this time is 100 mg / dL, the correction slope (A) is set to 10, and the current value of the biometric information is then multiplied by the slope to determine the user's biometric value.

[0101] Depending on the field to which the present invention is applied, the unit mismatch can be corrected by assigning weights depending on whether the biometric value is increasing or decreasing or the difference between the measured biometric value and the reference biometric value.

[0102] Second feature values ​​are extracted using corrected biometric information (S160). Here, the second feature values ​​are feature values ​​such as difference, slope, deviation, mean, root mean square root, and sharpness of corrected biometric information extracted from statistical techniques, or feature values ​​extracted by frequency domain analysis of corrected biometric information, such as Fourier transform or wavelet transform.

[0103] The first feature value and the second feature value are simply combined or resource-reduced to generate a feature vector value from the first feature value and the second feature value (S170).

[0104] The generated feature vector value is applied to the prediction model to generate a predicted biometric value after a certain period of time (S190).

[0105] The prediction model is used to generate predicted biometric values ​​after a certain period of time using feature vector values. For example, if the number of components constituting the feature vector values ​​(x) generated at certain intervals is 10, and the label (y) for each feature vector value is the predicted biometric value after a certain period of time, the input (x) and label (y) of each feature vector value are expressed as in the following mathematical equation (1).

[0106] [Mathematical Formula 1]

[0107]

[0108] Based on this, the feature vector value generated from feature value 1 and feature value 2 If you input it into the prediction model, it outputs the predicted biometric value y' after a certain period of time.

[0109] Here, the prediction model can be generated using machine learning such as SVM (Support Vector Machine) and GMM (Gaussian Mixture Model) using the user's feature value vector as training data, deep learning such as CNN (Convolution Neutal Network) and RNN (Recurrent Neural Network), reinforcement learning such as model-free RL and model-based RL, and deep reinforcement learning such as DQN (Deep Q Network). The prediction model generated in the present invention can be retrained using additional or new training data. Various known techniques can be used to generate the prediction model, and a detailed description thereof is omitted.

[0110] The first and second features have a large number of features. If you try to create a predictive model using this data, the learning speed will be slow due to the large dimensionality of the data, and the performance will likely be poor. To this end, you can create a predictive model by simply combining the first and second features, or by selecting or reducing features among them. Methods for reducing the data dimensionality, such as projection and manifold learning, as well as representative dimensionality reduction algorithms such as principal component analysis (PCA), or feature selection algorithms such as Lasso can be used. Various known techniques can be used to generate feature vector values ​​from the first and second features, and a detailed description of these techniques is omitted.

[0111] In the present invention, by generating a prediction model using both the first feature value and the second feature value generated from the user's biometric information, a prediction model that is personalized for the user and can accurately predict the user's biometric value can be generated without using data from other users around the user, and by applying the feature vector value generated from the first feature value and the second feature value to the prediction model, the user's biometric value can be accurately predicted.

[0112]

[0113] Figure 6 is a flowchart illustrating an example of a step of retraining a prediction model in the present invention.

[0114] Looking more specifically at Fig. 6, the prediction error is calculated from the difference between the predicted biometric value at the first prediction time point generated using the feature vector value and the actual biometric value measured using the actual biometric value at the first prediction time point (S211).

[0115] The prediction error is compared with a preset threshold value or critical ratio to determine whether the prediction error exceeds the threshold value or critical ratio (S213).

[0116] If the prediction error exceeds the threshold value or the threshold ratio, it is determined whether the relearning requirement is satisfied (S215). Here, the relearning requirement may be determined to be satisfied if the prediction error exceeds the threshold value or the threshold ratio, or if the prediction error exceeds the threshold value or the threshold ratio and the average value of the prediction error for a preset first time period thereafter exceeds the threshold average value, or if the prediction error exceeds the threshold value or the threshold ratio and the prediction error exceeds the threshold value or the threshold ratio for a second time period thereafter.

[0117] If the retraining requirements are met, the prediction model is retrained and the previously used prediction model is updated with the retrained prediction model (S217). The retraining of the prediction model is characterized by using a subsequent data set generated from the user's biometric information measured up to the current point in time, in addition to the previous data set used to generate the prediction model stored in the storage unit.

[0118] Here, the data set is characterized by a feature vector value generated from the first feature value and the second feature value.

[0119]

[0120] Figure 7 is a flowchart illustrating an example of a step of regenerating a prediction model in the present invention.

[0121] Referring to Figure 7, a more detailed examination will be made of the prediction error expression characteristics (S231). Here, the prediction error expression characteristics may include the average value of the prediction error per unit time, the number of times the prediction error exceeds a threshold value or a critical ratio per unit time, the number of times the prediction error continuously exceeds a threshold value or a critical ratio per unit time, and the ratio of the time the prediction error exceeds a threshold value or a critical ratio relative to the unit time.

[0122] Based on the expression characteristics of the prediction error, it is determined whether the regeneration condition of the prediction model is satisfied (S233). Here, the regeneration condition of the prediction model may be a combination of the following: the average value of the prediction error during a unit time exceeds a first threshold average value; the number of times the prediction error exceeds a threshold value or a threshold ratio during a unit time exceeds a first threshold number; the number of times the prediction error continuously exceeds a threshold value or a threshold ratio during a unit time exceeds a second threshold number; or the ratio of the time during which the prediction error exceeds a threshold value or a threshold ratio relative to a unit time exceeds a first threshold ratio.

[0123] If the conditions for regeneration of the prediction model are satisfied, the prediction model is regenerated using the training data, and the previously used prediction model is updated with the regenerated prediction model (S235). Here, the regeneration of the prediction model is characterized by using a subsequent data set generated from the user's biometric information measured up to the current point in time, in addition to the previous data set used to generate the prediction model stored in the storage unit.

[0124] Here, the data set is characterized by a feature vector value generated from the first feature value and the second feature value.

[0125]

[0126] Figure 8 is a drawing for explaining an example of time delay compensation.

[0127] As illustrated in Figure 8, a time delay occurs between the biometric information actually measured by the body-attached unit (represented by the dotted line) and the biometric information received by the communication terminal (represented by the solid line). This time delay can be corrected using known Kalman filters, ARIMA, etc.

[0128]

[0129] Figure 9 is a drawing for explaining an example of a method for correcting unit mismatch.

[0130] As illustrated in Fig. 9, the biometric information received from the body attachment unit in the continuous blood glucose monitoring system must be corrected for unit discrepancies using the reference biometric value at regular time intervals (t1, t2, t3, t4...), for example, every 12 hours or 24 hours, to ensure accuracy.

[0131] When correcting unit mismatch, weights can be assigned by considering whether the biometric information (indicated by the solid line) is rising or falling, and the correction can be performed by multiplying the assigned weight by the correction slope (A). Here, if the biometric information is rising, the weight can be assigned inversely proportional to the rising speed, and if the biometric information is falling, the weight can be assigned in proportion to the falling speed. For example, if the biometric information is rising, the weight can be assigned a low value inversely proportional to the rising speed (the higher the rising speed, the lower the value, such as 0.90, 0.80, 0.70, etc.), and if the biometric information is falling, the weight can be assigned a high value in proportion to the falling speed (the higher the falling speed, the higher the value, such as 1.10, 1.20, 1.30, etc.).

[0132]

[0133] Figure 10 is a drawing illustrating another example of a method for correcting unit mismatch.

[0134] As illustrated in Fig. 10, the biometric information received from the body-attached unit in the continuous blood glucose monitoring system must be corrected for unit discrepancies using the reference biometric value at regular time intervals (t1, t2, t3, t4...), for example, every 12 hours or 24 hours, to ensure accuracy.

[0135] When correcting for unit discrepancies, correction can be made based on the difference between the reference biometric value and the measured biometric value (indicated by the solid line). Weights can be assigned to reduce the difference based on the difference between the reference biometric value and the measured biometric value (for example, weights can be calculated and assigned so that the correction slope is the average of the reference biometric value and the measured biometric value), or weights can be assigned to reduce the difference only when the difference between the reference biometric value and the measured biometric value is outside the critical range.

[0136]

[0137] Figures 11 and 12 are diagrams for explaining an example of determining whether to retrain a prediction model.

[0138] As illustrated in FIG. 11, if the prediction error (d1) between the predicted biometric value predicted at the first prediction time point (t1) and the measured biometric value actually measured at the first prediction time point is outside the threshold value or critical range, or if the prediction error (d2) between the predicted biometric value predicted at the second prediction time point (t2) and the measured biometric value actually measured at the second prediction time point is outside the threshold value or critical range, the prediction model can be retrained at the first prediction time point and the second prediction time point, respectively.

[0139] Meanwhile, as shown in Fig. 12, for a certain period of time (t d ) If the average prediction error between the predicted biometric value predicted during the period and the actual measured biometric value at the corresponding time exceeds the critical average value, the prediction model can be retrained.

[0140]

[0141] Figure 13 is a diagram illustrating an example of regenerating a prediction model.

[0142] As shown in Figure 13, a certain unit time (t D ) the total number of times the predicted error between the predicted biometric value and the actual measured biometric value at the corresponding time exceeds the threshold or critical range, or exceeds the critical number of times, or a certain unit of time (t) D ) If the number of times the predicted error between the predicted biological value predicted during the period and the actual measured biological value at the corresponding time exceeds the threshold or the critical range, or if all combinations thereof are satisfied, the prediction model can be regenerated.

[0143]

[0144] Meanwhile, the embodiments of the present invention described above can be written as a program that can be executed on a computer, and can be implemented in a general-purpose digital computer that operates the program using a computer-readable recording medium.

[0145] The computer-readable recording medium includes storage media such as magnetic storage media (e.g., ROM, floppy disk, hard disk, etc.), optical readable media (e.g., CD-ROM, DVD, etc.), and carrier wave (e.g., transmission via the Internet).

[0146]

[0147] While the present invention has been described with reference to the embodiments illustrated in the drawings, these are merely exemplary, and those skilled in the art will understand that various modifications and equivalent embodiments are possible. Therefore, the true scope of technical protection of the present invention should be determined by the technical spirit of the appended claims.

Claims

1. A method for predicting a user's biometric value using biometric information measured from a sensor, A step of extracting a first feature value from the measured user's biometric information; A step of correcting the measured user's biometric information and extracting a second feature value from the corrected biometric information; A step of generating a feature vector value by reducing and combining the first feature value and the second feature value; and A method for predicting biometric values, characterized by including a step of predicting a user's biometric value by applying the generated feature vector value to a prediction model.

2. In paragraph 1, A method for predicting biometric values, characterized in that the above sensor is a sensor that is inserted into a part of a user's body for a certain period of time to continuously measure the user's biometric information.

3. In the second paragraph, the method for predicting the biometric information is It further includes a step of preprocessing the measured biometric information by removing noise from the measured biometric information. A method for predicting biometric values, characterized in that the first feature value and the second feature value are extracted from preprocessed biometric information.

4. In paragraph 3, The above first feature value is extracted directly from the preprocessed biometric information, A method for predicting biometric values, characterized in that the second feature value is extracted from corrected biometric information generated by correcting time delay and unit mismatch of preprocessed biometric information.

5. In paragraph 4, the unit mismatch is A method for predicting a biometric value, characterized in that the biometric value is corrected based on preprocessed biometric information or a reference biometric value.

6. In paragraph 5, the unit mismatch is A method for predicting biometric values, characterized in that the preprocessed biometric information is corrected by assigning weights when it rises or falls.

7. In paragraph 5, the unit mismatch is A method for predicting a biometric value, characterized in that the biometric value is corrected by a weight assigned according to the difference between the biometric value determined from the measured biometric information and the reference biometric value.

8. In the fourth paragraph, the method for predicting the biometric value is A step of calculating a prediction error from the difference between the predicted bio-value at the first prediction time point and the bio-value actually measured at the first prediction time point; and A method for predicting biometric values, characterized in that it further comprises a step of determining whether to retrain the prediction model based on the prediction error.

9. In paragraph 8, A method for predicting biometric values, characterized in that it is determined to retrain the prediction model when the above prediction error is greater than a threshold value or a critical ratio.

10. In the 8th paragraph, the method for predicting the biometric value is A method for predicting biometric values, characterized in that it further comprises a step of determining whether to regenerate the prediction model based on the occurrence characteristics of the prediction error during a unit time.

11. In clause 10, the expression characteristic is A method for predicting a biometric value, characterized in that at least one of the number of times the prediction error continuously exceeds a threshold value or a critical ratio during the unit time and the total number of times the prediction error exceeds a threshold value or a critical ratio during the unit time.

12. In the method of predicting the biometric value in paragraph 8, A method for predicting biometric values, characterized in that the retraining of the prediction model or the regeneration of the prediction model uses a subsequent data set generated from the user's biometric information measured up to the present point in time in addition to the previous data set used to generate the prediction model.

Images