RU2763700C1 - Method for monitoring the hydration of a living organism - Google Patents

Abstract

FIELD: medical equipment.

SUBSTANCE: group of inventions relates to medicine, namely, to a method and apparatus for estimating the hydration of the user, as well as to a smart watch with the function of estimation of the hydration of the user. The apparatus comprises an individual user data collection unit, sensors, a unit for collecting data from the sensors, a data preprocessing unit, a daily water requirement calculation unit, a unit for comparing the actual amount of water taken and the calculated daily water requirement, a unit for issuing recommendations to the user. The user data includes gender, age, weight. The sensors constitute an accelerometer, a gyroscope, a pulse sensor, a temperature sensor. The data preprocessing unit comprises a subunit for determining the level of physical activity, a subunit for determining the amount of water lost with sweat, a subunit for determining the amount of liquid taken, a subunit for determining the actual amount of water taken based on the amount of liquid taken by the body and the amount of water lost by the body. The data collection unit is therein connected with the data preprocessing unit. In the data preprocessing unit, the subunit for determining the amount of water lost with sweat and the subunit for determining the amount of liquid taken are connected with the subunit for determining the actual amount of water taken. The subunit for determining the actual amount of water taken is connected with the unit for comparing the actual amount of water taken and the calculated daily water requirement. The subunit for determining the level of physical activity is connected with the daily water requirement calculation unit. The individual data collection unit is connected with the daily water requirement calculation unit. The daily water requirement calculation unit is connected with the unit for comparing the actual amount of water taken and the calculated daily water requirement. The unit for comparing the actual amount of water taken and the calculated daily water requirement is connected with the unit for issuing recommendations to the user. When executing the method, the individual user data is placed in the individual user data collection unit. Data from the sensors is supplied to the unit for collecting data from the sensors and transmitted to the data preprocessing unit. In the data preprocessing unit, the values are estimated over a period of time. The amount of water lost by the body is therein determined by means of the subunit for determining the amount of water lost with sweat. The amount of liquid taken by the body is determined by means of the subunit for determining the amount of liquid taken, wherein the amount of liquid taken by the body is taken as equal to the amount of water taken by the body. The actual amount of water taken by the body is determined by means of the subunit for determining the actual amount of water taken based on the amount of liquid taken by the body and the amount of water lost by the body. The level of physical activity is determined by means of the subunit for determining the level of physical activity. The daily water requirement of the body of the user is calculated based on the level of physical activity and the individual user data in the daily water requirement calculation unit. The difference between the actual amount of water taken and the calculated daily water requirement is calculated in the unit for comparing the actual amount of water taken and the calculated daily water requirement. Based on the calculated difference between the actual amount of water taken and the calculated daily water requirement, at least one individual recommendation is issued to the user regarding the amount of water that the user should consume. The stages of the method therein occur continuously. The smart watch comprises the apparatus for estimating the hydration of the user.

EFFECT: automatic and continuous estimation of the amount of water lost by the body, the amount of liquid taken by the body, is provided by means of an apparatus suitable for placement in a smart watch, without actions on the part of the user, providing timely advice for the user on preventing dehydration of the body and oversatiation of the body with water.

31 cl, 8 dwg

Description

Technical field

The present invention relates to methods for monitoring human health, and can be used to monitor the state of the human body, which controls the optimal consumption of water.

Description of the Prior Art

Nowadays, smart watches have gained immense popularity among consumers. Smart watch manufacturers are looking for new ways to capture the imagination of the consumer, while wearing the watch, the user learns about himself a lot of information useful for tracking his health, for example, pulse, pressure, body temperature, the number of steps taken, the concentration and type of air pollution around him and etc., smart watches also use basic and necessary functions, such as synchronization of notifications with a smartphone, an alarm clock, a calendar with event reminders, an Internet connection, and payment of payments.

During intense sports, fitness or other intense physical activity, sweat loss may exceed the amount of water drunk. Water deficiency in the body can lead to dehydration, which can be accompanied by increased fatigue, decreased concentration, and headaches. Moreover, severe dehydration (dehydration) of the body leads to changes in the water and electrolyte balance of the body, kidney failure, that is, it can be life-threatening.

In clinical and research practice, to determine the status of hydration, the isotope dilution method using isotopes of heavy water (D2O), analysis of urine or saliva parameters, and assessment of skin turgor are used. All of these methods cannot be implemented through mobile devices in a user-friendly way.

The prior art solution disclosed in the document US 2016003615 A1 (publication date 01/07/2016). The known solution monitors a user's fluid intake from a container of fluid using a flexible smart bracelet connected to a network. The smart bracelet can detect the orientation of the liquid container and determine the sips taken by the user based on the detected orientation patterns. The smart bracelet can also estimate the volume of fluid intake using the numerical characteristics of the detected sips, for example, the duration of the sip. The detected sip, as well as the characteristics of the sip and the volume of fluid intake, can be displayed on the smart bracelet, and can also be transmitted to the server for storage and display on another device with network connectivity. The main disadvantage of the described solution is the inconvenience of its use, since in order to account for all the liquid consumed, the user must drink it only from this container. Also, as disadvantages, it can be noted that this solution has a complex design and does not take into account the individual characteristics of the user, and, therefore, does not allow him to issue recommendations to him, taking into account his physical activity and physiology.

The solution disclosed in the document US 10206619 B1 (publication date 02/19/2019) is known from the prior art. The known device provides monitoring of hydration by passing electromagnetic radiation with a given intensity at several wavelengths of the optical range through a part of the human body. For each of these wavelengths, the output intensities and absorption coefficients of radiation by the human body are measured. The slope of the absorption curve, described by the Bouguer - Lambert - Beer law, depends on the relative water content of the area of the human body in which the measurement is made. This dependence allows the device to estimate the level of hydration of the human body. The disadvantage of the proposed solution is that it does not provide individual recommendations to the user (for example, in case of dehydration) and does not provide monitoring of compliance with these recommendations.

The prior art solution disclosed in the document US 2016374588 A1 (publication date 12/29/2016). This document discloses examples that relate to monitoring body hydration levels based on galvanic skin response measurements obtained using a wearable electronic device. In one example, a wearable electronic device is provided, including a sensor configured to measure galvanic skin response, a logic device, and a memory device including instructions executed by the logic device to control a hydration monitoring mode, obtain multiple measurements of galvanic skin response over time, present data concerning multiple measurements of galvanic skin response. The disadvantage of the proposed solution is that it does not provide individual recommendations to the user (for example, in case of dehydration) and does not provide monitoring of compliance with these recommendations.

The solution described in the document US 2020000363 A1 (publication date 02/01/2020) is known from the prior art. This solution is the closest analogue of the present invention. A well-known solution discloses an approach in which data is collected from various sensors, based on the data obtained, the user receives information regarding dehydration of the body. The sensor interface includes a first sensor capable of performing a plurality of physiological measurements of a user over a period of time. The processing device is connected to the sensor interface. The processing device is configured to receive a plurality of physiological measurements, determine a change in the plurality of physiological measurements, and determine a measurement of the body's hydration state based on the change in the plurality of physiological measurements. The device displays an indicator of the user's hydration status. The known solution does not provide the user with information regarding the amount of fluid consumed, and the user does not receive individual recommendations regarding the required optimal amount of water consumption.

There are no devices in the prior art that allow comparison of fluid intake and water loss by a user, and based on such a comparison, give the user individual recommendations on water consumption.

The essence of the invention

Proposed device for monitoring the hydration of a living organism, containing:

a unit for collecting individual user data, including gender, age, weight;

sensors that are:

accelerometer,

gyroscope,

pulse sensor,

temperature sensor;

block for collecting data from sensors;

data preprocessing block containing

subblock for determining the level of physical activity,

sub-block for determining the amount of lost water,

subunit for determining the amount of liquid taken,

subblock for determining the actual amount of water received;

block for calculating the daily need for water;

a block for comparing the actual amount of water received and the calculated daily water requirement;

a block for issuing recommendations to the user;

and

the data acquisition unit from the sensors is connected to the data pre-processing unit, in which

- the subunit for determining the amount of lost water and the subunit for determining the amount of received liquid are connected to the subunit for determining the actual amount of water received,

- a sub-unit for determining the actual amount of water received is connected to a block for comparing the actual amount of water received and the calculated daily water requirement,

- the subunit for determining the level of physical activity is connected to the block for calculating the daily need for water;

the block for collecting individual data is connected to the block for calculating the daily need for water;

the unit for calculating the daily water requirement is connected to the unit for comparing the actual amount of water received and the calculated daily water requirement;

the unit for comparing the actual amount of water received and the calculated daily water requirement is connected to the unit for issuing recommendations to the user.

Moreover, the proposed device may additionally contain a block for checking whether the user follows the recommendations connected to the data collection block. Moreover, an additional individual parameter is growth. Moreover, the recommendations to the user are recommendations regarding the optimal consumption of water. Moreover, the recommendations to the user are a warning about the dehydration of the user's body or the saturation of the user's body with water. Moreover, the device additionally contains one or more of the following sensors: a bioimpedance sensor that measures the impedance of the human body, a galvanic skin response sensor, an optical sensor, an electrochemical sensor, a sensor that provides information about the water content in the user's subcutaneous layer, a geolocation sensor.

Also, a method of operation of the proposed device for monitoring the hydration of a living organism is proposed, containing the steps at which:

a) place individual user data in the block for collecting individual user data;

b) the data from the sensors enter the block for collecting data from the sensors and go to the block for preliminary data processing;

c) in the data pre-processing block, the following values are evaluated for a certain period of time:

- the amount of water lost by the body, by means of a sub-block for determining the amount of water lost,

- the amount of fluid ingested by the body, by means of a sub-unit for determining the amount of fluid ingested, wherein the amount of fluid ingested by the body is taken equal to the amount of water ingested by the body,

- the actual amount of water taken by the body, by means of a sub-block for determining the actual amount of water taken,

- the level of physical activity, through the sub-block of determining the level of physical activity;

d) calculating the daily need of the user's body for water based on the level of physical activity and individual data of the user in the block for calculating the daily need for water;

e) calculate the difference between the actual amount of water received and the calculated daily water requirement in the block for comparing the actual amount of water received and the calculated daily water requirement;

f) based on the calculated difference between the actual amount of water taken and the calculated daily water requirement, at least one individual recommendation is made to the user regarding the amount of water the user should consume;

and steps (b)-(d) occur continuously.

Moreover, the issuance of recommendations is a display of recommendations on the screen. At least one recommendation to the user is issued at the request of the user. Moreover, the amount of water lost by the body is the amount of water lost with sweat over a certain period of time. Moreover, the assessment of the level of physical activity is based on the number of steps taken per day. Additionally, the current state of the water balance in the user's body is output, which is data on the amount of fluid taken and the amount of water lost. Moreover, the frequency of issuing at least one individual recommendation to the user is determined by the user. Recommendations are continuously updated. Moreover, recommendations can be issued when a certain level of lack of water in the body is reached or when the state of dehydration of the body approaches. Moreover, with an insignificant imbalance of water in the body, which is less than 1% of the user's body weight, in the case of physical activity, the user receives a recommendation in the form of the current balance of water in the body and a recommendation on the amount of water that needs to be taken. With a significant imbalance of water in the body within 1% - 2% of the user's body weight, with a high physical activity of the user, the user receives a recommendation in the form of the current balance of water in the body and a recommendation on the amount of water that needs to be taken, while the activation stage is additionally performed a counter that displays to the user the decrease in the amount of water in the body. In case of dangerous dehydration of the body of more than 2% of the user's body weight, the user receives a warning about dehydration of the body and a recommendation to urgently take a certain amount of water. The level of physical activity can be one of low, medium, high. Additionally, the user's headphones are used, connected to the device for monitoring the hydration of a living organism, and microphones are built into the headphones that record sounds around the user, while the method additionally contains the stage at which, to determine the amount of liquid taken, analyze the sounds corresponding to the swallowing of the liquid. Moreover, the amount of fluid taken by the body is determined according to the data of the accelerometer and gyroscope, which fixes the user's gestures. The water lost amount determination sub-unit determines the amount of water lost by the body based on user activity periods based on sensor data. Moreover, the subunit for determining the amount of received liquid determines

the amount of liquid received by detecting the gesture of drinking and / or eating the user's hand on which the smart bracelet is worn, and the gesture of drinking and / or eating is determined by characteristic signals received from the gyroscope and accelerometer, which is responsible for the movement of the specified hand of the user when the user makes the gesture drinking and/or eating, and the amount of fluid taken is estimated by the time during which the gesture of drinking and/or eating continued. Moreover, the subunit for determining the amount of the received liquid determines the amount of the received liquid by detecting the microgesture of drinking and/or eating the user's hand, which is not wearing a smart bracelet, and the microgesture of drinking and/or eating is determined by the characteristic signals received from the gyroscope and the accelerometer responsible for the micro - movements of said user's hand when the user performs a microgesture of drinking and/or eating, wherein the amount of liquid taken is estimated by the time during which the microgesture of drinking and/or eating continued. Moreover, to determine the amount of liquid taken, a given dependence of the amount of liquid drunk on the total number of performed gestures and/or a given dependence of the amount of liquid drunk on the total duration of the gesture are used. In the subblock for determining the amount of received liquid, the time sequences of signals from the accelerometer and gyroscope are converted into a vector representation that describes the user's gestures at any time, based on the vector representation, using a regression model trained on a set of data sets, the gestures of receiving liquid are detected and the amount of received liquid is estimated. The regression model is trained to give more weight to the vector representations associated with the user's sip.

A smart watch containing the proposed device for monitoring the hydration of a living organism is also provided. Moreover, smart watches are connected to the “smart home” infrastructure.

A computing device is proposed that contains a processor and a memory that stores instructions for performing the steps of the proposed method for operating a device for monitoring the hydration of a living organism.

Also provided is a computer-readable medium storing instructions for causing a computing device to perform the steps of the proposed method for operating a living organism hydration monitoring device.

Brief description of the drawings

The above and other features and advantages of the present invention are explained in the following description, illustrated by drawings, in which the following is presented:

Fig. 1 illustrates a flow diagram of a method and apparatus for monitoring human body hydration.

Fig. 2 illustrates a user's fluid intake assessment.

Fig. 3 illustrates a fluid intake gesture detection model for a user.

Fig. 4 illustrates the training of a user's fluid intake gesture detection model.

Fig. 5 illustrates the operation of the block responsible for the ability of the model to use information only from specific gestures to predict the final result.

Fig. 6 schematically illustrates the general scheme for determining any user activity.

Fig. 7a illustrates the dependence of the amount of liquid drunk on the total duration of the gesture.

Fig. 7b illustrates the dependence of the amount of liquid drunk on the total number of gestures performed.

Fig. 8 illustrates an example of personalized recommendation notifications that a user may receive throughout the day.

Detailed description of the invention

A method for assessing hydration (saturation of the body with water) of the user is proposed. With the invention, it is possible to give individual recommendations to the user regarding optimal water intake, it is also possible to warn the user that the body is dehydrated or close to it, or that the user is drinking too much liquid. The proposed device provides high accuracy for users with any level of physical activity, and does not impose restrictions on the daily activities of the user. It also provides the ability to detect and count user gestures related to fluid intake. EFFECT: invention provides automatic monitoring of fluid intake into the body (events and assessment of fluid volume) by processing data from smart watch inertial sensors. The invention provides an automatic assessment of water loss due to sweating during sports and everyday life by processing data from inertial, optical and temperature sensors of smart watches. The invention monitors the hydration balance of the user's body and makes individual recommendations to the user regarding the amount of water needed for consumption in accordance with observed sweat losses and fluid intake. Thus, a fully automatic user-independent (no additional action is required) hydration balance monitoring device for smartwatch is proposed, which gives the user individual advice in a timely manner to prevent both dehydration and satiation of the body with water. In addition, saturating the body with water is just as dangerous as dehydration. When the body is oversaturated with water, the body temperature drops, salivation begins, nausea, vomiting, impaired coordination of movements, convulsions, muscle weakness, and headache appear.

It should be noted that a person does not always drink just water, and when eating, not pure water enters the body, however, taking into account the error, it is possible to assume that the amount of liquid that enters the body with food and drink is almost equal to the amount of water that enters the body. the body together with food and drink, the error from the mentioned assumption is small and is neglected in the scope of the present invention. In other words, when calculating the water balance, the amount of fluid taken in by the body is taken equal to the amount of water taken in by the body.

The invention can be useful not only for users engaged in strenuous physical activity, but also for any other users, including those leading an inactive lifestyle.

The present invention may be used in any suitable user computing device comprising a processor and a memory storing instructions for the processor to perform the steps of the proposed method. Such a device may be, but is not limited to, a smartphone, a smart watch, or other suitable device, hereinafter referred to as a smart bracelet. The memory may be any storage medium, in particular a computer-readable storage medium.

The invention is based on the analysis of data from sensors used in a smart bracelet. As mentioned above, the sensors can be inertial, optical and temperature, it is the presence of these sensors that is necessary and sufficient for the implementation of the proposed invention. Additionally/optionally, the following sensors can be used: a body resistance sensor (bioimpedance sensor) that measures the impedance of the human body, a galvanic skin response sensor; electrochemical sensors, and other sensors directly or indirectly related to the balance of water in the body. Additional sensors can serve to more accurately assess body hydration.

A machine learning algorithm is used, which, according to the data received from the sensors, produces:

assessment of water loss along with sweat, during physical exertion;

detecting user activity such as eating, in particular fluid consumption, by detecting individual gestures, and quantifying the volume of fluid consumed by the user.

Based on the aforementioned ratings, the user receives personalized recommendations for water consumption in accordance with the needs of the user's body at the moment. Also, when forming recommendations, it is necessary to take into account the age, gender of the user, his weight and height.

Thanks to the proposed invention, a new function is introduced into the smart bracelet, determining the balance of hydration of the user's body, which includes:

automatic assessment of perspiration, that is, the amount of water that is gone with the release of sweat;

automatic determination of the user's fluid intake with an estimate of the amount, that is, the user sees how many milliliters of water have been drunk;

providing a personal recommendation to the user on water consumption in accordance with the physiological needs of the user.

The proposed system is constantly trained and tested, that is, the database to which the device is connected is constantly updated, which leads to the issuance of more accurate recommendations to the user.

For the algorithm to work correctly, a large amount of data is needed, collected from people with different individual parameters, in addition, the conditions in which the subjects are located should also differ from each other. That is, to collect data in a variety of conditions, it is necessary to change the ambient temperature, humidity. To do this, some subjects run on a treadmill during data collection, while artificially creating various climatic conditions by changing the temperature and humidity of the environment. Other subjects run down the street in various weather conditions. Also, the subjects can take liquid at the table while sitting or standing, the height of the table and the height of the seats can also be taken into account. In addition, different subjects may ingest liquid with the hand wearing the smart band or with the other hand without the smart band. All of the above are just examples of what may be considered when collecting data to train the algorithm of the present invention.

When collecting data, the following necessary sensors are used: an accelerometer, a gyroscope, a pulse sensor (plethysmograph) and a temperature sensor. For each subject, the duration of physical activity is recorded. The signals received from the sensors are preprocessed using a Gaussian and median filter to remove outliers that do not carry the main information. Based on the data obtained from the signals, a set of features is calculated, which can be divided into statistical ones, which represent the total information about the signal or part of the signal of the change in value over time, and frequency, which are the harmonic changes in all or part of the signal for some or the entire period time. The signs that best correlate with the loss of water are determined and, therefore, if they are taken into account correctly, they make it possible to assess the loss of fluid.

Further, according to the collected data and calculated features, a regression model is trained (for example, “Boosting on trees”), the hyperparameters of which are at least the absolute error in cross-validation. After finding the hyperparameters, the regression model is trained on all the data.

The trained regression model, when working in a smart bracelet, can give individual characteristics to the user, based on the data that will be received from a particular user. To issue an individual recommendation, the water loss is estimated using a trained regression model .

In one of the implementations of the proposed invention, it is possible to use the following sensors: accelerometer, gyroscope, pulse sensor (plethysmograph), temperature sensor. To determine the characteristics of the user's individual water metabolism, in addition to the data obtained from these sensors, it is necessary to take into account the user's gender, age, height and weight, skin surface area, which is calculated from height and weight indicators.

In FIG. 1 shows a block diagram of a method and apparatus for monitoring human body hydration.

As shown in FIG. 1 living body hydration monitoring device contains:

a block 1 for collecting individual user data, including gender, age, weight;

sensors (not shown in Fig. 1), which are:

accelerometer,

gyroscope,

pulse sensor,

temperature sensor;

block 2 for collecting data from sensors;

data pre-processing block 3 containing

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

sub-block (b) determining the amount of lost water,

subblock (c) determining the amount of fluid taken,

subblock (d) determining the actual amount of water received;

block 4 for calculating the daily water requirement;

block 5 for comparing the actual amount of water received and the calculated daily water requirement;

block 6 issuing recommendations to the user;

and

block 2 for collecting data from sensors is connected to block 3 for preliminary data processing, in which

- the sub-unit (b) for determining the amount of water lost and the sub-unit (c) for determining the amount of received liquid are connected to the sub-unit (d) for determining the actual amount of water received,

- subunit (d) for determining the actual amount of water received is connected to the block 5 for comparing the actual amount of water received and the calculated daily water requirement,

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

block 1 for collecting individual data is connected to block 4 for calculating the daily need for water;

unit 4 for calculating the daily water requirement is connected to the unit 5 for comparing the actual amount of water received and the calculated daily water requirement;

block 5 for comparing the actual amount of water taken and the calculated daily water requirement is connected to block 6 for issuing recommendations to the user.

Moreover, recommendations to the user relate to the optimal consumption of water. If the body is found to be on the verge of dehydration or water oversaturation occurs, then the advice to the user is a warning 7 of the user's dehydration or water oversaturation of the user's body.

The calculation of the daily water requirement is made according to:

percentage (%) of body weight,

skin surface area (ml / m ^2),

sweat consumption ml/kcal or ml/kg body weight

Taken into account:

- Age and gender,

- The level of physical activity,

- Basal metabolic rate,

- Environmental factors.

The method for monitoring the hydration of the human body is as follows. The user puts on and turns on a smart bracelet, which may be a smart watch, or any other suitable device attached to the user's body. Hereinafter, for convenience, but not limitation, such a device is referred to as a "smart bracelet". The smart bracelet contains at least the following sensors: accelerometer, gyroscope, pulse sensor (plethysmograph), temperature sensor. To take into account environmental conditions, the device may contain sensors that measure, for example, ambient temperature, humidity, and the like.

In order for the user to use the proposed function of the smart bracelet, the user must enter their individual data regarding gender, age, height, weight, which are placed in block 1 for collecting individual user data, without these data the user will not be able to receive correct recommendations.

The smart bracelet starts collecting data using the indicated sensors, the collected data enters the block 2 for collecting data from the sensors, from where it goes to the block 3 for preliminary data processing, where, based on the readings of the sensors, the following values are calculated using algorithms:

- The amount of water lost by the body in sub-block (b) of determining the amount of water lost. The trained machine learning regression model as described above predicts the amount of water lost through sweat over a certain period of time, for example, during an exercise period or for the whole day, this occurs in sub-block (b) of determining the amount of water lost.

- The amount of fluid taken by the body, these data are determined in sub-block (c) determining the amount of fluid taken, and the amount of fluid taken by the body is taken equal to the amount of water taken by the body.

- The actual amount of water ingested by the body, in sub-block (d) determining the actual amount of water ingested, based on the amount of fluid taken in by the body and the amount of water taken in by the body.

- The level of physical activity, determined in sub-block (a) of determining the level of physical activity. Based on the data of inertial sensors (accelerometer and gyroscope), the level of general physical activity of a person is estimated, for example, by the number of steps taken per day, this is necessary to adjust the daily need of the user for water, as a rule, the activity level is divided into 3 classes: low, medium, high, and add 700 ml to the daily norm if a particular person is characterized by an average level of activity and 1100 ml if high. These data fall into sub-block (a) of determining the level of physical activity.

Based on the reconnaissance analysis, signs were found that correlate with the estimated water loss. To estimate water loss, correlated features are calculated and fed to the input of a trained regression model, which evaluates water loss in ml.

Based on the data on the level of physical activity and individual data of the user in block 4 for calculating the daily water requirement, the daily water requirement of the user is calculated.

Further, the calculated data, namely the amount of water lost by the body and the amount of fluid (that is, water) taken by the body, are converted into actual data on the amount of water taken by the user. And then the actual amount of water taken and the calculated daily water requirement are compared in the block for comparing the actual amount of water taken and the calculated daily water requirement, the difference between the actual amount of water taken by the body and the calculated daily water requirement of the body is calculated.

Based on the calculations made, a recommendation is issued regarding the amount of water that the user should drink, for example, before the end of the current day. The recommendation can be updated continuously, and the recommendation can be issued when a certain level of lack of water in the body is reached or when a state of dehydration is approaching.

If the amount of liquid received by the body is less than the critical threshold for water demand, the user, along with information about the amount of water to be taken, receives a warning about the likelihood of dehydration (dehydration).

If the amount of fluid received by the body is greater than the critical threshold of daily water requirement, the user, along with information about the amount of fluid, receives a warning about excessive fluid intake.

After giving a recommendation on the amount of liquid, the device continues to collect and calculate data, so it turns out that the device monitors whether the recommendations are met. In the preferred implementation - the frequency of repeated notifications can be set by the user. At the same time, the user can check the current state of the water balance at any time.

Thus, comprehensive monitoring of body hydration is carried out.

Obviously, there is a certain correlation between the received data and user behavior. Different users, depending on gender, age, weight and height, have certain levels of physical activity. To cluster users into groups, a regression model is used (for example, "Tree Boosting"), which breaks down the signs for each subgroup of users (weight, height, age), thus giving values not only based on the sensors (sensors) of the smart bracelet, but also using the data about the user that he enters. Based on this, features are selected that maximally generalize the model for all users.

Based on the data from the accelerometer and gyroscope sensors, it is possible to distinguish the drinking gesture from all other user gestures.

By detecting the time spent on the drinking gesture, it is possible to determine the amount of liquid the user has drunk.

Consider monitoring the hydration of the human body in detail.

The first factor for assessing the user's water balance is the assessment of the amount of water that has come out with sweat.

Based on the signals received from the accelerometer and gyroscope sensors, heart rate sensor, temperature sensor, it is possible to determine periods of user activity. Obviously, when the user starts to move, his speed, pulse and temperature begin to increase.

Features are distinguished that can be divided into statistical ones, which are summary information about the signal or part of the signal of the change in value over time, and frequency ones, which are harmonic changes in the whole or part of the signal for some or the entire period of time. After compiling all the features, the features that best correlate with the loss of water are determined and, therefore, if they are taken into account correctly, they allow estimating the loss of water. When estimating water loss, features are calculated that correlate with the estimated water loss and are fed to the input of the regression model, which produces an estimate of water loss.

The signals from the sensors enter the block 2 for collecting data from the sensors, then they are processed in the block 3 for preliminary data processing, where the characteristic features that most affect the assessment of water loss are selected from the set of signals, it is the signals that have the characteristic features that are transmitted to the sub-block (b ) estimating the amount of water lost, which uses a regression model based on machine learning that estimates the amount of water lost. The characteristic features of the sensor signals are determined at the stage of machine learning, therefore, during operation, the trained regression model will select signals that have characteristic features. The accelerometer signals provide information about the user's speed as well as the number of steps per minute. The plethysmogram provides information about the number of heart beats per minute. The temperature sensor provides information about the temperature of the body and the environment. The data is filtered using a median filter and a Gaussian filter. Further, the signals are normalized and characteristic features related to the loss of water by the body are selected.

There is a training sample and a test sample, which is split in such a way that there are no gaps in the distribution of height, weight, sex, and age. Next, all the features are calculated, then the features with the highest correlation between themselves are removed, since the correlation of random variables determines the linear change in one variable when another random variable changes. If the features are strongly correlated with each other, then their contribution will be almost the same, so you can throw out all strongly correlated features except for one, that is, features that do not correlate with each other get into the regression model. To this number of features in the regression model, features are iteratively added until the cross-validation error starts to increase. It should be clarified that during cross-validation, the sample is divided into N samples - training and validation - the algorithm is trained on the training sample and validated on the validation sample, after which the error on each of the validation samples is calculated, thus obtaining N estimates on different validation samples. After that, the average error is taken over all N errors on the validation partitions. This experiment is repeated for different regression models, with different trained regression models differing only in different parameters, such as the number of trees and the maximum depth of the tree. After calculating the cross-validation errors on different regression models that differ in parameters, the best model is selected by the cross-validation error, after which the regression model is determined, the error of which is not only the smallest, but also closest to the error on the test sample, and the test sample is formed from a large data set, about 25-30% of the entire data set is selected, which allows you to choose the best model in terms of model parameters.

- вектора реальных потерь воды и оцененных потерь воды соответственно, эти вектора принадлежат множеству действительных чисел. Тогда,

- среднее значение реальных потерь воды,

- тотальная квадратичная сумма,

- сумма квадратичных отклонений,

- коэффициент детерминации (R2), который и производит оценку качества алгоритма оценивания потери воды.The algorithm generates a calculation of the water loss estimate normalized to the body surface area. Let

- vectors of real water losses and estimated water losses, respectively, these vectors belong to the set of real numbers. Then,

- average value of real water losses,

is the total quadratic sum,

- sum of square deviations,

- coefficient of determination (R2), which evaluates the quality of the water loss estimation algorithm.

Further, the obtained value is multiplied by the body surface area and the calculated value of the amount of water loss excreted with sweat is obtained.

In order to obtain a better parameter determination accuracy, the living body hydration monitoring device can be additionally equipped with a sensor that provides information on the water content of the skin based on the optical scattering characteristics of the skin. Such a sensor can operate in both the visible and near infrared regions of the spectrum and collects data on the water content in the subcutaneous layer of the user.

Also, the proposed device can be additionally equipped with an impedance sensor to measure the water content in the whole body. To do this, the first electrode is located on the lower part of the smart bracelet adjacent to the skin, the second electrode is isolated and located on the body of the smart bracelet. To measure the impedance, the user must touch the second electrode with the hand that is not wearing the smart bracelet.

Also, the proposed device can be equipped with a galvanic skin response sensor. Two electrodes of such a sensor can be located at the bottom of the smart bracelet. Such a sensor measures the local sweating in a specific area. Based on the data from this sensor, a conclusion is made about how the pores of the whole body work.

Another sensor that can be additionally equipped with the proposed device is a chemical sensor containing microfluidic channels through which the user's sweat passes. The sensor analyzes the chemical composition of the electrolytes in the user's sweat. By changing the ratio of electrolytes, a conclusion is made about the state of hydration of the whole organism.

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

As an additional source of data, it is proposed to use smart scales that transmit information to a smart bracelet. Moreover, a sensor for determining bioimpedance can be built into smart scales.

An additional source of data can also be various applications installed on the smartphone associated with the smart bracelet.

And also an additional source of data can be other wearable devices of the user, for example, headphones, which are connected to a smart bracelet and / or smartphone wirelessly. Temperature sensors, pulse sensors, microphones can be built into the headphones to determine the process of taking liquids and food.

Also, an additional source of data can be the "smart home" infrastructure to which the user's smart bracelet is connected. It is possible to connect the smart bracelet wirelessly to the equipment in the fitness center where the user is exercising, and thus receive additional information from the equipment.

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

If the user performs a drinking gesture with the hand on which the smart bracelet is worn, then the gyroscope and accelerometer captures the movement signals of this hand. An example of the signals recorded by these sensors during fluid consumption is shown in Figure 2. The upper two graphs display the signals recorded by the device worn on the active hand, and the two lower graphs - on the inactive hand. Each of the graphs shown contains three components corresponding to linear accelerations and angular velocities relative to the three sensor coordinate axes (x, y, z). Rectangular areas on the graphs mark the areas of gyroscope signals corresponding to the drinking gesture and having a characteristic shape for it. Such repetitive fragments of signals are observed both on a smart bracelet worn on the hand, with which the user brings a vessel with liquid to his mouth, and on a second smart bracelet located on an inactive hand.

Signals for moving the hand specifically towards the mouth for drinking are different from those for moving the hand towards the mouth for eating or performing hand-raising exercises or other gestures. Each of the listed movements has its own specifics, for example, eating gestures are accompanied by pricking or scooping up food, drinking gestures are characterized by their own turn of the hand and a longer holding of the hand at the mouth, etc., the algorithm trained on many examples can recognize various movements , even if the user tries to simulate a drinking gesture with an empty hand, the algorithm recognizes that it was not a drinking gesture, and evaluates both the occurrence of the drinking gesture event itself and the duration of this gesture.

If the user performs a drinking gesture with a hand that is not wearing a smart bracelet, the accelerometer and gyroscope capture micro-movements of the inactive hand, corresponding to the movement of the whole body when making a drinking gesture with the active hand. In figure 2, the micromovements of the inactive hand and the movements of the active hand are marked with rectangles, along the X axes time is plotted, along the Y axes the level of the accelerometer and gyroscope signals, respectively.

As an example, consider the gesture that a person performs in the process of one sip. As a rule, this gesture begins with raising the hand with the vessel to the mouth. At the beginning of this gesture, the gyroscopic sensor captures the acceleration of the hand. At the moment of the end of the movement, when the vessel with liquid is near the human mouth, deceleration is observed, which manifests itself in the form of a signal of a characteristic shape in the measurements taken from the gyroscope. The downward movement of the hand with the vessel also has characteristic features that make it possible to detect gestures of fluid intake and to evaluate its volume.

The algorithms for detecting the event of fluid intake and estimating its volume are based on a neural network model that simultaneously solves both the problem of detecting a gesture of fluid intake and the problem of estimating the volume of this fluid. Accelerometer and gyroscope signals are pre-processed before they are fed to the input of the specified model. Such processing may include digital filtering to remove noise, decimation, scaling, normalization, and other such transformations.

A schematic representation of the neural network model for determining the drinking gesture is shown in Fig. 3.

The algorithm allows you to detect the gesture of receiving a liquid. By such a gesture, we mean the movement during which a person takes a sip. In machine learning, a model is understood as a functional transformation from input data to an estimated result, as well as a set of parameters necessary to implement this transformation.

During the operation of the proposed device, the accelerometer continuously measures the direction and magnitude of the acceleration vector, which includes the free fall acceleration, and the gyroscope measures the angular acceleration. The sequences of measurements from these sensors contain information about the trajectory of the device, and, consequently, the hand on which they are worn. This trajectory, in turn, describes the movements of a person.

In subblock (c) for determining the amount of liquid received, the time sequences of signals from the accelerometer and gyroscope are converted into a vector representation that describes what gestures the user is making at any moment of measurement. Next, the sequence of vector representations is analyzed.

The resulting sequence of vector representations can be considered as an input signal for the next functional level of the model. At the next functional level of the model, a decision is made which of the gestures described by the corresponding vector representation refers to the event of fluid intake. Part (a), in Fig. 3 is divided into two blocks. The left block is responsible for detecting the gesture itself, at the output of this block, data is received indicating the time period in which the person made the gesture related to drinking. The right block serves to estimate the amount of liquid that was drunk as a result of the perfect gesture. The assessment of the volume of the drunk liquid is performed on the basis of the duration of the gesture, the angle of inclination of the vessel with the liquid.

Part (b) of figure 3 shows the calculation of the amount of liquid taken by the user. The input of the neural network model, or rather its first layer, is a sequence of preprocessed measurements (the preprocessing operation is described above). An encoder is a sequence of convolutional layers, each of these layers performs an operation of convolution of a set of input signals with a set of kernels belonging to the layer. The convolution operation mixes the input signals according to the ratios given by the weights that describe each of the indicated kernels.

As a result of the successive application of the described operations for each of the encoder layers, a set of time series is formed at its output. The set of values of these time series for a particular point in time is a numerical description of the movements that a person carried out in a small period of time. This set of values is essentially a vector representation of such movements.

The set of time series generated at the output of the encoder is fed to two blocks that solve the problems of detecting a gesture of receiving a liquid during which a sip is performed and estimating the volume of this sip. In one embodiment, these blocks are built on the basis of recurrent layers (LSTM or GRU). Such layers in the process of model operation store a state vector that characterizes the time sequence of signals fed to the input of the layer. That is, this state vector describes a sequence of gestures that a person performed in a short time interval preceding the current gesture.

Fig. 4 illustrates the training of a user's fluid intake gesture detection model. The input of the model (b) is the data (a) of the accelerometer and gyroscope, which is a sequence of vectors containing 6 measurements for each moment of time: 3 numbers measured by the accelerometer, describing the linear acceleration vector, and 3 numbers measured by the gyroscope, specifying the angular velocity . A sequence of measurement sets is formed in such a way that the delay between two successive measurements is constant; if necessary, signals are interpolated to ensure this condition. Time is not explicitly transferred to the model, while the time is encoded as a sequence of sensor measurements.

The main task of the proposed model is to continuously estimate the amount of liquid drunk by a person. This property of the model is provided as a result of learning - the process of solving an optimization problem of finding model parameters, which are understood as weight coefficients in functions that describe the transformation over input variables. Upon completion of the learning process, the parameters are chosen so that the minimum value of the error function calculated on the data set used is achieved. An important feature of the model training process proposed in the present invention is that, in the course of searching for the necessary parameters of the regression model, the solution of two problems is optimized, consisting in the detection of a fluid intake gesture and the assessment of the volume of fluid intake. In the process of searching for the parameters of this model on the training data set, information is used not only on the volume of liquid, but also on the time intervals of gestures, as a result of which a person takes a sip. An error function is constructed that contains two components, the first of which characterizes the quality of detection of gestures containing a sip, and the second characterizes the quality of the estimate of the volume drunk. During optimization, errors on both tasks are taken into account. This approach greatly simplifies the learning process, thereby making it technically feasible.

The proposed approach to learning allows us to find such model weights that provide high quality both for solving the problem of classifying gestures into those related to drinking and others, and for estimating the amount of liquid drunk as a result of one gesture.

A large number of datasets are used in the model training process, since each person makes the drinking gesture differently, and the vessels from which the liquid can be taken can also vary greatly.

The proposed model can be modified in several ways. One way to increase the accuracy of the estimation is to add a block to the model that is responsible for the ability to use information only from specific gestures to predict the final result, as shown in Fig. 5. In other words, thanks to the applied block, the model gains the ability to pay attention to specific features of the signals in the entire temporal context. As shown in Fig.5, when receiving a segment of the signal change at a certain time interval to the input of the model with the mentioned block, the model uses information from various points of the specified segment.

In other words, the estimation accuracy of the model can be improved by adding an attention block to it. The presence of this block will allow the model to assign more weight to the vector representations of individual gestures, both associated with the sip itself and auxiliary. Such gestures may contain information about the type and volume of the vessel used in the process of drinking, which makes it possible to increase the accuracy of estimating the amount of liquid drunk. Also, as studies show, the attention mechanism allows the model to store a longer temporal context (the history of actions that the user performed). Obviously, this also has a positive effect on the quality of the model.

By supplementing the input signal processing algorithm, it is possible to determine various other user gestures besides the drinking (liquid taking) gesture. Fig. 6 schematically illustrates the general scheme for determining any user activity. That is, with the help of an accelerometer and a gyroscope, as well as a regression model trained on a large number of examples with different features, it is possible to classify user actions based on the detection of motion patterns. In this way, it is possible to recognize various user activities, for example, but not limited to, a gesture of eating, or, for example, it is possible to determine whether the user is drinking hot or cold, this is determined, for example, by the consumption rate. Also, by modifying the model, it is possible to determine the moments of the user's life when he is definitely not in the process of taking liquid. Also, the model can determine what the user eats and the approximate fluid content of their food to account for the amount of fluid received with the meal while balancing the amount of fluid taken. It is also possible to use the user's headphones, which are connected to the smart bracelet wirelessly, microphones are built into the headphones, which capture the sounds around the user while the user is wearing the headphones, in order to detect cases of liquid consumption by analyzing sound patterns corresponding to, for example, swallowing a liquid. That is, for detecting fluid intake, it is possible to monitor the movement characteristics of the user and the sounds around the user. Certain movements that the trained model recognizes as drinking are recorded, and the amount of fluid consumed by the user is estimated from their duration in time, as well as taking into account the individual parameters of the user. It should be noted that it is precisely the assessment of the amount of liquid that is made, and not an exact calculation.

Additional data sources to improve the accuracy of estimating the amount of fluid consumed by the user can be data from other user devices, such as smartphones, smart watches, smart home infrastructure.

It is also possible to use an additional algorithm that determines on which hand the smart bracelet is worn, on the dominant one, which a person uses more often, or not.

The following describes an illustrative example of estimating the amount of fluid drunk for a single user. Fig. 7a illustrates the dependence of the amount of liquid drunk on the total duration of the gesture. Fig. 7b illustrates the dependence of the amount of liquid drunk on the total number of gestures performed. That is, the volume of the drunk liquid can be determined quite accurately even by these two dependencies, without using information about the shape of the signals that describe the trajectory of gestures. An example is given for a simple algorithm with only two features. If we complicate the algorithm, for example, using additional features from the trajectory of the hand movement, it is obvious that a smaller error can be achieved.

Using only this information, it is possible to obtain a result with an error of the order of 10-15% for a particular user.

In the case of a negative hydration balance, i.e. when the volume of liquid consumed by the user is less than the volume of water excreted by the user with sweat, there are three scenarios:

A) Minor water imbalance in the body, i.e. the water imbalance is less than 1% of the user's body weight. In case of low physical activity of the user, there is no need to notify the user. If the user is physically active, the user receives a notification in the form of the current balance of water in the body and recommendations for water intake.

B) Significant imbalance of water in the body, that is, the imbalance of water is within 1% - 2% of the user's body weight. In case of low physical activity of the user, the user receives a one-time notification about the current balance of water in the body. If the user is physically active, the user receives information about the current balance and a notification about the need to drink water. With high physical activity in the smart watch, a counter may turn on, which reflects a decrease in the amount of water in the body.

C) Dangerous dehydration of the body, that is, when the water imbalance is more than 2% of the user's body weight. In case of low physical activity of the user, the user receives a notification with an urgent recommendation to take a certain amount of water. If the user is physically active, the user is notified of the required amount of water to drink, but here it is necessary to take into account that a single large portion of water can be harmful.

Fig. 8 illustrates an example of personalized recommendation notifications that a user may receive throughout the day. For example, the user is a male, age 30, weight 75 kg, level of physical activity - average. Height is an optional parameter, since the body's need for water is usually estimated based on weight with different coefficients for the age category and for the level of physical activity. The level of physical activity of the user is determined by the average number of steps taken per day and the average duration of increased physical activity per day (running, other sports / fitness).

In this example, when the user wakes up, the smart bracelet gives a notification with a recommendation for the estimated daily water requirement, in this case: “You should drink at least 2800 ml of water per day. Drink 250 ml of water." After the user has gone in for sports, the smart bracelet, based on the data assessed during the day on the volume of fluid consumed, on the volume of water coming out with sweat, including after training, issues a notification that contains the following information: “Drank 400 ml liquids. Lost 550 ml of water. Drink 450 ml of water." During office work, based on the data assessed during the day on the volume of fluid consumed, on the volume of water coming out with sweat, the smart bracelet issues a notification: “Drank 1400 ml of liquid. Drink 200 ml of water." At the end of the day, the smart bracelet can issue the following notification: “Drank 3200 ml of liquid. Well done! Today you received a health benefit! All notifications are provided to the user on the screen of the smart bracelet, depending on the data received.

A person without a smart bracelet is guided only by the feeling of thirst, while a non-athletic person who decides to run for the first time probably will not drink more water than he used before, since this is not part of his habit. If such a person drinks too little water, they may be at risk of dehydration and experience symptoms such as dizziness, nausea, and other gastrointestinal problems. If a person drinks too much water, then problems and health risks arise, in particular, if the water does not have time to be consumed by the body, this is a big burden on the kidneys. If a person is a user of the proposed smart watch, then already during the run he will be given information about the lack of water in the body, and in case of a risk of dehydration, the smart bracelet will trigger a signal that draws the user's attention to dehydration.

It is possible to implement the invention when the assessment of water loss is not performed, but only the assessment of the amount of received liquid is performed. In this case, the user can also receive personalized recommendations on water consumption, however, the water balance monitoring algorithm will work with low accuracy.

If the user has not entered their individual data regarding gender, age, weight and height, in this case the system will still function. For example, the amount of sweat in this case can be estimated from the duration and intensity of physical activity. At the same time, the accuracy of estimating water losses will also be reduced.

It should be noted that within the scope of the present invention, when monitoring body hydration, it does not make sense to take into account the excretion of water in the urine. The kidneys maintain the internal balance of water in the body, if necessary, retain water or excrete excess water in the urine. At the same time, being already in the bladder and when excreted from the body, urine does not affect the hydration of body tissues in any way. That is, the need for water and the excretion of urine from the body is a natural background process of a healthy person. While the proposed invention is intended to detect the following events:

- the user drank very little for a long time;

- the user had a high physical load and sweated a lot of water (more than he drank);

- the user drank too much water and his kidneys may not be able to cope with such a load.

In the events under consideration, it does not matter how often and how much urine is excreted from the body.

Although the invention has been described in connection with some illustrative embodiments, it should be understood that the invention is not limited to these specific embodiments. On the contrary, the summary is intended to include all alternatives, corrections, and equivalents that may be included within the spirit and scope of the claims.

In addition, the invention retains all equivalents of the claimed invention, even if the claims are changed in the course of consideration.

Claims (65)

1. A user hydration assessment device, comprising: a unit for collecting individual user data, including gender, age, weight; sensors that are: accelerometer, gyroscope, pulse sensor, temperature sensor; block for collecting data from sensors; data preprocessing block containing: subblock for determining the level of physical activity, subunit for determining the amount of water lost with sweat, subunit for determining the amount of liquid taken, a sub-block for determining the actual amount of water taken, based on the amount of fluid taken by the body and the amount of water lost by the body; block for calculating the daily need for water; a block for comparing the actual amount of water received and the calculated daily water requirement; a block for issuing recommendations to the user; and the data acquisition unit from the sensors is connected to the data pre-processing unit, in which - the subunit for determining the amount of water lost with sweat and the subunit for determining the amount of liquid taken are connected to the subunit for determining the actual amount of water taken, - a sub-unit for determining the actual amount of water received is connected to a block for comparing the actual amount of water received and the calculated daily water requirement, - the subunit for determining the level of physical activity is connected to the block for calculating the daily need for water; the block for collecting individual data is connected to the block for calculating the daily need for water; the unit for calculating the daily water requirement is connected to the unit for comparing the actual amount of water received and the calculated daily water requirement; the unit for comparing the actual amount of water received and the calculated daily water requirement is connected to the unit for issuing recommendations to the user. 2. The device according to claim 1, further comprising a unit for checking whether the user follows the recommendations, connected to the data collection unit. 3. The device according to any one of claims 1, 2, in which an additional individual parameter is height. 4. Apparatus according to claim 3, wherein the recommendations to the user are recommendations regarding optimal water consumption. 5. The device according to claim 1, wherein the advice to the user is a warning about the dehydration of the user's body or the saturation of the user's body with water. 6. The device according to claim 1, further comprising one or more of the following sensors: a bioimpedance sensor that measures the impedance of the human body, a galvanic skin response sensor, an optical sensor, an electrochemical sensor, a sensor that provides information about the water content in the subcutaneous layer of the user, a sensor geolocation. 7. A method for assessing a user's hydration, comprising the steps of: a) place individual user data in the block for collecting individual user data; b) data from sensors, which are an accelerometer, a gyroscope, a pulse sensor, a temperature sensor, enters the data acquisition unit from the sensors and is transmitted to the data pre-processing unit; c) in the data pre-processing block, the following values are evaluated for a period of time: - the amount of water lost by the body - by means of a sub-block for determining the amount of water lost with sweat, - the amount of liquid taken by the body - by means of a sub-block for determining the amount of liquid taken, and the amount of liquid taken by the body is taken equal to the amount of water taken by the body, - the actual amount of water taken by the body - by means of a sub-block for determining the actual amount of water taken, based on the amount of fluid taken by the body and the amount of water lost by the body; - level of physical activity - by means of a sub-block for determining the level of physical activity; d) calculating the daily need of the user's body for water based on the level of physical activity and individual data of the user in the block for calculating the daily need for water; e) calculate the difference between the actual amount of water received and the calculated daily water requirement in the block for comparing the actual amount of water received and the calculated daily water requirement; f) based on the calculated difference between the actual amount of water taken and the calculated daily water requirement, at least one individual recommendation is made to the user regarding the amount of water the user should consume; and steps (b)-(d) occur continuously. 8. The method of claim 7, wherein the recommendation is displaying the recommendations on a screen. 9. The method of claim 7, wherein at least one recommendation is provided to the user at the request of the user. 10. The method of claim 9, wherein the amount of water lost by the body is the amount of water lost through sweat over a period of time. 11. The method according to claim 7, in which the assessment of the level of physical activity is based on the number of steps taken per day. 12. The method of claim 7, further comprising providing the current state of the user's body water balance, which is data on the amount of fluid taken and the amount of water lost through sweat. 13. The method of claim 7, wherein the frequency of issuing at least one individual recommendation to the user is determined by the user. 14. The method of claim 7 wherein the recommendations are continuously updated. 15. The method of claim 7, wherein the advice may be given if the amount of fluid received by the body is less than a critical water requirement threshold, or when a dehydration state of the body is approaching. 16. The method according to any one of claims 7 to 15, in which, in case of an insignificant imbalance of water in the body, which is less than 1% of the user's body weight, in the case of physical activity, the user receives a recommendation in the form of a current balance of water in the body and a recommendation on the amount of water, which needs to be accepted. 17. The method according to any one of claims 7-15, in which, with a significant imbalance of water in the body within 1 - 2% of the user's body weight, with a high physical activity of the user, the user receives a recommendation in the form of a current balance of water in the body and a recommendation for the amount of water that needs to be taken, while additionally performing the step of turning on the counter, displaying to the user a decrease in the amount of water in the body. 18. The method according to any one of claims 7 to 15, wherein in case of dangerous dehydration of the body of more than 2% of the user's body weight, the user receives a warning about dehydration of the body and a recommendation to urgently take the required amount of water. 19. The method of claim 7, wherein the level of physical activity may be one of low, medium, high. 20. The method of claim 7, further comprising the user's headphones connected to the user's hydration monitoring device, the headphones having built-in microphones that pick up sounds around the user, the method further comprising analyzing the sounds to determine the amount of fluid taken. corresponding to the swallowing of liquid. 21. The method according to claim 7, in which the amount of fluid taken by the body is determined according to the data of the accelerometer and gyroscope, fixing the user's gestures. 22. The method of claim 7, wherein the sweat-lost-water-determining sub-unit determines the amount of water lost by the body based on user activity periods based on sensor data. 23. The method according to claim 7, in which the sub-unit for determining the amount of ingested liquid determines the amount of ingested liquid by detecting the drinking and/or eating gesture of the user with the hand on which the smart bracelet is worn, and the drinking and/or eating gesture is determined from the characteristic signals received from a gyroscope and an accelerometer responsible for moving said user's hand when the user makes a drinking and/or eating gesture, the amount of liquid taken is estimated by the time during which the drinking and/or eating gesture lasted. 24. The method according to claim 7, in which the sub-unit for determining the amount of ingested liquid determines the amount of ingested liquid by detecting the drinking and/or eating microgesture of the user with a hand that is not wearing a smart bracelet, and the drinking and/or eating microgesture is determined by characteristic signals, obtained from a gyroscope and an accelerometer responsible for micro-movements of said user's hand when the user performs a drinking and/or eating micro gesture, the amount of liquid taken is estimated by the time during which the drinking and/or eating micro gesture lasted. 25. The method according to any one of claims 7 to 15, wherein a predetermined dependence of the amount of liquid drunk on the total number of gestures performed and/or a predetermined dependence of the amount of liquid drunk on the total duration of the gesture is used to determine the amount of fluid taken. 26. The method according to claim 7, moreover, in the sub-block for determining the amount of received liquid, the time sequences of signals from the accelerometer and gyroscope are converted into a vector representation that describes the user's gestures at any time, gestures are detected based on the vector representation using a regression model trained on data sets fluid intake and estimate the amount of fluid taken. 27. The method of claim 26, wherein the regression model is trained to assign more weight to vector representations associated with the user's sip. 28. A smart watch comprising a user's hydration assessment device according to claim 1. 29. The smart watch of claim 28, wherein the smart watch is connected to a smart home infrastructure. 30. A computing device for a user's hydration assessment device, comprising a processor and a memory storing instructions for performing the method steps of claim 7. 31. A computer-readable medium for the user's hydration assessment device, storing instructions for causing the computing device to perform the steps of the method of claim 7.

Images