Intelligent multi-path collaborative weight loss regulation method and system based on multi-dimensional health data

Abstract

The application relates to the technical field of health data processing, in particular to an intelligent multi-path cooperative weight loss regulation method and system based on multi-dimensional health data; the method comprises the following steps: aligning weight, diet and exercise records into a weight link, an intake link and a consumption link, filtering and differentiating to obtain an actual change sequence; subtracting an individual basal metabolic rate from the difference value of the intake link and the consumption link and converting the difference value into a predicted change sequence, comparing the predicted change sequence with the actual change sequence to generate a balance residual sequence, decomposing the balance residual sequence into a sudden change component and a slow change component, and correcting the individual basal metabolic rate by using the slow change component increment in the metabolic drift interval; positioning a distortion link according to a record vacancy state and a sudden change component sign, filling in missing records and correcting deviations, updating an intake target and a consumption target according to the corrected energy balance sequence, and pushing a supplementary recording prompt. The application can realize record distortion repair and metabolic parameter adaptive tracking in the weight loss process, and guarantee that the weight loss target adjustment is based on accurate data.

Description

Technical Field

[0001] This application relates to the field of health data processing technology, specifically to an intelligent multi-path collaborative weight loss regulation method and system based on multi-dimensional health data. Background Technology

[0002] Existing smart weight loss apps typically integrate weighing data from smart scales, dietary records entered by users on the app, and exercise data from wrist-worn devices. They calculate daily energy deficits using an energy balance model and generate intake and exercise goals for the next cycle. However, in long-term real-world use, the weight changes obtained from weighing and the energy balance calculated from the records often show discrepancies in direction or magnitude. When these discrepancies occur, existing apps still treat all recorded values ​​as reliable and directly substitute them into the calculation of goals for the next cycle.

[0003] For example, a person trying to lose weight attended a dinner party on a Saturday. The side dishes from the hot pot and the sugary drinks were not recorded, resulting in a total food intake of only 1100 kcal that day. According to the record, there was a significant energy deficit. However, the next morning, the person's weight had increased by 0.4 kg instead of decreasing. The application still judged the exercise as successful based on the record and further reduced the intake target by 150 kcal in the next cycle. The person experienced increased hunger for several days afterward but saw no weight loss, leading them to suspect a recording error, but they couldn't pinpoint the specific day, meal, or whether the error was due to a missed food or exercise record. Another example: in the eighth week of weight loss, the weight remained around 81.5 kg for two consecutive weeks without further decrease. Both the food and exercise records were completed as planned. The deviation from the previous two weeks should have been included in the execution deviation. However, the energy deficit was increased twice, reducing the intake target from 1500 kcal to 1250 kcal, drastically increasing the difficulty of execution. The person eventually gave up recording the exercise.

[0004] If the above problems are not effectively addressed, intake and expenditure targets will continue to be based on distorted data. The direction and magnitude of target adjustments will deviate from actual needs. Users will either be unknowingly subjected to continuously reduced intake or be vaguely prompted to supplement data without verification. The recording behavior and target execution will become disconnected, making it difficult to maintain the weight loss plan in the long term. Therefore, it is necessary to address the problem of data distortion across multiple recording channels such as weight, diet, and exercise, as well as the slow changes in body energy status, which leads to distorted data on the basis for adjusting weight loss targets.

[0005] In view of this, this application proposes an intelligent multi-path collaborative weight loss regulation method and system based on multidimensional health data. Summary of the Invention

[0006] To address the problems mentioned in the background section, this application provides an intelligent multi-path collaborative weight loss regulation method and system based on multi-dimensional health data, the specific technical solution of which is as follows:

[0007] Firstly, this application provides an intelligent multi-pathway collaborative weight loss regulation method based on multidimensional health data, including:

[0008] The weight, diet, and exercise records of the weight loss subjects are aligned into weight, intake, and expenditure links according to a unified time window. The weight links are filtered by moving average to generate a weight trend sequence that filters out short-term physiological fluctuations. The actual change sequence is obtained by differentiating the weight trend sequence.

[0009] The energy balance sequence is obtained by subtracting the individual's basal metabolic rate from the difference between the intake and consumption pathways. This sequence is then converted into a predicted change sequence based on the individual's energy conversion coefficient. The actual change sequence is subtracted from the predicted change sequence to generate the balance residual sequence. The balance residual sequence is decomposed into a mutation component and a gradual change component. The individual's basal metabolic rate is corrected by the increment of the gradual change component within a preset metabolic drift range.

[0010] The time periods of mutation components are divided into missing segments and deviation segments according to the missing status of intake and consumption links. The distorted links are located according to the symbol of the mutation component. The missing segments are filled with the historical statistical values ​​of the distorted links in the same period, and the distorted link records are corrected according to the amplitude of the mutation component in the deviation segments. The intake and consumption targets for the next cycle are updated based on the corrected energy income and expenditure sequence, and a prompt to supplement the distorted links is pushed.

[0011] Preferably, the alignment of the weight link, intake link, and consumption link according to a unified time window includes: dividing the unified time window into natural days, and assigning the collected raw data to the corresponding time window according to the collection timestamp; taking the first weighing value within a preset morning period and writing it into the weight link, and using the first weighing value of the next natural day morning as the end-of-period weight record value of the current time window; weighing values ​​that deviate from the weight trend sequence by more than a preset judgment margin are judged as not being weighed by the individual and are not written into the weight link; time windows lacking morning weighing values ​​are marked as weight missing windows, and are filled by linear interpolation of the nearest valid weight record values ​​on both sides, with the missing mark retained in the interpolation window.

[0012] Preferably, the intake link and the consumption link include: the intake energy value of the intake link is obtained by accumulating all recorded energy values ​​that fall within the same time window at the time of eating, and the recorded energy value is obtained by multiplying the portion by the energy density retrieved from the food energy density database; the exercise consumption value of the consumption link is obtained by accumulating the activity consumption of all minutes within the time window, and the minute-level activity consumption is converted according to metabolic equivalent, the weight of the weight loss target and the duration, the metabolic equivalent is determined according to the intensity level divided by the percentage of heart rate reserve, and the metabolic equivalent involved in the conversion only retains the activity increment other than basal metabolism; the minute-level wearing status flag is stored with the consumption link.

[0013] Preferably, the step of decomposing the income and expenditure residual sequence into abrupt and gradual components includes: applying a median filter with a preset window length to the income and expenditure residual sequence; applying a moving average with a preset window length to the output of the median filter, using the cascaded output as the gradual component; subtracting the gradual component from the income and expenditure residual sequence point by point to obtain the abrupt component; when the number of accumulated time windows at the beginning of the sequence is insufficient to correspond to the preset window length, the two-stage filtering uses the actually covered time windows in the calculation.

[0014] Preferably, the step of correcting the individual's basal metabolic rate (BMR) with the incremental change of the slow-varying component within a preset metabolic drift interval includes: the initial value of the individual's BMR is estimated based on the gender, age, height, and weight of the weight loss subject; for each complete correction determination cycle, the average value of the slow-varying component within the current correction determination cycle is used as the incremental change of the slow-varying component; the incremental change of the slow-varying component is converted into an energy offset according to the individual's energy conversion coefficient; when the energy offset falls within the preset metabolic drift interval, the individual's BMR is corrected according to the energy offset; the corrected individual's BMR participates in the energy balance sequence calculation from the next time window, and the correction is performed iteratively cycle by cycle.

[0015] Preferably, the preset metabolic drift range includes: an energy range defined by a preset proportion of the currently effective individual basal metabolic rate; when the energy deviation exceeds the preset metabolic drift range, correction is performed using the boundary value of the preset metabolic drift range, and the excess portion is retained in the balance residual sequence.

[0016] Preferably, the step of converting the individual energy conversion coefficient into a predicted change sequence includes: setting the initial value of the individual energy conversion coefficient based on the energy density of adipose tissue; collecting body composition measurement data of the weight loss subjects according to a preset measurement cycle, and calculating the lean body mass loss ratio based on the ratio of lean body mass loss to body mass loss; lowering the individual energy conversion coefficient when the lean body mass loss ratio reaches a preset ratio threshold, and the lowered value takes effect from the subsequent time window.

[0017] Preferably, the mutation component time period includes: a time period composed of mutation windows, wherein the mutation window is a time window in which the absolute value of the energy converted from the mutation component exceeds the mutation determination threshold; the mutation determination threshold is set according to a preset multiple of the standard deviation of the energy converted from the residual sequence on the stable window set, and the stable window set is the time window that has not been determined as a mutation window in the most recent preset number of time windows; when the accumulated number of time windows is insufficient, the mutation determination threshold takes a preset initial value, and after the stable window samples are accumulated, it switches to the rolling calculation value; when the mutation components of adjacent mutation windows have the same sign, they are merged into one mutation component time period, and when the signs are opposite, they belong to two mutation component time periods.

[0018] Preferably, the missing status records for the intake and consumption links include: dividing the time window into meal periods; if there are no food record entries in any meal period, the corresponding meal period is marked as a missing intake record; snack entries outside of meal periods are included in the accumulation of intake energy values ​​but are not included in the missing status determination; the cumulative non-wearing time during the waking period is calculated according to the minute-level wearing status flag stored with the consumption link; if the cumulative non-wearing time exceeds the preset determination time, the current time window is marked as a missing consumption record.

[0019] Preferably, dividing the mutation component time period into omission segments and deviation segments includes: when the mutation component is negative, the ingestion link is considered a distorted link; when the mutation component is positive, the consumption link is considered a distorted link; the time window with a recording gap in the distorted link is classified as an omission segment; the time window with no recording gaps in either the ingestion or consumption link is classified as a deviation segment; when the sign of a link with a recording gap is inconsistent with that of the mutation component, the sign of the mutation component is used as the standard, and the segment is treated as a deviation segment; filling and correction are performed independently window by window within the omission segment and deviation segment.

[0020] Preferably, the method of filling the missed segment with historical statistical values ​​of the distorted link includes: when the distorted link is an intake link, taking time windows with the same day pattern as the time window in the missed segment within a preset number of weeks to form a time window set, with the day pattern divided into weekdays and rest days, and using the median of the total recorded energy values ​​corresponding to the recorded gap in the time window set as the filling value and writing it into the intake link; when the distorted link is a consumption link, taking time windows with the same week number as the time window in the missed segment within a preset number of weeks, and using the median of the exercise consumption value as the filling value and writing it into the consumption link; the filling value is stored with an estimation flag.

[0021] Preferably, the deviation segment correction of the distorted link record value according to the amplitude of the mutation component includes: converting the amplitude of the mutation component into a record correction amount according to the individual energy conversion coefficient, adding the record correction amount to the record value of the distorted link in the current time window; when the record correction amount exceeds a preset proportion of the record value of the distorted link in the current time window, only the preset proportion is used for correction, and the excess part is retained in the income and expenditure residual sequence and listed as a pending confirmation item in the supplementary recording prompt; when there is a weight missing measurement window filled by linear interpolation in the difference interval of the current time window, the record correction is temporarily suspended, and a verification prompt containing a record verification request and a weight supplementary measurement request is pushed.

[0022] Preferably, updating the intake and consumption targets for the next cycle includes: recalculating the corrected energy balance sequence using the supplemented and corrected intake and consumption pathways; converting the weekly weight loss target value into a target weekly deficit using the individual energy conversion coefficient, and allocating it as a target daily deficit; taking the opposite of the corrected energy balance value as the deficit value, and calculating the average deficit value within the target update cycle as the average deficit; adjusting the difference between the target daily deficit and the average deficit to the intake and consumption targets respectively according to a preset allocation ratio; the intake target uses the individual's basal metabolic rate as the lower limit, and when the lower limit is reached, the remaining difference is transferred to the consumption target.

[0023] Preferably, the push notification for supplementing the distorted link includes: the supplementary notification includes the distorted time window date, the distorted link identifier, the judgment type, the numerical adjustment amount already executed, and a confirmation entry; the confirmation entry is pre-filled with a list of high-frequency items corresponding to the distorted link and the numerical adjustment amount; in response to the weight reduction target's selection operation of the actual item, the actual value is backfilled, replacing the written numerical adjustment amount with the backfilled value, and the corrected energy income and expenditure sequence is recalculated; if no backfill is made, the numerical adjustment amount continues to be effective, and the backfill replacement only applies to the distorted time window, and the targets already issued are not traced back.

[0024] Secondly, this application also provides an intelligent multi-path collaborative weight loss regulation system based on multi-dimensional health data, which is used to implement the intelligent multi-path collaborative weight loss regulation method based on multi-dimensional health data, including: a link alignment module, a residual decomposition module and a record repair module.

[0025] The link alignment module is used to align the weight records, diet records, and exercise records of the weight loss subject into weight links, intake links, and consumption links according to a unified time window; to filter the weight links by moving average to generate a weight trend sequence that filters out short-term physiological fluctuations; and to obtain the actual change sequence by differentiating the weight trend sequence.

[0026] The residual decomposition module is used to subtract the individual's basal metabolic rate from the difference between the intake and consumption pathways to obtain an energy balance sequence, which is then converted into a predicted change sequence according to the individual's energy conversion coefficient; the predicted change sequence is subtracted from the actual change sequence to generate a balance residual sequence; the balance residual sequence is decomposed into a mutation component and a slow change component, and the individual's basal metabolic rate is corrected by the increment of the slow change component within a preset metabolic drift interval.

[0027] The record repair module is used to divide the period of mutation component into missing segment and deviation segment according to the missing status of intake link and consumption link records, and locate the distorted link according to the symbol of mutation component; the missing segment is filled with the historical statistical value of the distorted link in the same period, and the deviation segment is corrected according to the amplitude of mutation component to correct the distorted link record value; the intake target and consumption target of the next cycle are updated according to the corrected energy income and expenditure sequence, and a supplementary recording prompt is pushed to the distorted link.

[0028] The beneficial effects of this application are as follows: The technical solution of this application uses the filtered weight path as an objective benchmark to perform reverse diagnosis and window-by-window repair on the intake and consumption paths. Metabolic drift is absorbed by parameterization through a restricted correction path, and the intake and consumption targets are always generated based on the repaired energy balance sequence. Under real-world usage conditions where users often miss or under-record their energy levels, weight loss products can still maintain the direction and magnitude of target adjustments that match the actual energy state, reducing misjudgments during plateau periods, excessively low targets, and invalid data entry checks. This reduces manual verification costs, extends user retention periods, and is suitable for large-scale deployment scenarios where smart scales, wearable devices, and weight loss applications work together. Attached Figure Description

[0029] Figure 1 Flowchart of the intelligent multi-path collaborative weight loss regulation method based on multidimensional health data provided in this application;

[0030] Figure 2 This is a schematic diagram illustrating the weight loss data processing scenario of this application;

[0031] Figure 3 This is a schematic diagram of the unified time window alignment for the three links in this application;

[0032] Figure 4 This is a schematic diagram of the weight link moving average filter in this application;

[0033] Figure 5 This is a schematic diagram illustrating the separation of the two categories of revenue and expenditure residuals in this application;

[0034] Figure 6 This is a schematic diagram illustrating the metabolic drift range limitation correction in this application;

[0035] Figure 7 This is a schematic diagram illustrating the symbol localization of the mutation components in this application;

[0036] Figure 8 The structural diagram of the intelligent multi-path collaborative weight loss regulation system based on multidimensional health data provided in this application. Detailed Implementation

[0037] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the specific embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0038] Many specific details are set forth in the following description in order to provide a full understanding of this application. However, this application may also be implemented in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0039] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of this application. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.

[0040] Example 1

[0041] Reference Figures 1 to 7 This is the first embodiment of the present application, which provides an intelligent multi-path collaborative weight loss regulation method based on multi-dimensional health data.

[0042] The method in this embodiment operates within a weight loss data processing system comprised of a smart scale, a diet recording client, a wrist-worn exercise monitoring device, and a data processing server. The smart scale has a built-in strain gauge weighing sensor and wireless communication module, with a weighing resolution of 0.1 kg. After each weighing, the weighing value and time are packaged and uploaded to the data processing server. The diet recording client is installed on the mobile terminal held by the person trying to lose weight, allowing them to enter their diet records. Each record includes the food name, portion size, and time of consumption. The diet recording client also includes a built-in food energy density database. The wrist-worn exercise monitoring device has a built-in triaxial accelerometer and a photoplethysmography (PPG) heart rate sensor, outputting heart rate values ​​and wearing status indicators at minute-by-minute granularity, which are then transmitted to the data processing server via the mobile terminal. The data processing server performs all calculations, including link alignment, filtering, residual decomposition, record correction, and target update. The calculation results are fed back to the person trying to lose weight via the diet recording client.

[0043] Figure 2 This is a schematic diagram illustrating a weight loss data processing scenario in which this method is implemented. Figure 2 The weight loss participants wear wrist-worn motion monitoring devices that output minute-level heart rate values ​​and wearing status indicators, which are then transmitted and uploaded via a mobile terminal. The smart scale uploads the weighing value and weighing time to the data processing server. The diet recording client on the mobile terminal allows the weight loss participants to enter diet record items and receive feedback on intake targets, consumption targets, and supplementary recording prompts. The data processing server performs all calculations for link alignment, filtering, residual decomposition, record correction, and target updates. Figure 2 The dashed arrows indicate the data uplink and feedback downlink relationships between each terminal and the data processing server.

[0044] In weight management scenarios, weight data, dietary data, and exercise data come from the scale, manual input, and wearable devices, respectively, and the error characteristics of these three types of data are different. Weight values ​​are affected by water retention, gastrointestinal contents residue, and hormone cycles, resulting in short-term fluctuations unrelated to fat mass; dietary records rely on manual input, and omissions and underestimations of portion sizes are common; exercise expenditure relies on sensor conversion, which has limited ability to recognize certain types of exercise. When these three data links are used independently, distortion in any link will directly affect the adjustment basis of weight loss goals, causing intake and expenditure targets to deviate from actual needs. This embodiment incorporates the three links into a unified energy budget framework, using the filtered weight data as an objective benchmark to perform reverse diagnosis, filling, and correction of the recording quality of the intake and expenditure data links, specifically including steps S1 to S3.

[0045] Step S1: Align the weight records, diet records, and exercise records of the weight loss subject with a unified time window to form a weight link, an intake link, and a consumption link; apply a moving average filter to the weight link to generate a weight trend sequence that filters out short-term physiological fluctuations; and obtain the actual change sequence by differentiating the weight trend sequence.

[0046] Specifically, a unified time window is first defined using calendar days as the unit, with the window starting at 0:00 local time and lasting 24 hours. The time windows are numbered sequentially by date. Using calendar days as the window length is because dietary and exercise behaviors have a basic daily repetition cycle, and weight measurement typically occurs only once or twice a day. A window length shorter than one day would result in numerous gaps in the data, while a window length longer than one day would mask the differences in daily behavior. The data processing server categorizes the raw data uploaded from the scale, input terminal, and wearable device into the corresponding time window according to the collection timestamp, forming a three-link aligned data structure where weight records, energy intake, and exercise expenditure values ​​correspond one-to-one under the same window number.

[0047] Figure 3 The diagram illustrates the correspondence between the three data links aligned according to a unified time window. Figure 3 The top center displays three data sources: the smart scale uploads the weighing value and time; the diet recorder client inputs the food name, portion size, and time of consumption; and the wrist-worn fitness tracker outputs minute-level heart rate values ​​and wearing status indicators. The raw data from these three devices are categorized into unified time windows divided by calendar days, with each window lasting 24 hours. These time windows are numbered sequentially by date. Figure 3 The data also shows the one-to-one correspondence between weight records, energy intake, and exercise expenditure under the same window number. Thus, the weight, intake, and expenditure links have a window-by-window correspondence on the time axis, providing a data foundation for subsequent energy expenditure comparison.

[0048] All three data uploads carry a timestamp of the data collection time. When the wrist-worn motion monitoring device is offline, minute-level data is cached locally on the device. After connecting to the network, the data is re-transmitted according to the original timestamp. The data processing server then assigns the data to windows according to the timestamp. Out-of-order re-transmission does not affect the assignment result. The mobile terminal and the data processing server synchronize their time according to the network time protocol to avoid cross-window misassignment caused by local clock drift on the terminal.

[0049] The weight tracking system uses the first weigh-in between 4:00 AM and 10:00 AM as the recorded weight value. If multiple weigh-ins occur within this timeframe, only the first weigh-in is retained; weigh-ins outside this period are not recorded. The reason for limiting the timeframe to the morning is that the interference between gastrointestinal contents and water intake on the weight-indicator is minimal in the fasting state of the morning, ensuring consistent weigh-in conditions daily and comparability between time windows. Regarding time window allocation, morning weigh-ins reflect the body's state up to the end of all intake and expenditure activities the previous day. Therefore, the first weigh-in on the morning of the (t+1)th calendar day is used as the final weight record for the t-th time window. Food records with meals between 0:00 AM and 4:00 AM are synchronously assigned to the previous time window. This final caliber ensures that weight changes correspond temporally to energy intake and expenditure within the same time window. When a certain time window lacks a weighing record that meets the value selection rules, the current time window is marked as a weight loss window. The two most recent valid weight records on both sides of the loss window are used for linear interpolation to fill the gap. The interpolation is only used to maintain the sequence continuity of the weight link, and the interpolation window retains the loss mark.

[0050] Considering the reality that family members often share scales, even after a smart scale is linked to the account of the person trying to lose weight, it may still receive weight values ​​that are not theirs. The data processing server will determine that weight values ​​that deviate from the most recent value in the weight trend sequence by more than a certain margin are not theirs and will not be written into the weight tracking system. The basis for setting the certain margin is that the sum of the actual weight change and water retention fluctuations in a single window generally does not exceed 2 kg. The certain margin needs to be higher than the sum of these two values ​​to allow for a balance between excluding other people's weights and retaining the actual weight change. In this embodiment, the certain margin is set to 5 kg.

[0051] When constructing an intake link, the diet tracking client searches the food energy density database by food name and lists candidate items. After the weight loss target selects and confirms, the energy value of a single record is obtained by multiplying the serving size by the energy density of the selected item, with the unit uniformly in kilocalories. The serving size field supports two methods: entering by grams and entering by commonly used units of measurement. Commonly used units of measurement are pre-converted to grams by the food energy density database. The data processing server accumulates the energy values ​​of all records where the eating time falls within the same time window to obtain the intake energy value of the intake link in the current time window. The eating time of each record is stored along with the intake link.

[0052] When constructing the consumption pathway, the data processing server reads the heart rate value output by the wrist-worn motion monitoring device every minute. The heart rate reserve percentage is determined by the ratio of the difference between the current heart rate and the resting heart rate to the heart rate reserve. Heart rate reserve is the difference between the maximum heart rate and the resting heart rate. The resting heart rate is the average of the lowest heart rates during the last 7 nights of sleep. The maximum heart rate is estimated by subtracting age from 220. Both estimation methods are well-known in the field of exercise physiology. Minutes with a heart rate reserve percentage below 20% are classified as daily activity level; 20% to 40% is low-intensity level; 40% to 60% is moderate-intensity level; and above 60% is high-intensity level. These four levels correspond to metabolic equivalents of 2, 3, 5, and 8, respectively. The level boundaries are set according to the intensity classifications commonly used in exercise physiology. Minute-level activity expenditure is calculated using metabolic equivalents, the known conversion relationship between the target's weight and duration, and the metabolic equivalent used in the conversion is reduced by 1 from its original value, retaining only the activity increment beyond basal metabolism. The activity expenditure of all minutes within the time window is summed to obtain the exercise expenditure value of the expenditure pathway in the current time window. The reason for subtracting 1 is that the metabolic equivalent itself includes the resting metabolic share, while basal metabolism is deducted separately in subsequent energy expenditure calculations; including both simultaneously would result in duplicate deduction of the same energy component. The minute-level wear status indicator output by the wrist-worn motion monitoring device is stored along with the expenditure pathway.

[0053] Directly differencing the weight pathway can misinterpret water-related fluctuations as weight changes related to fat mass. Therefore, a moving average filter is applied to the weight pathway before differencing. The main sources of short-term physiological fluctuations include water and sodium retention caused by changes in sodium intake, changes in bound water associated with changes in glycogen reserves, changes in gastrointestinal contents residue, and hormonal water retention during the female menstrual cycle. These fluctuations typically range from 0.5 kg to 2 kg and last for 3 to 7 days, large enough to obscure the true fat changes within a single window. The filter window length should cover the typical duration of water-related fluctuations, as well as the complete behavioral cycle consisting of weekdays and rest days. In this embodiment, the filter window length is... Seven time windows are used. The weight trend series is calculated as follows: ;in, Indicates the first Weight trend values ​​over a time window Indicates the first Weight records for each time window, Indicates the length of the filter window. Indicates the time window number. Indicates the offset index within the filtering window. Insufficient cumulative weight at the beginning of the link. When there is a time window, the average of all accumulated weight records is used as the output to avoid invalid values ​​at the beginning.

[0054] Figure 4This demonstrates the input and output of the weight-linked moving average filter. Figure 4 The dotted line represents the weight records for each time window, superimposed with short-term physiological fluctuations caused by water and sodium retention, glycogen-bound water, and hormonal water retention. The direction of these fluctuations is unrelated to changes in body fat. The thick solid line represents the filtered weight trend sequence, with a filter window of 7 time windows, covering the typical duration of water-related fluctuations and the complete behavioral cycle consisting of weekdays and rest days. Figure 4 The coverage area of ​​the filter window is indicated by a double-headed arrow. After filtering, short-term physiological fluctuations are removed, and the weight trend sequence reflects the net change related to the increase or decrease of fat mass, providing a stable objective benchmark for obtaining the actual change sequence through differencing.

[0055] The actual change sequence is obtained by performing a first-order difference on the weight trend sequence. The difference is calculated as follows: ;in, Indicates the first The actual change in weight over each time window is expressed in kilograms. Positive values ​​represent an upward trend in weight, while negative values ​​represent a downward trend. The actual change sequence is valid from the second time window onwards.

[0056] After executing step S1, the three links have a window-by-window correspondence on the time axis. The final weight caliber eliminates the temporal misalignment between the weighing time and the behavior period. Short-term fluctuations in the weight link that are unrelated to energy balance are filtered out. The actual change sequence reflects the net change related to the increase or decrease of fat mass. As a result, the subsequent energy balance comparison obtains a stable objective benchmark.

[0057] Step S2: Subtract the individual's basal metabolic rate from the difference between the intake and consumption pathways to obtain the energy balance sequence, and convert it into a predicted change sequence according to the individual's energy conversion coefficient; subtract the actual change sequence from the predicted change sequence to generate the balance residual sequence; decompose the balance residual sequence into mutation components and slow-change components, and correct the individual's basal metabolic rate with the increment of the slow-change component within the preset metabolic drift interval.

[0058] The initial value of an individual's basal metabolic rate (BMR) was calculated by substituting the subject's gender, age, height, and weight into a general BMR estimation formula. This estimation formula is well-known in the field of exercise physiology and will not be elaborated upon here. In this embodiment, for a male subject who is 175 cm tall, weighs 85 kg, and is 35 years old, the initial BMR is set to 1780 kcal per time window. There is usually a deviation between the initial value given by the estimation formula and the individual's actual metabolic level. In this embodiment, the initial value is only used as the starting point for iteration; the accuracy of the initial value does not affect the operation of subsequent mechanisms, and the deviation will be gradually absorbed through the correction path of the slowly varying components.

[0059] If an individual's basal metabolic rate (BMR) is kept at a fixed initial value for an extended period, a recurring problem in weight loss practice arises: after several weeks of energy restriction, the body's actual BMR gradually declines, the predicted weight loss under fixed parameters consistently exceeds the actual weight loss, the energy deficit is continuously increased, intake targets are gradually lowered, the difficulty of weight loss becomes more pronounced, and the actual metabolism is further adjusted downwards, creating a negative cycle. This embodiment establishes a restricted closed-loop correction path for the individual's BMR in step S2 to break the aforementioned cycle.

[0060] The energy budget sequence is calculated as follows: ;in, Indicates the first Energy balance for each time window Indicates the intake link at the 1st Energy intake values ​​within a time window, Indicates the consumption link in the 1st... The exercise consumption value for each time window This represents the current basal metabolic rate of an individual, with all units in kilocalories. A positive energy balance indicates an energy surplus within the current time window, while a negative value indicates an energy deficit.

[0061] The physical meaning of the individual energy conversion coefficient (ECC) is the energy equivalent corresponding to a unit change in body weight for a weight-loss subject. It is based on the following: the energy density of adipose tissue is approximately 7700 kcal / kg, a well-known value; the components lost during weight loss are not entirely fat, and the higher the percentage of lean body mass loss, the lower the energy equivalent corresponding to a unit change in body weight. In this embodiment, the initial ECC is set at 7700 kcal / kg. Bioelectrical impedance analysis (BIA) data of the weight-loss subject is collected monthly. The percentage of lean body mass loss is calculated as the ratio of lean body mass loss to total body weight loss. When the ratio reaches 20%, the ECC is lowered to 6600 kcal / kg, and the lowered value takes effect from subsequent time windows. The conversion of the predicted change sequence is as follows: ;in, Indicates the first The predicted change over each time window, in kilograms. This represents the individual's energy conversion coefficient. The predicted change in weight is the expected weight change within the current time window, assuming accurate intake, expenditure, and basal metabolic rates.

[0062] The income and expenditure residual series is generated by subtracting the actual change series from the predicted change series, that is: ;in, Indicates the first The residual value of income and expenditure within a time window. When all three links are recorded accurately and the individual's basal metabolic rate is consistent with the actual metabolic level, the residual value fluctuates randomly around zero. In actual operation, the residual value carries two types of error information with different characteristics. The first type is sudden recording errors, such as the subject missing a meal, visually estimated portion size being lower than the actual portion size, cooking oil and sugary drinks not being recorded, the entire scene of a group meal being missing, and exercise occurring while the wrist-worn motion monitoring device is being removed for charging not being measured. All of these situations manifest as isolated high-amplitude residuals within individual time windows. The second type is a systematic and slow shift. Adaptive thermogenesis causes the actual basal metabolic rate to gradually decrease under energy restriction. Weight loss itself also reduces basal metabolic rate and the actual energy consumption of equivalent exercise. Long-term underestimation of portion size on the intake side can also create a stable shift in direction. All of these situations manifest as the mean of the residual value slowly drifting on a weekly to monthly time scale. The two types of errors are clearly different on the time scale. The former is an isolated spike, while the latter is a low-frequency drift. They meet the conditions for separation based on frequency characteristics. Separating and processing the two types of errors is the core of step S2.

[0063] The decomposition employs a low-frequency extraction method that combines median filtering and cascaded moving average. The first stage applies a window of length [missing information] to the income and expenditure residual sequence. Median filtering within a time window, within the window After sorting the residual values, the middle value is taken as the output. Median filtering is insensitive to isolated spikes and can exclude spikes from low-frequency estimation without widening them. Spikes caused by a single missed record or a single gathering last for 1 to 2 time windows. Median filtering with a window length of 5 can completely remove spikes with a width not exceeding 2 time windows. Odd window lengths ensure that the output is aligned with the window center. In this embodiment... Take 5. The second stage applies a window length of 5 to the median filter output. The moving average is applied over several time windows to smooth out residual intra-weekly fluctuations. The timescale for systematic shifts ranges from weekly to monthly. The moving average window length is two complete behavioral cycles, which smooths out intra-weekly residuals without erasing monthly drift. In this embodiment... Take 14. The cascaded output is the slowly varying component. Subtracting the slowly varying component point by point from the income and expenditure residual sequence yields the abrupt component, i.e.: ;in, Indicates the first The abrupt component values ​​of each time window Indicates the first The slowly varying component values ​​of each time window. When the accumulated value at the beginning of the sequence is insufficient for the corresponding window length, both stages of filtering use the actual covered time windows in the calculation.

[0064] Figure 5 This demonstrates the mechanism by which the income and expenditure residual series separates into two types of components according to the time scale. Figure 5In the upper sequence, isolated spikes correspond to sudden recording errors such as missed meals, underestimated portions, and missed measurements during periods when wrist-worn motion monitoring devices are removed. Low-frequency drift corresponds to a systematic, slow shift caused by adaptive thermogenesis and long-term underestimation of portions. After low-frequency extraction using median filtering and cascaded moving averages, the slowly varying components show a systematic shift, while the abrupt components retain the timing, direction, and amplitude of sudden recording errors at zero baseline. The two types of error information do not interfere with each other; a single meal will not be misinterpreted as a metabolic change, and a decline in metabolism will not be misinterpreted as an anomaly in a single day's recording.

[0065] After decomposition, sudden recording errors are concentrated in abrupt components, while systematic shifts are concentrated in slowly varying components. The two types of error information do not interfere with each other, and the subsequent processing paths each obtain clean input. A single meal will not be misread as a metabolic change, and a metabolic decline will not be misread as an anomaly in the recording of a certain day.

[0066] Slow-varying components do not entirely originate from actual metabolic changes. Recording bias caused by long-term underestimation of these components can also lead to a continuous negative accumulation of slow-varying components. If all slow-varying components are indiscriminately recorded into an individual's basal metabolic rate (BMR), systematic errors at the recording end will be mistakenly absorbed into the physiological parameters. The preset metabolic drift range is designed to distinguish between these two factors. The range is based on experimental findings in exercise physiology, which show that the decrease in BMR caused by adaptive thermogenesis under energy restriction generally does not exceed 15% of the BMR. Therefore, the metabolic drift range is set as an energy range defined by plus or minus 15% of the current effective individual's BMR. Slow-varying component increments falling within this range are treated as actual metabolic changes, while those exceeding the range are treated as systematic biases at the recording end and are not recorded into the individual's BMR. For example, when an individual's BMR is 1780 kcal per time window, the metabolic drift range is from -267 kcal to +267 kcal.

[0067] Figure 6 This demonstrates how metabolic drift ranges constrain the correction of an individual's basal metabolic rate. Figure 6 The thick solid line represents the current effective individual basal metabolic rate (BMR). The metabolic drift range above and below this line is defined by ±15%, indicated by a gray band and dashed upper and lower boundaries. The left arrow indicates cases where the energy shift calculated from the incremental change in the slowly varying component falls within the range; this is treated as a true metabolic change and fully recorded in the individual's BMR. The right arrow crosses the lower boundary of the range; the out-of-bounds segment is drawn as a dashed line, indicating that correction is performed based on the range boundary value. Any deviation exceeding this boundary is treated as a systematic bias at the recording end and retained in the balance residual sequence. Corrections are performed periodically according to the correction determination cycle. The magnitude of a single correction is constrained by the range, ensuring that recorded shifts are not mistakenly absorbed into physiological parameters.

[0068] The correction of individual basal metabolic rate is performed according to the correction judgment period. The correction judgment period should cover the complete monthly behavioral cycle and the female menstrual cycle, and the sample size within the period should be sufficient to smooth single-window noise. In this embodiment, the correction judgment period is 28 time windows. After each correction judgment period, the average value of the slow variation component within the current correction judgment period is calculated and recorded as the slow variation component increment. The unit is kilograms. After each correction takes effect, the balance sheet residual sequence is re-centered at zero. The average value of the slowly varying components in the next correction determination period reflects the newly accumulated drift after the correction takes effect; drift already absorbed by the correction will not be included again. Energy shift converted from the increment of slowly varying components. When falling within the metabolic drift range, the corrected basal metabolic rate is expressed as follows: ;in, This represents the corrected individual basal metabolic rate; when the energy shift exceeds the metabolic drift range, the correction is performed using the range boundary value, and the excess portion is retained in the income and expenditure residual sequence.

[0069] After the amendment takes effect, renew From the next time window onwards, the energy balance sequence will be updated according to the new timeline. The calculation and correction are performed iteratively on a cycle basis, and the individual's basal metabolic rate continuously tracks the actual metabolic level as weight loss progresses. The magnitude of a single correction is naturally constrained by the metabolic drift range, and parameter updates will not oscillate. The monthly updates of the individual's energy conversion coefficient and the periodic corrections of the individual's basal metabolic rate are independent of each other and do not affect the timing of each other's effectiveness.

[0070] Step S2 separates the two types of error information in the income-expenditure residuals according to the time scale. The gradually changing component is written back to the individual's basal metabolic rate in a restricted manner, and the prediction model can adjust to the decline in the actual metabolic level during the weight loss process. When a weight loss plateau occurs, the income-expenditure residuals show a gradually changing drift and are absorbed as a downregulation of metabolic parameters. The predicted change sequence then converges to the actual change sequence, and the calculation basis of the intake and expenditure targets remains valid. The plateau will not be misjudged as a recording error. The abrupt component fully retains the information on the timing, direction, and magnitude of sudden recording errors.

[0071] Step S3: Divide the time period of mutation component into missing segment and deviation segment according to the missing status of intake and consumption link records, and locate the distorted link according to the symbol of mutation component; fill the missing segment with the historical statistical value of the distorted link in the same period, and correct the distorted link record value according to the amplitude of mutation component in the deviation segment; update the intake target and consumption target of the next cycle according to the corrected energy income and expenditure sequence, and push the supplementary recording prompt to the distorted link.

[0072] The judgment and correction in step S3 are both performed on the energy caliber. The mutation component and the income and expenditure residual are first multiplied by the individual energy conversion coefficient and converted into kilocalories before being compared. The unified caliber facilitates the expression of thresholds and the display of prompts.

[0073] The determination of a mutation component time period is based on a mutation threshold. A time window where the absolute value of the energy converted from the mutation component exceeds this threshold is considered a mutation window. The mutation threshold is set at 2.5 times the standard deviation of the energy converted from the residual sequence on a stable window set. The stable window set refers to the time windows that have not been identified as mutation windows in the most recent 28 time windows. Under stable conditions, the residuals have approximately zero mean and random fluctuations, with the percentage of samples deviating from 2.5 times the standard deviation on the order of 2%, resulting in an acceptable false positive rate. Simultaneously, the energy converted from the threshold should be on the order of a single meal; deviations below this level are considered normal recording noise and have no correction value. For example, when the energy converted from the standard deviation on the stable window set is 200 kcal, the mutation threshold is set to 500 kcal. During the cold start phase, if less than 28 time windows have been accumulated, the mutation threshold is directly set to 500 kcal; once 28 time windows have been accumulated, the threshold is switched to a rolling calculation value. When adjacent mutation windows have the same mutation component sign, they are merged into one mutation component time period. When the signs are opposite, they belong to two time periods. Within each time period, filling and correction are performed independently for each window. The time period unit is only used for aggregated display of prompts.

[0074] The recording gap status of the intake tracking is determined based on meal times. Each time window is divided into three meal times: 6:00-9:00, 11:00-14:00, and 17:00-20:00, based on the distribution range of a typical three-meal-a-day schedule. If there are no food records within any meal time period, the corresponding meal time period is marked as a gap in intake recording; if there are no records within the entire time window, the entire window is marked as a gap in intake recording. Snack entries outside of meal times are included in the normal accumulation of energy intake and are not included in the gap determination. Gap determination depends only on the presence or absence of an entry, not on the reliability of the entry's value. The determination result is stored with the intake tracking as a binary label, resulting in low computational complexity and stable results.

[0075] The record gap status of the consumption link is determined based on the wearing status. The data processing server reads the minute-level wearing status flag stored with the consumption link in step S1, and counts the cumulative non-wearing time during the waking period from 7:00 to 23:00. When the cumulative non-wearing time exceeds the judgment time, the current time window is marked as a consumption record gap. The judgment time should be sufficient to accommodate one complete exercise behavior. The longer the non-wearing window, the higher the probability of missing the entire exercise segment. If the judgment time is too short, fragmented removal caused by washing and charging will be mistakenly judged as a gap. In this embodiment, the judgment time is 4 hours; if the wrist-worn exercise monitoring device does not support underwater wear and the weight loss object has a regular swimming schedule, the judgment time is reduced to 2 hours.

[0076] The division between omission and deviation segments is based on the location of the distortion link, which is determined by the difference in direction between the dominant errors at the intake and consumption ends. At the intake end, manual recording errors are mainly due to underestimation; visually estimated portions are generally lower than actual portions, and the energy from cooking oil and seasonings is often missed. Underestimating intake values ​​leads to a lower predicted change than the actual change, resulting in a negative residual. At the consumption end, errors during normal device wear are concentrated on insufficient activity type recognition. Activities like cycling and strength training, which involve few wrist movements but high actual energy consumption, are underestimated. Underestimating consumption values ​​leads to a higher predicted change than the actual change, resulting in a positive residual. The gradually varying components have already been separated in the low-frequency extraction in step S2. The direction of sudden recording errors is directly reflected in the sign of the abrupt change component: the abrupt change component corresponding to the dominant error at the intake end is negative, and the abrupt change component corresponding to the dominant error at the consumption end is positive. The dominant error directions of the two links are opposite, thus the sign of the abrupt change component has a definite link orientation capability.

[0077] Figure 7 It demonstrates the basis for locating distorted links using mutation component symbols; Figure 7 The negative peak below the zero baseline corresponds to the dominant error on the intake side. Underestimation of portion size, omission of cooking oil and seasonings, missed meal recording, and complete absence of mealtime scenarios all lead to lower intake records, resulting in a negative residual. The distortion link is identified as the intake link. The positive peak above the zero baseline corresponds to the dominant error on the consumption side. Insufficient recognition of activities with few wrist movements but high actual energy consumption, such as cycling and strength training, as well as missed measurements during device charging, lead to lower consumption records, resulting in a positive residual. The distortion link is identified as the consumption link. The dominant errors of the two links are in opposite directions, and the sign of the abrupt change component has a definite link orientation capability.

[0078] The specific classification rules are as follows: Within the mutation component period, the time window where the mutation component is negative and there is a record gap in the ingestion link is classified as an ingestion link omission segment; the time window where the mutation component is positive and there is a record gap in the consumption link is classified as a consumption link omission segment; the time window where neither link has a record gap and the mutation component is negative is classified as an ingestion link deviation segment, and the distorted link is the ingestion link; the time window where neither link has a record gap and the mutation component is positive is classified as a consumption link deviation segment, and the distorted link is the consumption link; when the sign of a link with a record gap is inconsistent with that of the mutation component, the sign of the mutation component shall prevail, and the deviation segment rules shall apply. The classification rules provide a definite conclusion for all combinations of signs and gaps, and there are no unclassifiable time windows or branches that rely on manual judgment within the mutation component period.

[0079] Before performing the filling and correction, a data reliability check is performed. The difference interval refers to the range of time windows involved in calculating the actual change in the current time window, consisting of the current time window and the L time windows preceding it. If there is a missing weight measurement window filled by linear interpolation within the difference interval of the abrupt change window, the record correction for the current abrupt change window is temporarily suspended, and only a verification prompt is pushed. The verification prompt includes the abrupt change window date, the record verification request, and the weight re-measurement request, and is sent through the same message channel as the supplementary recording prompt. The reason for setting up the check is that interpolated weight does not carry true measurement information, and using interpolation difference to drive record correction will amplify the interpolation error and write it into the record link.

[0080] The method for filling missing segments in the intake link is as follows: The time window within the missing segment is designated as the missing window. A set of time windows with the same day type as the missing window within the most recent 4 weeks is selected, with the day type divided into weekdays and rest days. The total recorded energy values ​​corresponding to the missing meal times are extracted from the time window set, and the median of the total values ​​is used as the filler value and written into the intake link of the missing window. Statistically, the 4-week window strikes a balance between behavioral timeliness and sample size. A window with the same day type can provide 8 to 20 samples within 4 weeks. The median is used because meal peaks can raise the mean, and the median is not sensitive to outliers. For example, if a weight loss subject has no entries for lunch on a certain weekday and the mutation amount is negative, and the median of the total recorded energy values ​​for lunch times on weekdays within the most recent 4 weeks is 650 kcal, then 650 kcal will be written. The filler value is stored with an estimated flag to distinguish it from the actual recorded value.

[0081] The same method applies to filling in missing segments in the energy expenditure tracking. A time window with the same week number as the missing window is selected from the most recent four weeks. The median of the energy expenditure value is used as the filler value and written into the energy expenditure tracking of the missing window, with an estimation flag. The reason for using the same week number is that regular exercise schedules are usually tied to a particular day of the week, and the consistency of energy expenditure within the same week number window is highest. For example, if a person trying to lose weight habitually trains every Wednesday evening, and a certain Wednesday window has a missing energy expenditure record, it is filled with the median of the energy expenditure values ​​of the most recent four Wednesday windows, which is 320 kcal.

[0082] The correction for the deviation segment is performed based on the amplitude of the abrupt change component. The amount of distortion link correction is expressed as follows: ;in, Indicates the first The correction amount for the distortion link within a time window is expressed in kilocalories, with the vertical bar indicating absolute value calculation. When the distortion link is an intake link, the correction amount is added to the intake energy value of the current time window; when the distortion link is a consumption link, the correction amount is added to the exercise consumption value of the current time window. Both paths are corrected upwards, consistent with the positioning basis that the dominant errors at both the intake and consumption ends are low recorded values. To prevent abnormalities in the weight link itself from being mixed into the record correction, a single-window correction upper limit is set. When the correction amount exceeds 50% of the current time window recorded value of the distortion link, only 50% is executed, and the excess is retained in the income and expenditure residual sequence and listed as a pending confirmation item in the supplementary recording prompt. A difference of more than half of the recorded value increases the probability of mixing in weight link anomalies such as weighing with clothing or changing scales; writing the full amount will pollute the record link. For example, if the intake recorded value in a certain deviation time window is 1400 kcal, the energy equivalent of the mutation component is 900 kcal, and the correction amount executed according to the upper limit is 700 kcal.

[0083] After the filling and correction are completed, the data processing server recalculates the energy balance sequence using the corrected intake and consumption pathways and the currently effective individual basal metabolic rate, resulting in the corrected energy balance sequence. The target update for the next cycle is based solely on the corrected energy balance sequence. The target update cycle consists of 7 time windows, maintaining the same behavioral cycle caliber as the filter window length. During the update, the system first reads the weekly weight loss target set by the individual in the diet tracking client. The weekly weight loss target is limited to 0.25 kg to 1 kg per week on the client side, based on the generally accepted range of healthy weight loss rates. The weekly weight loss target is multiplied by the individual's energy conversion factor to obtain the target weekly deficit, and divided by 7 to obtain the target daily deficit. The negative of the corrected energy balance is recorded as the deficit value. The average deficit value is calculated by averaging the deficit values ​​in the most recent target update cycle. The difference between the target daily deficit and the average deficit is calculated. If the difference is positive, the intake target for the next cycle is reduced by 60% of the difference based on the current intake target, and the consumption target for the next cycle is increased by 40% of the difference based on the current consumption target. If the difference is negative, the intake target is increased by 60% of the absolute value of the difference, and the consumption target is decreased by 40% of the absolute value of the difference. Energy regulation at the intake end has greater scope and adherence than at the expenditure end. Boosting at the expenditure end is constrained by exercise time and joint load; a 60% to 40% allocation strikes a balance between regulation speed and feasibility. The intake target has a lower limit, taken as the currently effective individual basal metabolic rate. When the intake target reaches the lower limit, the remaining difference is transferred to the expenditure target. Long-term intake below the basal metabolic rate will exacerbate adaptive thermogenesis decline, contradicting the sustainability of weight loss.

[0084] When the goals are set, the intake goals are broken down and displayed in a 3:4:3 ratio across three main meal periods, based on the energy distribution of three regular meals. The expenditure goals are converted into exercise duration at a moderate intensity level for easy execution by the target individual. For example, if a target weighing 85 kg increases their expenditure goal by 120 kcal, the increase is converted into approximately 20 minutes of moderate intensity exercise, based on the increase in activity after subtracting 1 from the metabolic equivalent of 5 at the moderate intensity level.

[0085] The supplementary recording prompt will be pushed out in the next time window after the time period division is completed, with the push time set to 8:00 AM, via the message channel of the diet record client. The prompt content includes the date of the distortion time window, the distortion link identifier, the judgment type, the amount of numerical adjustment already performed, and the confirmation entry. When the judgment type is "missed recording," the numerical adjustment amount is the fill-in value; when the judgment type is "deviation," the numerical adjustment amount is the correction amount. When the distortion link is a consumption link and the judgment type is "missed recording," the prompt content will include a wearing reminder. The system confirms the pre-filled numerical adjustment amount in the entry and the list of high-frequency items corresponding to the distorted link. When the distorted link is an intake link, it lists the high-frequency dietary items of the same period; when the distorted link is a consumption link, it lists the high-frequency exercise items of the same period. The high-frequency items of the same period are selected in descending order of the frequency of occurrence of the items in the same daily time window (intake link) or the same week ordinal time window (consumption link) within the last 4 weeks. After the weight loss target selects the actual item, the actual value is filled in. The data processing server replaces the adjustment value written by the server with the filled value. The replacement of the filled value simultaneously clears the estimation flag and recalculates the corrected energy income and expenditure sequence. If the weight loss target does not fill in the value, the adjustment value continues to be effective, and the target update is not interrupted due to waiting for the fill in. The fill replacement only applies to the distorted time window. The issued target is not retrospectively applied, and the target for the next cycle is calculated based on the replaced data.

[0086] The targeted design of the prompts allows weight loss participants to check only one record for a specific date, specific link, and specific meal time, eliminating the need to manually check the entire record set. The execution cost of supplementary recording is compressed to the granularity of a single record. The supplemented real data, in turn, replaces the estimated value, and the record quality continuously improves with usage time. Step S3 thus completes the entire process from anomaly detection, distortion localization, record repair to target closure. Omissions and deviations are distinguished according to the state of record gaps, and the handling methods are matched to the causes of distortion. The symbols of the mutation components provide a definite link direction, avoiding the misapplication of intake-side problems to the consumption-side. Filling and correction ensure that the energy balance sequence remains usable during the distortion period, and target updates are always based on the repaired data.

[0087] The method provided in this embodiment uses the filtered weight tracking as an objective benchmark, transforming the recording quality issues of the intake and expenditure tracking into detectable signals on the energy balance residual sequence. It separates metabolic drift and recording distortion by decomposing the data on a time scale. Metabolic drift is absorbed through parameter correction pathways, and recording distortion is repaired window-by-window through filling and correction pathways. Intake and expenditure targets are always generated based on the repaired energy balance sequence. The three tracking paths mutually verify each other, and the inherent reliability defects of manual recording are compensated by the data processing stage. The weight loss subjects do not need nutritional calculation capabilities, and the method remains adaptive with data accumulation over long-term use, solving the problem of data-level distortion in adjusting weight loss plans.

[0088] Example 2

[0089] Reference Figure 8 This is the second embodiment of the present application, which provides an intelligent multi-path collaborative weight loss regulation system based on multi-dimensional health data.

[0090] The system includes: a link alignment module, a residual decomposition module, and a record repair module.

[0091] The link alignment module is used to align the weight records, diet records, and exercise records of the weight loss subject into weight links, intake links, and consumption links according to a unified time window; to filter the weight links by moving average to generate a weight trend sequence that filters out short-term physiological fluctuations; and to obtain the actual change sequence by differentiating the weight trend sequence.

[0092] The residual decomposition module is used to subtract the individual's basal metabolic rate from the difference between the intake and consumption pathways to obtain an energy balance sequence, which is then converted into a predicted change sequence according to the individual's energy conversion coefficient; the predicted change sequence is subtracted from the actual change sequence to generate a balance residual sequence; the balance residual sequence is decomposed into a mutation component and a slow change component, and the individual's basal metabolic rate is corrected by the increment of the slow change component within a preset metabolic drift interval.

[0093] The record repair module is used to divide the period of mutation component into missing segment and deviation segment according to the missing status of intake link and consumption link records, and locate the distorted link according to the symbol of mutation component; the missing segment is filled with the historical statistical value of the distorted link in the same period, and the deviation segment is corrected according to the amplitude of mutation component to correct the distorted link record value; the intake target and consumption target of the next cycle are updated according to the corrected energy income and expenditure sequence, and a supplementary recording prompt is pushed to the distorted link.

[0094] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings or direct couplings or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0095] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments under the guidance of this application without departing from the spirit and scope of protection of the claims. All of these variations are within the protection scope of this application.

[0096] It should be noted that the collection, storage, and use of all personal information involved in the technical solution of this application must be carried out only after obtaining the explicit authorization and separate consent of the information subject. The processing of personal information-related data strictly complies with the requirements of currently effective national laws, regulations, and relevant standards and specifications. The collected personal information is limited to use within the specific purpose necessary to achieve the technical solution of this application, and no processing beyond that purpose will be performed. Necessary technical and management measures will be taken to ensure the security of personal information.

Claims

1. A smart multi-path collaborative weight loss regulation method based on multidimensional health data, characterized in that, include: The weight records, diet records, and exercise records of the weight loss participants are aligned into a weight tracking, intake tracking, and consumption tracking according to a unified time window. The weight trend sequence is generated by filtering the weight link moving average to remove short-term physiological fluctuations. The actual change sequence is obtained by differentiating the weight trend sequence. The energy balance sequence is obtained by subtracting the individual's basal metabolic rate from the difference between the intake and consumption pathways. This sequence is then converted into a predicted change sequence based on the individual's energy conversion coefficient. The actual change sequence is subtracted from the predicted change sequence to generate the balance residual sequence. The balance residual sequence is decomposed into a mutation component and a gradual change component. The individual's basal metabolic rate is corrected by the increment of the gradual change component within a preset metabolic drift range. Based on the missing status of intake and consumption links, the time period of mutation component is divided into missing segment and deviation segment, and the distorted link is located according to the sign of the mutation component; the missing segment is filled with the historical statistical value of the distorted link in the same period, and the distorted link record value is corrected according to the amplitude of the mutation component in the deviation segment. The intake and consumption targets for the next cycle are updated based on the revised energy balance sequence, and supplementary recording prompts are pushed to the distorted link.

2. The intelligent multi-path collaborative weight loss regulation method based on multidimensional health data according to claim 1, characterized in that, The alignment of the weight, intake, and consumption pathways according to a unified time window includes: Divide the time window into a unified time window based on the calendar day, and classify the collected raw data into the corresponding time window according to the collection timestamp; The first weighing value within the preset morning time period is written into the weight link, and the first weighing value in the morning of the next natural day is used as the final weight record value of the current time window. Weight values ​​that deviate from the most recent value in the weight trend sequence by more than the preset judgment margin are judged as not being weighed by the individual and are not recorded in the weight tracking. The time window with missing morning weigh-in values ​​is marked as the weight loss window, and is filled by linear interpolation of the two nearest valid weight records. The interpolation window retains the loss mark.

3. The intelligent multi-path collaborative weight loss regulation method based on multidimensional health data according to claim 1, characterized in that, The intake and consumption pathways include: The energy intake value of the intake link is obtained by summing all recorded energy values ​​that fall within the same time window at the time of eating. The recorded energy value is obtained by multiplying the portion size by the energy density retrieved from the food energy density database. The exercise consumption value of the consumption link is obtained by summing up the activity consumption of all minutes within the time window. The minute-level activity consumption is converted according to metabolic equivalent, the weight of the weight loss target and the duration. The metabolic equivalent is determined by the intensity level divided by the heart rate reserve percentage. The metabolic equivalent used for conversion only retains the activity increment other than the basal metabolism. Minute-level wear status flags are stored along with the consumed link.

4. The intelligent multi-path collaborative weight loss regulation method based on multidimensional health data according to claim 1, characterized in that, The process of decomposing the income and expenditure residual sequence into abrupt and gradual components includes: Apply a median filter with a preset window length to the income and expenditure residual sequence; Apply a moving average with a preset window length to the output of the median filter, and use the cascaded output as the slowly varying component; Subtract the slowly varying components from the income and expenditure residual sequence point by point to obtain the abrupt components; When the number of accumulated time windows at the beginning of the sequence is insufficient to correspond to the preset window length, the two-stage filtering uses the actual covered time windows in the calculation.

5. The intelligent multi-path collaborative weight loss regulation method based on multidimensional health data according to claim 1, characterized in that, The method of correcting an individual's basal metabolic rate by gradually varying component increments within a preset metabolic drift range includes: The initial value of an individual's basal metabolic rate was estimated based on the gender, age, height, and weight of the person trying to lose weight. For each full correction determination period, the average value of the slowly varying components within the current correction determination period is used as the increment of the slowly varying components; The incremental change of the slowly varying component is converted into an energy shift according to the individual energy conversion coefficient. When the energy shift falls within the preset metabolic drift range, the individual basal metabolic rate is corrected according to the energy shift. The corrected individual basal metabolic rate will be used in the energy balance sequence calculation starting from the next time window, and the correction will be performed iteratively on a cycle-by-cycle basis.

6. The intelligent multi-path collaborative weight loss regulation method based on multidimensional health data according to claim 5, characterized in that, The preset metabolic drift range includes: The energy range is limited by a preset ratio based on the currently effective individual basal metabolic rate; When the energy offset exceeds the preset metabolic drift range, correction is performed using the boundary value of the preset metabolic drift range, and the excess portion is retained in the balance residual sequence.

7. The intelligent multi-path collaborative weight loss regulation method based on multidimensional health data according to claim 1, characterized in that, The conversion of individual energy conversion coefficients into predicted change sequences includes: The initial value of the individual energy conversion coefficient is set according to the energy density of adipose tissue; Body composition measurement data of the weight loss subjects are collected according to the preset measurement cycle, and the lean body mass loss ratio is calculated as the ratio of lean body mass loss to body mass loss. When the percentage of lean body mass loss reaches a preset threshold, the individual energy conversion coefficient is lowered, and the lowered value takes effect from the subsequent time window.

8. The intelligent multi-path collaborative weight loss regulation method based on multidimensional health data according to claim 1, characterized in that, The mutation component time period includes: The time period constituted by the mutation window is the time window in which the absolute value of the energy converted from the mutation component exceeds the mutation determination threshold. The mutation determination threshold is set according to a preset multiple of the standard deviation of the energy of the income and expenditure residual sequence on the stable window set. The stable window set is the time window that has not been determined as a mutation window among the most recent preset number of time windows. When the number of accumulated time windows is less than the preset number, the mutation determination threshold is set to a preset initial value. After the number of accumulated time windows reaches the preset number, the calculation value is switched to rolling. When the mutation components of adjacent mutation windows have the same sign, they are merged into one mutation component time period; when the signs are opposite, they belong to two separate mutation component time periods.

9. The intelligent multi-path collaborative weight loss regulation method based on multidimensional health data according to claim 1, characterized in that, The missing status records for the intake and consumption links include: Divide the time window into meal periods. If there are no food record entries in any meal period, mark the corresponding meal period as an intake record gap. Snacks outside of regular meal times are included in the cumulative energy intake but are not included in the sparse food count. The cumulative non-wearing time during the waking period is counted based on the minute-level wearing status flags stored along the consumption link. When the cumulative non-wearing time exceeds the preset judgment time, the current time window is marked as a consumption record gap.

10. The intelligent multi-path collaborative weight loss regulation method based on multidimensional health data according to claim 1, characterized in that, The division of the mutation component time period into the missed recording period and the deviation period includes: The time window with negative mutation components is considered as the distorted link when the mutation component is ingested, and the time window with positive mutation components is considered as the distorted link when the mutation component is consumed. The time window where there are missing records in the distorted link is classified as a missed recording segment; Time windows with no recorded gaps in either the intake or consumption pathways are designated as deviation segments. When there is a missing record in a link and the sign of the mutation component is inconsistent, the sign of the mutation component shall prevail and the segment shall be treated as the deviation segment. The missing and deviation segments are filled and corrected independently window by window.

11. The intelligent multi-path collaborative weight loss regulation method based on multidimensional health data according to claim 1, characterized in that, The missing segments are filled with historical statistical values ​​from the same period of the distorted link, including: When the distortion link is the ingestion link, a time window set is formed by taking time windows with the same day pattern as the time window in the missed segment within a preset number of weeks. The day pattern is divided into weekdays and rest days. The median of the total recorded energy value corresponding to the recording gap in the time window set is used as the filling value and written into the ingestion link. When the distorted link is a consumption link, take the time window with the same week number as the time window in the missing segment within the preset number of weeks, and write the median value of the motion consumption value into the consumption link as the filling value. The filled value is stored with an estimated flag.

12. The intelligent multi-path collaborative weight loss regulation method based on multidimensional health data according to claim 1, characterized in that, The deviation segment corrects the distorted link record value according to the amplitude of the mutation component, including: The amplitude of the mutation component is converted into a recording correction amount according to the individual energy conversion coefficient, and the recording correction amount is added to the recording value of the distorted link in the current time window; When the amount of correction recorded exceeds the preset proportion of the value recorded by the distorted link in the current time window, only the preset proportion will be corrected, and the excess will be retained in the income and expenditure residual sequence and listed as an item to be confirmed in the supplementary recording prompt. If there is a missing weight measurement window filled by linear interpolation within the difference interval of the current time window, the recording correction will be temporarily suspended, and a verification prompt containing a record verification request and a weight re-measurement request will be pushed.

13. The intelligent multi-path collaborative weight loss regulation method based on multidimensional health data according to claim 1, characterized in that, The updated intake and consumption targets for the next cycle include: The corrected energy balance sequence was obtained by recalculating the intake and consumption links after filling and correcting them; The weekly weight loss target is converted into a target weekly deficit based on the individual energy conversion coefficient, and then distributed as a target daily deficit. The opposite of the corrected energy balance is taken as the gap value, and the average gap value within the target update cycle is taken as the average gap. The difference between the target daily deficit and the average deficit will be adjusted to the intake target and the consumption target respectively according to the preset allocation ratio; The intake target is based on the individual's basal metabolic rate as the lower limit. When the lower limit is reached, the remaining difference is transferred to the consumption target.

14. The intelligent multi-path collaborative weight loss regulation method based on multidimensional health data according to claim 1, characterized in that, The push notification for supplementing the distorted link includes: The supplementary information includes the date of the distortion time window, the distortion link identifier, the judgment type, the amount of numerical adjustment already performed, and the confirmation entry. Confirm the list of high-frequency entries corresponding to the pre-filled numerical adjustment amount in the input and the distorted link; In response to the selection operation of the actual item by the weight reduction object, the actual value is backfilled, the backfilled value replaces the written numerical adjustment, and the corrected energy income and expenditure sequence is recalculated. The numerical adjustment remains in effect when backfilling is not performed. Backfilling and replacement only apply to the distortion time window, and targets that have already been issued will not be retrospectively applied.

15. An intelligent multi-path collaborative weight loss regulation system based on multidimensional health data, used to implement the intelligent multi-path collaborative weight loss regulation method based on multidimensional health data as described in any one of claims 1 to 14, characterized in that, include: Link alignment module, residual decomposition module, and record repair module; The link alignment module is used to align the weight records, diet records, and exercise records of the weight loss subject into weight links, intake links, and consumption links according to a unified time window; to filter the weight links by moving average to generate a weight trend sequence that filters out short-term physiological fluctuations; and to obtain the actual change sequence by differentiating the weight trend sequence. The residual decomposition module is used to subtract the individual's basal metabolic rate from the difference between the intake and consumption pathways to obtain an energy balance sequence, which is then converted into a predicted change sequence according to the individual's energy conversion coefficient; the predicted change sequence is subtracted from the actual change sequence to generate a balance residual sequence; the balance residual sequence is decomposed into a mutation component and a slow change component, and the individual's basal metabolic rate is corrected by the increment of the slow change component within a preset metabolic drift interval. The recording and repair module is used to divide the time period of mutation component into missing segment and deviation segment according to the missing status of intake link and consumption link records, and locate the distorted link according to the sign of mutation component; the missing segment is filled with the historical statistical value of the distorted link in the same period, and the deviation segment is corrected according to the amplitude of mutation component to correct the distorted link record value. The intake and consumption targets for the next cycle are updated based on the revised energy balance sequence, and supplementary recording prompts are pushed to the distorted link.

Images