KR20220058374A - Method of monitoring the hydration of a living body - Google Patents

Abstract

A method for monitoring the water supply of a user body according to embodiments of the present invention comprises the steps of: obtaining personal data of a user; obtaining sensor data related to the user body from a plurality of sensors; estimating the actual water intake of the user body and the activity level of the user body based on the sensor data; calculating the daily water requirement of the user body based on the activity level of the user body and the personal data of the user; comparing the actual water intake of the user body with the calculated daily water requirement of the user body; and providing at least one personalized recommendation related to the amount of water to be consumed by the user to the user based on the comparison result.

Description

Method and device for monitoring the water supply of the living body

The present disclosure relates to a method for monitoring human health, and more particularly, to a method that can be used to monitor a human condition that regulates optimal water intake of a person.

Recently, smartwatches have gained great popularity among consumers. Smartwatch manufacturers are looking for new opportunities to capture the imagination of consumers, and while wearing a smartwatch, users can find a lot of information useful to track their health, such as heart rate, blood pressure, body temperature, steps taken, and the air around them. Knowing about the concentration and type of pollution, etc., the smartwatch also uses basic and necessary functions such as synchronization of notifications with a smartphone, an alarm clock, a calendar with event notifications, internet connection, payment.

During strenuous sports, fitness activities, or other strenuous physical activity, sweat loss may exceed water intake. A lack of water in the body can lead to dehydration, which can be accompanied by increased fatigue, loss of concentration, and headaches. Moreover, severe dehydration of the body may lead to changes in the body's water-electrolyte balance, renal failure, that is, life-threatening.

In clinical and research practice, isotope dilution with heavy water isotopes (D2O), urine or saliva parameter testing, and skin turgor observations have been used to estimate hydration. Not all of these methods are implemented in a mobile device in a user-friendly manner.

The solution disclosed in US Patent Publication US 2016003615 A1 (published on 01.07.2016) is known in the prior art. The above prior art solution uses a flexible network-connected smart band to track liquid consumption by the user in the liquid container. The smart band can detect the direction of the liquid container and identify swallows by the user based on the pattern of the sensed direction. The smart band can also estimate liquid intake based on sensed swallowing characteristics, such as the duration of the swallowing action. The detected swallowing and swallowing characteristics and liquid intake may be displayed on the smart band, and may also be transmitted to a server for storage and display on other network-enabled devices. The biggest disadvantage of the above solution is that it is inconvenient to use because the user has to drink only from such a container considering the total liquid intake. In addition, as another disadvantage, it can be understood that these solutions have a complex design and do not take into account the individual characteristics of the user, and therefore cannot provide a recommendation suitable for him/herself considering his/her physical activity and physiological characteristics. There will be.

The solution disclosed in US Patent Publication US 10206619 B1 (published on February 19, 2019) is known in the prior art. The above prior art device provides hydration monitoring using electromagnetic radiation having a predetermined intensity at several wavelengths in the optical range transmitted through a body part. The output intensity and absorption coefficient of radiation by the human body are measured for each of these wavelengths. The slope of the absorption curve described by the Beer-Lambert law is dependent on the relative water content of the part of the body being measured. Through this dependence, the device makes it possible to assess the level of hydration of the human body. A disadvantage of the proposed solution is that it does not provide the user with individual recommendations (eg in case of dehydration) and does not monitor whether these recommendations have been followed.

One solution is known from the prior art as disclosed in US Patent Publication No. US 2016374588 A1 (published on December 29, 2016). The disclosed document discloses an example related to monitoring a body hydration level based on a galvanic skin response measurement obtained by a wearable electronic device. operating a sensor configured to measure a galvanic skin response, a logic device, and a hydration monitoring mode as an example and obtaining multiple measurements of the skin galvanic response over time, relating to the multiple measurements of the galvanic skin response A wearable electronic device is provided that includes a storage device comprising instructions executable by the logic device to present data. A disadvantage of the proposed solution is that it does not provide the user with individual recommendations (eg in case of dehydration) and does not monitor whether these recommendations have been followed.

Another solution is also known from the prior art as disclosed in US Patent Publication US 2020000363 A1 (published on 2/1/2020). This solution is a technique most similar to the method proposed above. The known solution described above discloses one approach in which data is collected from various sensors based on the obtained data, and the user receives information about dehydration of the body. The sensor interface includes a first sensor operable to perform a plurality of physiological measurements on the user over a period of time. The processing device may be coupled to the sensor interface. The processing device may be operable to receive a plurality of physiological measurements, determine a change in the plurality of physiological measurements, and determine a hydration status measurement for the body based on the change in the plurality of physiological measurements. there is. The device displays the user's hydration status indicator. The above prior art solutions do not provide the user with information on liquid intake, and the user does not receive individual recommendations on the optimal water intake required.

There is no device in the prior art that makes it possible to compare liquid intake and water loss by a user and, based on such comparison, provide the user with individual recommendations for water intake.

The present disclosure aims to provide a method and apparatus for providing individual recommendations on water intake to a user.

A method of monitoring the water supply of a user's body according to embodiments of the present disclosure includes a process of acquiring individual data of a user, a process of acquiring sensor data related to the user's body from a plurality of sensors, and the estimating an actual water intake of the user's body and an activity level of the user's body based on the and comparing the actual water intake of the user's body with the calculated daily water requirement of the user's body, and at least one individual recommendation related to the amount of water to be consumed by the user to the user based on the comparison result process to provide.

In the method of monitoring the water supply of a user's body according to embodiments of the present disclosure, in the process of estimating the actual water intake of the user's body based on the sensor data, the water intake of the user's body based on the sensor data and estimating the water loss amount of the user's body, and calculating the actual water intake amount of the user's body based on the water intake amount of the user's body and the water loss amount of the user's body.

In the method of monitoring a user's body water supply according to embodiments of the present disclosure, the at least one recommendation is provided by the user's request.

In the method of monitoring the water supply of the user's body according to embodiments of the present disclosure, the amount of water loss of the user's body includes the amount of water loss due to sweat over a certain period of time.

In the method of monitoring water supply of a user's body according to embodiments of the present disclosure, the amount of water loss of the user's body is estimated based on the sensor data and the activity level of the user's body.

In the method of monitoring a water supply of a user's body according to embodiments of the present disclosure, the activity level of the user's body is estimated based on the number of steps taken during the day of the user.

The method for monitoring water supply of a user's body according to embodiments of the present disclosure further includes a process of providing a current state of the water balance of the user's body related to the water intake of the user's body and the water loss amount of the user's body include

In the method of monitoring the water supply of a user's body according to embodiments of the present disclosure, the individual data of the user includes at least one of the user's gender, the user's age, and the user's weight.

In the method of monitoring the water supply of a user's body according to embodiments of the present disclosure, the recommendation is continuously updated.

In the method of monitoring water supply of a user's body according to embodiments of the present disclosure, the activity level of the user's body includes one of low, medium, and high.

The method of monitoring the water supply of a user's body according to embodiments of the present disclosure is a method of providing at least one individual recommendation related to an amount of water to be consumed by the user to the user based on the comparison result, the comparison result Accordingly, when the water imbalance of the user's body is less than the first threshold of the user's body weight, if the activity level of the user's body is low, the at least one individual recommendation is not provided to the user, and the and providing the at least one personalized recommendation to the user if the activity level is high.

The method of monitoring the water supply of a user's body according to embodiments of the present disclosure is a method of providing at least one individual recommendation related to an amount of water to be consumed by the user to the user based on the comparison result, the comparison result According to the above, when the water imbalance of the user's body is greater than or equal to the first threshold value and less than the second threshold value of the user's body weight, if the activity level of the user's body is low, the at least one individual recommendation is given to the user. and providing the at least one individual recommendation to the user repeatedly if the user's body activity level is high.

The method of monitoring the water supply of a user's body according to embodiments of the present disclosure is a method of providing at least one individual recommendation related to an amount of water to be consumed by the user to the user based on the comparison result, the comparison result and providing a warning indicating that the user's body is in a dehydrated state and the at least one individual recommendation to the user when the water imbalance of the user's body is equal to or greater than the second threshold value of the user's body weight. .

The electronic device for monitoring the water supply of a user's body according to embodiments of the present disclosure, when executed by a plurality of sensors, one or more processors, and the one or more processors, causes the electronic device to acquire user's individual data and obtaining sensor data related to the user's body from the plurality of sensors, and estimating an actual water intake of the user's body and an activity level of the user's body based on the sensor data, and the activity level of the user's body and calculating the daily water requirement of the user's body based on the individual data of the user, comparing the actual water intake of the user's body with the calculated daily water requirement of the user's body, and based on the comparison result and a memory storing instructions for providing the user with at least one personalized recommendation related to an amount of water to be consumed by the user.

The electronic device for monitoring water supply of a user's body according to embodiments of the present disclosure stores instructions that, when executed by the one or more processors, cause the electronic device to perform any one of the methods described above. and the memory.

The foregoing and other features and advantages of the present disclosure are illustrated in the following description with reference to the accompanying drawings
1 shows a block diagram of a method and apparatus for monitoring hydration of a human body according to embodiments of the present disclosure;
2 illustrates estimation of liquid intake by a user according to embodiments of the present disclosure;
3 illustrates a model for confirming a liquid intake gesture by a user according to embodiments of the present disclosure.
4 illustrates training of a model for confirming a liquid intake gesture by a user according to embodiments of the present disclosure.
5 illustrates an operation of a unit performing a function of a model predicting a final result using only information from a specific gesture according to embodiments of the present disclosure.
6 schematically illustrates a general method for confirming an activity of an arbitrary user according to embodiments of the present disclosure.
7A illustrates the dependence of the amount of liquid ingested on the total duration of a gesture according to embodiments of the present disclosure.
7B illustrates the dependence of the amount of ingested liquid on the total number of gestures performed according to embodiments of the present disclosure.
8 illustrates an example of individual recommendations that a user may receive throughout the day according to embodiments of the present disclosure.

A method for evaluating the user's hydration status (water saturation of the body) is proposed. The present disclosure makes it possible to provide individualized recommendations to the user on optimal water intake, and also to warn the user about whether the body is dehydrated or close to dehydrated, or the user's excessive fluid intake. The proposed device provides high accuracy to a user with any level of physical activity, and does not place any restrictions on the user's daily activities. It is also possible to detect and calculate the user's liquid intake gesture. The present disclosure provides automatic tracking of fluid ingress (generation and estimation of fluid volume) into the body by processing data from the smartwatch's inertial sensor. The present disclosure processes data from the smartwatch's inertial, optical and temperature sensors to provide an automatic estimation of water loss due to sweating during sports and daily activities. The present disclosure monitors the user's body's water balance with respect to the amount of water to be consumed based on observed sweat loss and fluid intake and provides personalized recommendations to the user. Therefore, a user-independent fully automatic device (without the need for additional user action) for monitoring the balance of hydration for the smart watch is proposed, which is both dehydration and over-hydration of the body. to provide users with timely and individual recommendations to prevent Moreover, overhydration is just as dangerous as dehydration. If the body is excessively hydrated, body temperature may drop and salivation, nausea, vomiting, impaired movement control, convulsions, muscle weakness, and headaches may occur.

A person does not always drink water, and impure water enters the body when eating food, but taking into account the error, the amount of liquid that a liquid enters the body with food and drink is It should be noted that it can be assumed that it is substantially equal to the amount of water entered, and an error from this assumption is small and is ignored in the present disclosure. In other words, when calculating the water balance, the body's fluid intake is treated the same as the body's water intake.

The present disclosure may be useful to any other user, including users who engage in vigorous physical activity as well as users who lead an inactive lifestyle.

The present disclosure may be used in any suitable computing device for a user, comprising a processor and a memory storing instructions for causing the processor to perform the steps of the proposed method. Such devices may be, but are not limited to, smartphones, smartwatches and other suitable devices, which will be referred to as smart bands below. The memory may be any medium for storing data, in particular a computer-readable storage medium.

The present disclosure is based on data analysis from sensors used in a smart band. As mentioned above, the sensor may be an inertial, optical and temperature sensor, and a necessary and sufficient condition for the implementation of the present disclosure is the presence of such sensors. Also, additionally/optionally, body resistance sensors (bio-impedance sensors), galvanic skin response sensors, electrochemical sensors, and other sensors directly or indirectly related to body water balance that measure the impedance of the body can be used Additional sensors may play a role to more accurately predict body hydration.

A machine learning algorithm is used to perform the following actions based on the data received from the sensors:

estimation of water loss that accompanies sweating during physical activity; and

Detecting individual gestures and quantifying the user's fluid intake to detect user activity such as nutrition, especially fluid intake.

Based on the above estimation, the user receives a personalized recommendation for water intake according to the needs of the current user's body. It is also necessary to consider the age, gender, weight and height of the user when making recommendations.

A new function for checking the user's body's hydration balance is introduced into the smart band by the present disclosure, which also includes the following operations:

automatic estimation of sweating, that is, the amount of water lost with sweating;

Automatic identification of fluid intake by the user, ie how many milliliters of water the user has drank, along with an estimate of the amount; and

Providing individualized recommendations for water intake to users based on their physiological needs.

The proposed system is continuously being trained and tested, ie the database to which the device is connected is updated repeatedly to provide more accurate recommendations to the user.

For the correct operation of the algorithm, a lot of data must be obtained from people with different individual parameters, and the conditions under which the subject is located must also be different. In other words, in order to collect data under various conditions, it is necessary to change the ambient temperature and humidity. To this end, during data collection, some test personnel run on a treadmill, while others change the temperature and humidity of the environment so that different climatic conditions are artificially created. Other test personnel may run the distance in various weather conditions. Also, the subject may consume liquid from the table while sitting or standing, taking the table height and seat height into account. Also, different test personnel could consume liquids with one hand wearing the smart band or the other without the smart band. All of the foregoing are merely examples of what may be considered when collecting data to train the algorithms of the present disclosure.

During data collection, essential sensors such as accelerometers, gyroscopes, heart rate sensors (plethysmographs) and temperature sensors are used. The duration of physical activity is recorded for each test person. The signal received from the sensor is preprocessed using a Gaussian filter and a median filter to remove outliers that do not convey basic information. Based on the data obtained from the signal, a set of features is computed, which is a statistical feature that is summary information about a portion of a signal or value-changing signal over time, and all or part of the harmonics of a signal over a specific or all period of time. It can be divided into a frequency characteristic that is a change. The properties most relevant to water loss are identified and, therefore, allow estimation of the corresponding liquid loss when properly described.

Further, according to the obtained data and calculated features, a regression model is trained (eg, "Boosting on Trees"), and corresponding hyperparameters are at least cross-validated. It is an absolute error in cross-validation. After finding the hyper-parameters, the corresponding regression model is trained on all data.

A trained regression model implemented in a smart band can provide individual characteristics to users based on data to be obtained from specific users. Water loss is estimated using the trained regression model to provide individualized recommendations.

In one of the embodiments of the present disclosure, sensors such as an accelerometer, a gyroscope, a heart rate sensor (volume meter), and a temperature sensor may be used. In order to determine the characteristics of a user's individual water metabolism, in addition to the data obtained from these sensors, it is necessary to consider the user's gender, age, height and weight, and the skin surface area calculated from indicators of height and weight.

1 is a block diagram of a method and apparatus for monitoring human body hydration according to embodiments of the present disclosure.

As shown in FIG. 1 , the apparatus for monitoring water supply of a living body includes:

a user's individual data acquisition unit 1 including gender, age, and weight;

accelerometer,

gyroscope,

pulse sensor,

with temperature sensor

Sensors (not shown in FIG. 1 ):

sensor data acquisition unit (2);

sub-unit (a) for determining the level of physical activity,

a sub-unit (b) for determining the amount of water loss;

sub-unit (c) for ascertaining liquid intake;

a sub-unit (d) for ascertaining actual water intake;

data pre-processing unit (3);

a unit (4) for calculating the daily water requirement;

a unit (5) for comparing the actual intake with the calculated daily water requirement;

a unit 6 for providing recommendations to the user;

Here, the sensor data acquisition unit (2) is connected to the data pre-processing unit (3),

The sub-unit (b) for ascertaining the water loss amount and the sub-unit (c) for ascertaining the liquid intake amount are connected to the sub-unit (d) for ascertaining the actual water intake amount,

the sub-unit (d) for ascertaining the actual water intake is connected to the unit (5) for comparing the actual water intake with the calculated daily water requirement;

the sub-unit (a) for ascertaining the level of physical activity is connected to the unit (4) for calculating the daily water requirement,

the individual data acquisition unit (1) is connected to the unit (4) for calculating the daily water requirement;

the unit (4) for calculating the daily water requirement is connected to the unit (5) for comparing the calculated daily water requirement with the actual water intake; And

A unit 5 for comparing the actual water intake with the calculated daily water requirement is connected to a unit 6 for providing a recommendation to the user.

Recommendations to users relate to optimal water intake. If the body is on the verge of dehydration or is over-hydrated with water, the recommendation to the user indicates a warning (7) about the user's body dehydration or over-hydration of the user's body.

The daily water requirement calculation is based on:

- percentage of body weight (%),

- skin surface area (ml/m2), and

- Sweat ratio of body weight (ml/kcal or ml/kg).

Here, the following are also taken into account:

- age and gender;

- level of physical activity,

- basal metabolic rate, and

- Environmental factors.

The human body's water supply monitoring method is performed as follows. The user wears and activates a smart band, which may be a smartwatch or other suitable device that can be attached to the user's body. Hereinafter, such a device is referred to as a “smart band” for convenience. The smart band may include at least an accelerometer, a gyroscope, a heart rate sensor (plethysmograph), a temperature sensor, and the like. In order to take into account environmental conditions, the device may include sensors that measure, for example, ambient temperature, humidity, and the like.

In order for the user to use the suggested function of the smart band, the user must input personal information about gender, age, height, and weight provided by the user's personal information collection unit 1, without this information, the user cannot receive a correct recommendation. there will be no

The smart band starts acquiring data using these sensors, the acquired data is sent to the sensor data acquisition unit 2, from which the data is sent to the data preprocessing unit 3, where the readings of the sensors are Based on the following values are calculated using the algorithm.

Body water loss: As described above, the trained machine learning regression model predicts the amount of water lost due to sweat over a period of time (eg, during periods of physical activity or throughout the day), which occurs in the sub-unit (b) for

- body fluid intake: this data is checked in sub-unit (c) for ascertaining the liquid intake, the body fluid intake is equal to the body water intake.

- Actual body water intake: identified in sub-unit (d) for ascertaining actual water intake based on body fluid intake and body water loss.

- Physical activity level: identified in sub-unit (a). Based on data from inertial sensors (accelerometers and gyroscopes), a person's general level of physical activity is estimated, for example, by the number of steps per day, which is usually necessary to correct the user's daily water needs, and that activity level is It is divided into three grades: low, medium, and high. If a particular person is engaged in an average level of activity, 700 ml is added to the daily rate, and 1100 ml is added to the daily rate if the activity level is higher. This data is input into sub-unit (a) for ascertaining the level of physical activity.

Based on exploratory analysis, a feature(s) that correlated with the estimate of water loss was found. To estimate water loss, the correlation feature(s) is computed and input to the trained regression model to estimate water loss in ml.

Based on the data on the level of physical activity and the user's personal data, the daily water requirement of the user is calculated in the unit 4 for calculating the daily water requirement.

In addition, the calculated data, ie, body water loss and body fluid (ie, water) intake, are recalculated as actual data of the user's water intake. And the difference between the actual water intake and the calculated daily water requirement is calculated by comparing the actual water intake and the calculated daily water requirement in a unit for comparing the actual water intake and the calculated daily water requirement.

Based on the above calculation, a recommendation is provided, for example, on the amount of water the user should drink within the day of the day. These recommendations may be continuously updated, and such recommendations may be provided as the body reaches a certain level of dehydration or approaches dehydration.

If the body fluid intake is less than the threshold for water needs, the user will be warned about possible dehydration along with information about the amount of water to be consumed.

If the body fluid intake is greater than the threshold for the daily water requirement, the user will be warned of excessive fluid intake along with information about that fluid amount.

After a recommendation for the amount of liquid is provided, the device continues to acquire and calculate data, so the device monitors and judges whether the recommendation was followed. In a preferred implementation, the frequency of repeated notifications is user configurable. The user can check the current status of the water balance at any time.

Thus, overall hydration monitoring of the body is performed.

It is clear that there is a certain correlation between the received data and user behavior. There is a certain level of physical activity for each user based on gender, age, weight and height. To cluster users into groups, a regression model (e.g., "boosting in the tree") is used, which refines features (weight, height, age) for each subgroup of users, It provides values not only based on the sensors of the smart band, but also using data about the user input by the user. Based on this, features that maximize the generalization of the model for all users are selected.

Based on data from the accelerometer and gyroscope sensors, the gesture of drinking can be distinguished from all other user gestures.

By detecting the time spent in the drinking gesture, it is possible to ascertain the amount of liquid the user has drank.

Now, the description will be given in consideration of the detailed monitoring of the human body's moisture supply.

The first factor in estimating a user's water balance is to estimate the amount of water lost due to perspiration.

Based on the signals received from the accelerometer and gyroscope sensors, the heart rate sensor, and the temperature sensor, the user activity period can be checked. As the user begins to move, it is evident that speed, heart rate and temperature begin to increase.

The above characteristics can be divided into statistical characteristics, which are summary information about a signal or part of a signal of a change in value over time, and a frequency characteristic, which is a harmonic change of all or part of a signal over some or all of the time period. Once all properties have been collected, the features that best correlate with water loss are identified and, if properly accounted for, this allows estimation of water loss. When estimating water loss, features correlating with the water loss estimate are computed and fed into the regression model estimating water loss.

The signals from the sensors are input to the sensor data acquisition unit 2, which is then processed in the data preprocessing unit 3, where the characteristic features most influencing the water loss estimation are selected from a series of signals, i.e. , that is to say, the signals with the above characteristic characteristics are passed to the sub-unit (b) for ascertaining the amount of water loss, which uses a machine learning regression model to estimate the amount of water loss. Since the unique characteristics of the sensor signal are identified during the machine learning phase, the trained regression model will select signals with specific characteristics during operation. Accelerometer signals provide information about the user's speed and steps per minute. A plethysmogram provides information about heart rates per minute. Temperature sensors provide information about body and ambient temperature. The data is filtered using a median filter and a Gaussian filter. The signals are then normalized and characteristic features associated with body water loss are selected.

Training samples and test samples that are partitioned so that there are no gaps in the distribution of height, weight, gender, and age are provided. Next, all the features are computed, and then the features with the greatest correlation with each other are removed because the correlation of the random variables determines the linear change of one value when the other random variable is changed. When the features are strongly correlated with each other, their contributions are nearly identical, so that all but one strongly correlated features are dropped out, i.e., features that are not correlated with each other are added to the regression model. will belong Features are iteratively added to this number of features in the regression model until the cross-validation error starts to increase. In the cross-validation process it is necessary to explain that the sample is divided into N samples (training and validation), the algorithm is trained on the training sample and validated on the validation sample, and then the error is calculated in each validation sample Thus, N estimates are obtained for different verification samples. Then, the average error is applied to all N errors on the validation partitions. This experiment is repeated for several different regression models, where the different regression models trained differ only in other parameters such as number of trees and maximum tree depth. After calculating the cross-validation error on several different regression models with different parameters, the best model is selected by the error on the cross-validation, and then the regression model is determined, the error of which is the smallest as well as the test sample It is closest to the error of the image, and also the test sample is constructed from a large dataset, and about 25-30% of the entire dataset is selected, making it possible to select the best model in terms of model parameters.

를 각각 실제 수분 손실 및 추정 수분 손실의 벡터라고 하자. 여기서, 이들 벡터는 실수 집합(set of real numbers)에 속한다. 다음으로,

는 실제 수분 손실의 평균값이고,

는 총 2차 합(total quadratic sum)이고,

는 제곱 편차(square deviations)의 합이고,

는 수분 손실 추정 알고리즘의 품질을 평가하는 결정 계수(coefficient of determination)(R2)이다.The algorithm provides the calculation of an estimate of water loss normalized to body surface area. now,

Let be vectors of actual and estimated water loss, respectively. Here, these vectors belong to the set of real numbers. to the next,

is the average value of the actual water loss,

is the total quadratic sum,

is the sum of square deviations,

is a coefficient of determination (R2) that evaluates the quality of the water loss estimation algorithm.

Then, the resultant value is multiplied by the body surface area, and a calculated value of the amount of water loss due to sweat is obtained.

In order to increase the accuracy of determining the parameters, a sensor for providing information on the moisture content of the skin according to an optical characteristic of skin scattering may be additionally installed in the device for monitoring the moisture supply of a living body. These sensors can operate in both the visible and near-infrared regions of the spectrum, collecting data about the moisture content in the user's subcutaneous layer.

In addition, the proposed device may be additionally equipped with an impedance sensor for measuring the water content of the whole body. To this end, the first electrode is located under the smart band adjacent to the skin, and the second electrode is separated and disposed on the body of the smart band. In order to measure the impedance, the user will have to touch the second electrode with a hand not wearing the smart band.

In addition, the proposed device may be equipped with a galvanic skin response sensor. The two electrodes of this sensor may be located at the lower end of the smart band. This sensor measures local sweat production in a specific area. Based on the data from these sensors, conclusions are made about how the pores of the body work.

Another sensor that can be further mounted on the proposed device is a chemical sensor comprising a micro-fluidic channel through which the user's sweat passes. The sensor analyzes the chemical composition of electrolytes in the user's sweat. Changes in the proportion of electrolytes lead to conclusions about the state of hydration throughout the body.

A user position detection sensor may also optionally be used. The user's location is confirmed by location information data (GPS), for example, the vehicle in which the user is located may be identified by speed and location information. In this case, data on the weather at the user's location is requested via the Internet, since the loss of body water depends, among other things, on weather conditions.

As an additional data source, it is proposed to use a smart scale that transmits information to the smart band. A sensor for determining bio-impedance may be built into the smart balance.

Additional data sources may be various applications installed on the smartphone associated with the smart band.

Further data sources may also be the smart band and/or other wearable devices of the user, such as headphones, wirelessly connected to the smartphone. The headphones may incorporate a temperature sensor, a pulse sensor, and a microphone to monitor the drinking and eating process.

In addition, the additional data source may be a smart home infrastructure to which the user's smart band is connected. Users can wirelessly connect the smart band to the fitness equipment they train to receive additional information from the equipment.

The second factor for evaluating a user's water balance is the user's estimate of the user's fluid intake. Figure 2 shows that such an estimation can be made based on the detection of a drinking gesture.

When the user makes a drinking gesture with one hand wearing the smart band, the gyroscope and accelerometer record the movement signal of this hand. An example of the signal recorded by these sensors during ingestion of liquid is shown in FIG. 2 . The upper two graphs illustrate the signals recorded by the hand-worn device in an active state, and the two lower graphs illustrate the signals of the inactive hand-worn device. Each of the graphs shown above includes three components corresponding to the linear acceleration and the angular velocity with respect to the three coordinate axes (x, y, z) of the sensors. Rectangular areas on the graph indicate areas of gyroscope signals corresponding to drinking motions and have a characteristic shape. Fragments of these repeated signals are observed on both the smart band worn by the user on the hand that brings the liquid container to the mouth and the second smart band on the inactive hand.

The signals of hand movement to the mouth for the drinking action are different from the signals of the hand movement to the mouth for eating or exercising by raising the arm or using other gestures. Each of these gestures has its own specificity: for example, the eating gesture involves pricking or scooping food, and the drinking gesture involves turning one's hand and holding the hand longer in the mouth. In many examples, the trained algorithm can recognize various movements even if the user imitates the gesture of drinking with empty hands, and the algorithm recognizes that it is not a gesture of drinking, and the occurrence of the drinking event itself and Evaluate both the duration of the gesture.

If the user makes a drinking motion with a hand that is not wearing the smart band, the accelerometer and gyroscope record micromovements of the inactive hand that correspond to the movement of the whole body when the user makes a drinking motion with the active hand. record In FIG. 2 , minute movements of the inactive hand and the active hand movement are indicated by rectangles, time is indicated on the X axis, and the levels of the accelerometer and gyroscope signals are indicated along the Y axis, respectively.

As an example, consider the gesture a person makes during a single swallow. Typically, these gestures begin with a vascularized hand to the mouth. When initiating these gestures, the gyroscope sensor records the hand's acceleration. At the end of the movement, if a container with a liquid is placed in a person's mouth, inhibition is observed, which appears as a characteristic shaped signal in measurements taken on the gyroscope. The downward movement of the hand holding the container has a characteristic characteristic that allows it to detect a liquid intake gesture and to estimate its volume.

The algorithm for detecting a liquid ingestion event and estimating its volume is based on a neural network model that simultaneously solves the problem of detecting a liquid ingestion gesture and estimating the volume of this liquid. The accelerometer and gyroscope signals are pre-processed before being input to the specified model. Such processing may include digital filtering to remove noise, decimation, scaling, normalization, and the like.

A schematic representation of a neural network model for confirming a drinking gesture is shown in FIG. 3 .

The algorithm makes it possible to detect a liquid intake gesture. This gesture is referred to as a human swallowing gesture. A model in machine learning is understood as a functional transformation from input data to an estimated result and a set of parameters required to implement this transformation.

During the operation of the proposed device, the accelerometer continuously measures the direction and magnitude of the acceleration vector including the gravitational acceleration, and the gyroscope measures the angular acceleration. The measurement sequence of these sensors includes information about the trajectory of the device and hence the hand wearing it. These trajectories in turn describe the movements of a person.

In sub-unit (c) for ascertaining liquid intake, temporal sequences of signals from the accelerometer and gyroscope are converted into vector representations describing which gestures the user makes at which point in time during the measurement. Next, a series of vector representations are analyzed.

The resulting sequence of vector representations can be regarded as the input signal to the next functional level of the model. A decision is made at the next functional level of the model, which indicates which of the gestures described by that vector representation refers to the fluid intake event. Part (a) of FIG. 3 is divided into two units. The left unit is responsible for detecting the gesture itself, and at the output of this device data is received indicating the time at which the person makes the drinking gesture. The right unit is used to estimate the amount of liquid ingested as a result of that gesture. The amount of liquid ingested is estimated based on the duration of the gesture and the angle of inclination of the container with respect to the liquid.

Part (b) of Figure 3 shows the calculation of the user's liquid intake. A series of preprocessed measurements (the preprocessing task described above) is fed into the neural network model or rather its first layer. An encoder is a series of convolutional layers, each of which performs a convolution operation of a set of input signals using a set of kernels belonging to the layer. The convolution operation includes mixing input signals according to a ratio given by a weight describing each of the specified kernels.

As a result of sequentially applying the operations described above for each of the encoder layers, a set of time series is formed at its output. This set of time series values for a specific point in time is a numerical description of movements performed by a person in a short period of time. This set of values is essentially a vector representation of that motion.

The set of time series formed at the output of the encoder is fed into two units that solve the problem of detecting a liquid intake gesture during a swallow and estimating the volume of this swallow. In one of the embodiments, these units are built based on recurrent layers (LSTM or GRU). During the model operation, these layers store state vectors characterizing the time sequence of signals input to the signal layer. That is, this state vector describes a series of gestures performed by a person at a short time interval prior to the current gesture.

4 illustrates training of a model for confirming a liquid intake gesture by a user according to embodiments of the present disclosure. In model (b), accelerometer and gyroscope data (a) are input, which is a set of vectors containing six measurements for each instant in time, the three numbers measured by the accelerometer represent a vector of linear acceleration. On the other hand, the three numbers measured with the gyroscope designate the angular velocity. A set of series of measurement values is formed in such a way that the delay between two successive measurements is constant and, if necessary, interpolation of the signals is performed to ensure this condition. Time is not explicitly sent to the model, but time is encoded as a series of sensor measurements.

The main task of the proposed model is to continuously estimate the amount of liquid a person drank. These properties of the model are provided as a result of training, which is a process of solving an optimization problem in order to find parameters of the model that mean weighting coefficients in a function describing transformations with respect to input variables. Upon completion of the learning process, the parameters are selected such that a minimum value of the error function computed on the dataset used is achieved. An important feature of the training process of the model proposed in this disclosure is that, while retrieving the necessary parameters of the regression model, the solution of two problems existing in detecting a liquid intake gesture and estimating its liquid intake is optimized. In the process of retrieving the parameters of this model from the training dataset, information about the time interval as well as the amount of liquid as a result of a human swallowing gesture is also used. An error function is constructed comprising two components, a first component characterizing the quality of gesture detection including swallowing, and a second component characterizing the quality of drinking intake estimation. Errors for both problems are taken into account during optimization. This approach makes it technically feasible by greatly simplifying the learning process.

The proposed approach to training allows to discover the weights of this model, which solves the problem of classifying gestures as drinking and other related as well as estimating the amount of liquid drunk as a result of that gesture. to ensure high quality solutions for

A large number of datasets are used in the training of the model, since each person has a different drinking gesture and the containers from which they can drink liquid can also be significantly different.

The proposed model can be modified in several ways. One way to increase the accuracy of estimation is to supplement the model with a unit responsible for predicting the final result using only information from specific gestures, as shown in FIG. 5 . In other words, thanks to the unit used, the model makes it possible to pay attention to specific characteristics of signals in the overall temporal context. As shown in Fig. 5, when segments of signal changes at specific time intervals are input to a model having the unit, the model uses information from various points in the segment.

In other words, the accuracy of model estimation can be increased by adding an attention unit to it. The presence of this device will allow the model to assign more weight to the vector representations of the individual gestures associated with the swallow itself and the assistive gesture. Since these gestures may include information on the type and volume of a container used in the drinking process, it is possible to increase the accuracy of estimating the volume of the ingested liquid. Research also shows that attention mechanisms allow models to store longer temporal contexts (history of actions performed by users). This has a positive effect on the quality of the model.

By supplementing the input signal processing algorithm, it is possible to define various other user gestures besides the drinking gesture (liquid intake). 6 schematically shows a general method for confirming any user activity according to embodiments of the present disclosure. That is, by using a regression model trained on a number of examples with various functions as well as an accelerometer and a gyroscope, user actions can be classified based on the detection of movement patterns. Thus, it is possible to recognize various activities of the user, for example, but not limited to, a drinking gesture, for example, what kind of beverage the user is consuming, eg, hot, by the intake rate. Or you can check if it is cold. It is also possible to modify the model to identify a moment in the user's life when the user is definitely not in the process of ingesting a liquid. The model may also identify what the user is eating and the approximate liquid content of his food to take into account the amount of liquid ingested with the food in the overall balance of liquid intake. In addition, the user can use headphones that are wirelessly connected to the smart band and have a built-in microphone, which captures sounds around the user while the user is wearing the headphones, for example the sound equivalent to swallowing liquid. Patterns can also be analyzed to detect drinking cases. That is, it is possible to monitor the user's movement characteristics and sounds around the user in order to detect liquid intake. Specific movements that the trained model recognizes as liquid intake are recorded, and the user's liquid intake is estimated according to their duration and taking into account the user's individual parameters. It should be noted that this is only an estimate of the amount of liquid ingested, not an exact calculation.

An additional data source for improving the accuracy of estimating a user's liquid intake may be data from other user devices such as smartphones, smartwatches, and smart home infrastructure.

It is also possible to use additional algorithms to determine which hand is wearing the smartband and which hand is the dominant hand that the person uses more often.

The following describes an illustrative example of estimating the amount of liquid ingested for a user. 7A illustrates the dependence of the amount of liquid ingested on the total duration of a gesture according to embodiments of the present disclosure. 7B illustrates the dependence of the amount of ingested liquid on the total number of gestures according to embodiments of the present disclosure. That is, even without using information about the shape of a signal that describes the trajectory of a gesture, the amount of ingested liquid can be very accurately confirmed even by these two dependencies. The above example is a simple algorithm with only two functions. For example, it is clear that by complicating the algorithm with additional features in the trajectory of hand movements, smaller errors can be obtained.

If only this information is used, results can be obtained with an error of about 10 to 15% for a specific user.

When the hydration balance is negative, that is, when the user's liquid intake is less than the amount of water the user releases through sweat, the following three options exist for event development.

(a) The slight imbalance of body water, that is, the water imbalance may be less than 1% of the user's body weight. There is no need to notify the user if the user's level of physical activity is low. If the user is physically active, the user is informed of their current body water balance and fluid intake recommendations.

(b) A significant imbalance of body water, that is, water imbalance may be within 1 to 2% of the user's body weight. If the user's physical activity level is low, the user will receive one notification of the current body water balance. If the user is physically active, the user receives information about their current balance and a notification of their water intake needs. When the user's physical activity is high, the smartwatch can turn on the counter, which reflects a decrease in the amount of water in the body.

(c) Dangerous dehydration of the body, i.e. water imbalance, can exceed 2% of the user's body weight. When the user's level of physical activity is low, the user receives a notification with an urgent recommendation to drink a certain amount of water. If the user is physically active, it will notify the user of the water intake required, but it should be noted here that large amounts of water at a time can be harmful.

8 illustrates an example of individual recommendations that a user may receive throughout the day according to embodiments of the present disclosure. For example, suppose the user is a 30-year-old male, weighing 75 kg, and the level of physical activity is moderate. Height is an optional parameter because the body's water needs are generally estimated based on body weight using different coefficients for age categories and levels of physical activity. A user's level of physical activity is identified by the average number of steps per day and the duration of the increase in physical activity during the day (running, other sports/fitness, etc.).

In this example, when the user wakes up, the smartband provides a reminder about the expected daily water needs, for example, in the above case, "Drink at least 2800ml of water per day. Drink at least 250ml of water." do. After the user has exercised, the smart band provides a notification including the following information based on the data on the volume of water lost through sweat and the volume of liquid ingested during the day, including after exercise: "Liquid intake of 400 ml. Water loss of 550 ml. Drink 450 ml of water". During office work, based on the data on the volume of sweat and liquid intake estimated during the day, the smart band provides a reminder that "1400ml of liquid intake. 200ml of water will be drunk". At the end of the day, the smartband may display a reminder: "3200ml of liquid consumed. Good job! Got health benefits today!" All these notifications are displayed to the user on the smartband screen according to the data received.

A person without a smart band will be guided only by thirst, whereas a non-exercise person who decides to go jogging for the first time will be less likely to drink more water than he did before, because this is part of his habit. because it wins If such a person drinks too little water, they are at risk of dehydration and will experience symptoms such as dizziness, nausea and other gastrointestinal tract problems. On the other hand, even if a person drinks too much water, health problems and risks occur, especially if the body does not have time to consume water, which puts a great strain on the kidneys. If a person is a user of the proposed smart watch, information about the lack of water in the body is provided even while running, and if there is a risk of dehydration, a signal is triggered from the smart band to draw the user's attention to dehydration .

Implementation of the present disclosure is possible even when no estimation of water loss is performed and only estimation of liquid intake is performed. Again in this case, the user may receive individual recommendations for water intake, but the water balance monitoring algorithm works with low accuracy.

The system still works even if the user has not entered personal data regarding gender, age, weight and height. For example, the amount of sweat in this case can be estimated by the duration and intensity of physical activity, but at the same time, the accuracy of estimating water loss will also decrease.

It should be noted that, within the scope of the present disclosure, it is meaningless to consider the excretion of water in the urine when monitoring the hydration of the body. The kidneys maintain an internal balance of water in the body, retaining water when needed or excreting excess water in the urine. At the same time, when it is already present in the bladder and excreted from the body, urine does not in any way affect the hydration of body tissues. In other words, the need for water and the removal of urine from the body are natural background processes in a healthy person. The proposed method seeks to detect the following events:

- the user drank little or no water for a long time;

- the user did a lot of physical activity and a lot of water was excreted in sweat (more than he drank);

- The user drank too much water, and the kidneys may not be able to cope with this load.

For the events under consideration above, it is not important how often and in what quantity urine is excreted from the body.

Above, although the present disclosure has been described in connection with some exemplary embodiments, it should be understood that the essence of the present disclosure is not limited to these specific exemplary embodiments. On the contrary, the substance of the present disclosure is intended to cover all alternatives, modifications and equivalents that may be included within the spirit and scope of the following claims.

Claims (26)

A method of monitoring a user's body water supply, comprising:
the process of acquiring user's personal data;
acquiring sensor data related to the user's body from a plurality of sensors;
A process of estimating an actual water intake of the user's body and an activity level of the user's body based on the sensor data;
calculating the daily water requirement of the user's body based on the activity level of the user's body and the personal data of the user;
comparing the actual water intake of the user's body with the calculated daily water requirement of the user's body; and
and providing the user with at least one individual recommendation related to an amount of water to be consumed by the user based on the comparison result.
The method of claim 1, wherein the estimating of the actual water intake of the user's body based on the sensor data comprises:
estimating the user's body water intake and the user's body water loss based on the sensor data; and
and calculating the actual water intake of the user's body based on the water intake amount of the user's body and the water loss amount of the user's body.
2. The method of claim 1, wherein the at least one recommendation is provided at the request of the user.
The method according to claim 1, wherein the amount of water loss of the user's body includes an amount of water loss due to sweat over a period of time.
The method of claim 2, wherein the amount of water loss of the user's body is estimated based on the sensor data and the activity level of the user's body.
The method of claim 1, wherein the activity level of the user's body is estimated based on the number of steps taken during the day of the user.
The method of claim 1, further comprising: providing a current state of the user's body's water balance in relation to the user's body's water intake and the user's body's water loss amount.
The method of claim 1, wherein the individual data of the user comprises at least one of a gender of the user, an age of the user, and a weight of the user.
2. The method of claim 1, wherein the recommendation is continuously updated.
The method of claim 1 , wherein the activity level of the user's body comprises one of low, medium, or high.
11. The method of claim 10, wherein the method of providing at least one individual recommendation related to the amount of water to be consumed by the user to the user based on the comparison result comprises:
When the water imbalance of the user's body is less than the first threshold value of the user's body weight according to the comparison result, if the activity level of the user's body is low, the at least one individual recommendation is not provided to the user, and the user and providing the at least one personalized recommendation to the user if the activity level of the body is high.
The method of claim 11, wherein the method of providing at least one individual recommendation related to the amount of water to be consumed by the user to the user based on the comparison result comprises:
When the water imbalance of the user's body is equal to or greater than the first threshold value and less than the second threshold value of the user's body weight according to the comparison result, if the activity level of the user's body is low, the user is informed of the at least one and providing the individual recommendation once, and repeatedly providing the at least one individual recommendation to the user when the activity level of the user's body is high.
The method of claim 12, wherein the method of providing at least one individual recommendation related to the amount of water to be consumed by the user to the user based on the comparison result comprises:
When the water imbalance of the user's body is equal to or greater than the second threshold of the user's body weight according to the comparison result, providing a warning indicating that the user's body is in a dehydrated state and the at least one individual recommendation to the user A method comprising a.
In the electronic device for monitoring the water supply of the user's body,
a plurality of sensors;
one or more processors; and
When executed by the one or more processors, the electronic device acquires the user's personal data, acquires sensor data related to the user's body from the plurality of sensors, and based on the sensor data, estimating water intake and the activity level of the user's body, calculating the daily water requirement of the user's body based on the activity level of the user's body and the personal data of the user, and calculating the actual water intake of the user's body and and a memory for storing instructions for comparing the calculated daily water requirement of the user's body and providing at least one individual recommendation related to the amount of water to be consumed by the user to the user based on the comparison result .
15. The method of claim 14, wherein the memory, when executed by the one or more processors, causes the electronic device to estimate the amount of water intake of the user's body and the amount of water loss of the user's body based on the sensor data, and and storing instructions for calculating the actual water intake of the user's body based on the water intake amount of the user and the water loss amount of the user's body.
The electronic device of claim 14 , wherein the at least one recommendation is provided by the user's request.
The electronic device of claim 14 , wherein the amount of water lost by the user's body includes an amount of water lost due to sweat over a predetermined period of time.
The electronic device of claim 15 , wherein the water loss amount of the user's body is estimated based on the sensor data and the activity level of the user's body.
The electronic device of claim 14 , wherein the activity level of the user's body is estimated based on the number of steps taken during the day of the user.
15. The method of claim 14, wherein the memory, when executed by the one or more processors, causes the electronic device to cause a current state of the user's body's water balance in relation to the user's body's water intake and the user's body's water loss. Electronic device characterized in that it stores instructions to provide.
The electronic device of claim 14 , wherein the individual data of the user includes at least one of the user's gender, the user's age, and the user's weight.
15. The electronic device of claim 14, wherein the recommendation is continuously updated.
The electronic device of claim 14 , wherein the activity level of the user's body comprises one of low, medium, or high.
24. The method of claim 23, wherein the memory, when executed by the one or more processors, causes the electronic device to, according to the comparison result, determine that when the water imbalance of the user's body is less than a first threshold value of the user's body weight, the user Storing instructions for not providing the at least one individual recommendation to the user if the activity level of the body is low, and providing the at least one individual recommendation to the user if the activity level of the user's body is high electronic device with
25. The method of claim 24, wherein the memory, when executed by the one or more processors, causes the electronic device to determine that the water imbalance of the user's body is greater than or equal to the first threshold value of the user's body weight and the water imbalance of the user's body according to the comparison result. If it is less than a second threshold, the at least one individual recommendation is provided to the user once if the activity level of the user's body is low, and if the activity level of the user's body is high, the at least one individual recommendation is provided to the user An electronic device characterized in that it stores instructions for repeatedly providing a recommendation.
26. The method of claim 25, wherein the memory, when executed by the one or more processors, causes the electronic device to cause the water imbalance of the user's body to be greater than or equal to the second threshold of the user's body weight according to the comparison result; and storing instructions for providing a warning indicating that the user's body is dehydrated and the at least one individual recommendation to the user.

Images