Age-based personalized obesity prescription generation method based on deep learning

Abstract

The application relates to the technical field of obesity exercise analysis, and discloses a method for generating age-section personalized obesity prescriptions based on deep learning, which comprises the following steps: acquiring metabolic data of target objects in different age sections within 24 hours after exercise in real time, analyzing the metabolic data, obtaining metabolic phase angle maps of different age sections, and taking the target objects as obese people; through age-section analysis of 24-hour metabolism and multi-organ data after exercise, a personalized exercise prescription is generated, from a metabolism phase angle map to a rhythm step-out amount, to a metabolism resonance factor and a target direction amount, each link focuses on age differences, accurately captures metabolic fluctuation characteristics of young and middle-aged obese people, and finally, the prescription is combined with a resonance sensitive period, a metabolic elasticity interval and the like, dynamic adaptation of intensity, time length and time period is realized, the problem that a traditional prescription is disconnected with actual metabolism is effectively solved, and the accuracy and effect of obesity intervention are greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of obesity exercise analysis, and particularly relates to a method for generating age-segmented personalized obesity prescriptions based on deep learning. BACKGROUND

[0002] As a key risk factor for various chronic diseases, obesity poses an increasingly severe challenge to the public health system. Exercise therapy, as a first-line method for obesity intervention, has the advantages of non-drug dependence and multiple metabolic benefits.

[0003] In exercise therapy, the exercise prescription for obese people is mainly based on static physiological indicators such as body fat percentage, waist circumference, age, etc., and is generated through pre-set weight loss formulas or experiential exercise templates. Although the prescription can be adjusted during exercise, there are still the following defects in the exercise prescription for different age groups, such as middle-aged people and young people: when adjusting the exercise prescription the day after exercise, the adjustment logic is often based on the relevant data of the previous day's exercise or directly applies a unified metabolic rebound rule, which may not actually consider the fluctuations in resting metabolic rate within 24 hours after exercise for people of different ages, causing the exercise prescription for the next day to be out of touch with the actual metabolic condition and affecting the effectiveness of obesity intervention. SUMMARY

[0004] To overcome the deficiencies of the prior art, the present application provides a method for generating age-segmented personalized obesity prescriptions based on deep learning, which solves the above problems.

[0005] The above technical purposes of the present application are achieved through the following technical solutions:

[0006] The method for generating age-segmented personalized obesity prescriptions based on deep learning comprises:

[0007] Step 1: Real-time acquisition of metabolic data within 24 hours after exercise of target objects of different ages, analysis of the metabolic data, and obtaining of metabolic phase angle maps of different ages, the target objects being obese people;

[0008] Step 2: Real-time acquisition of whole-body multi-organ metabolic data of target objects of different ages within 24 hours after exercise, analysis of the multi-organ metabolic data, and generation of rhythm step-out quantities;

[0009] Step 3: Extraction of the metabolic phase angle maps to generate real-time metabolic physiological matrices, combination analysis of the real-time metabolic physiological matrices and the rhythm step-out quantities, and obtaining of metabolic resonance factors;

[0010] Step 4: Regulation based on the metabolic resonance factors to generate target direction quantities;

[0011] Step 5, calculate the target direction quantity to generate the final movement prescription.

[0012] Further, analyze the metabolic data to obtain the metabolic phase angle atlas of different age groups, including:

[0013] Obtain the exercise intensity of the target object of different age groups the day before, calibrate the pre-processed metabolic data based on the exercise intensity, and generate the exercise coupling correction value;

[0014] Based on the exercise coupling correction value, divide the 24-hour metabolic data into multiple continuous short time sequence segments, calculate the mutation degree of metabolic indicators between adjacent segments, and generate a metabolic time sequence mutation index;

[0015] According to the metabolic time sequence mutation index, for target objects of different age groups, respectively construct dynamic correlation nodes between metabolic indicators, and generate an age-segmented metabolic synergy feature matrix;

[0016] Analyze the age-segmented metabolic synergy feature matrix to generate a metabolic phase shift amount;

[0017] Fuse and analyze the age-segmented metabolic synergy feature matrix and the metabolic phase shift amount to construct a metabolic phase angle atlas of different age groups.

[0018] Further, analyze the multi-organ metabolic data to generate a rhythm out-of-step amount, including:

[0019] Perform core index screening on the pre-processed multi-organ metabolic data to generate an organ metabolic core index set;

[0020] Based on the organ metabolic core index set, calculate the interaction intensity of different organ metabolic signals in time sequence changes, and generate an organ metabolic interaction coefficient;

[0021] Based on the organ metabolic interaction coefficient, analyze the phase matching degree of metabolic signals of liver, adipose tissue and muscle in periodic changes, and generate an organ metabolic phase difference value;

[0022] According to the organ metabolic phase difference value, calculate the phase dislocation degree of the metabolic signals of the liver-adipose tissue-muscle whole, and generate a rhythm out-of-step amount.

[0023] Further, based on the organ metabolic interaction coefficient, analyze the phase matching degree of metabolic signals of liver, adipose tissue and muscle in periodic changes, and generate an organ metabolic phase difference value, including:

[0024] According to the organ metabolic interaction coefficient, extract the rhythm characteristics of the metabolic signals of liver, adipose tissue and muscle within a 24-hour cycle, and generate organ metabolic periodic feature points;

[0025] Based on the feature points of the organ metabolic cycle, the time difference between the peak time of the main cycle of any two organ metabolic signals is calculated, and the organ metabolic phase difference value is generated;

[0026] The organ metabolic phase difference value is analyzed to generate the organ metabolic phase difference value.

[0027] Further, the metabolic phase angle atlas is extracted to generate the real-time metabolic physiological matrix, including:

[0028] The metabolic phase angle atlas of different age groups is dynamically analyzed, and the phase change within 24 hours is disassembled to generate a metabolic phase function unit set;

[0029] The metabolic signal conduction effect between different phase units in the metabolic phase function unit set is analyzed to generate a phase unit linkage coefficient;

[0030] Based on the phase unit linkage coefficient, the activity of the dynamic correlation node is analyzed to generate an age-specific phase master factor;

[0031] Based on the age-specific phase master factor, the metabolic phase function unit set and the phase unit linkage coefficient are reorganized to form a real-time metabolic physiological matrix.

[0032] Further, the real-time metabolic physiological matrix and the rhythm out-of-step amount are combined and analyzed to obtain a metabolic resonance factor, including:

[0033] The connection density and regulation path of each node in the real-time metabolic physiological matrix are analyzed to obtain a metabolic network elasticity coefficient;

[0034] The metabolic network elasticity coefficient and the rhythm out-of-step amount are calculated to generate a metabolic-organ dynamic response rate;

[0035] Based on the metabolic-organ dynamic response rate, the interaction between the metabolic network elasticity coefficient and the rhythm out-of-step amount is calibrated to generate a metabolic resonance factor.

[0036] Further, the connection density and regulation path of each node in the real-time metabolic physiological matrix are analyzed to obtain a metabolic network elasticity coefficient, including:

[0037] The connection strength between each node in the real-time metabolic physiological matrix is analyzed to generate an age-specific node connection weight;

[0038] Based on the age-specific node connection weight, the parallel regulation path existing in the real-time metabolic physiological matrix is identified to generate a metabolic regulation path redundancy;

[0039] Based on the metabolic regulation path redundancy, the fluctuation range of the node connection strength is calculated to generate a node connection fault tolerance coefficient;

[0040] The metabolic network elasticity coefficient is obtained by comprehensively calculating the node connection weight of the age segment, the metabolic regulation path redundancy, and the node connection fault tolerance coefficient.

[0041] Further, based on the metabolic resonance factor, a targeted direction quantity is generated, including:

[0042] Based on the metabolic resonance factor, the resonance sensitive points of the target object metabolism in different age segments are analyzed, and a resonance sensitive map of the age segment is generated;

[0043] The dynamic correlation between the metabolic resonance factor and the metabolic network elasticity coefficient is analyzed, and a metabolic elasticity regulation interval is generated;

[0044] According to the metabolic resonance factor and the organ metabolism interaction coefficient, the optimal coupling strength of each organ metabolism signal is calculated, and an organ signal coupling correction value is generated;

[0045] According to the metabolic rhythm stability reflected by the metabolic resonance factor, a rhythm remodeling time effect parameter is generated;

[0046] The resonance sensitive map of the age segment, the metabolic elasticity regulation interval, the organ signal coupling correction value, and the rhythm remodeling time effect parameter are analyzed, and a targeted direction quantity is generated.

[0047] Further, the dynamic correlation between the metabolic resonance factor and the metabolic network elasticity coefficient is analyzed, and a metabolic elasticity regulation interval is generated, including:

[0048] A dynamic response curve of the metabolic resonance factor and the metabolic network elasticity coefficient is constructed, the corresponding fluctuation law of the elasticity coefficient when the metabolic resonance state changes is captured, and a resonance-elasticity linkage characteristic value is generated;

[0049] The resonance-elasticity linkage characteristic value and the metabolic network elasticity coefficient are calculated, and an age segment elasticity critical value is generated;

[0050] According to the resonance-elasticity linkage characteristic value and the age segment elasticity critical value, the regulation amplitude range of the target object metabolism when the metabolic resonance factor is at different levels is calculated, and a metabolic elasticity regulation interval is generated.

[0051] Further, the targeted direction quantity is calculated, and a final exercise prescription is generated, including:

[0052] The targeted direction quantity is extracted, and the key features directly related to the exercise intervention are screened out, and an exercise intervention core feature value is generated;

[0053] The exercise intervention core feature value and the resonance sensitive map of the age segment are calculated, the metabolic response difference of the target object to different exercise types is analyzed, and an age segment exercise adaptation coefficient is generated;

[0054] Based on the age segment motion adaptation coefficient, the metabolic elasticity regulation interval and the rhythm remodeling time effect parameter are fused, the optimal combination range of motion intensity, time length and implementation time period is calculated, and a motion parameter coordination scheme is generated;

[0055] According to the motion parameter coordination scheme, the target direction quantity is analyzed, and a final motion prescription is formed.

[0056] In summary, the present application mainly has the following beneficial effects:

[0057] By analyzing the metabolic data and the metabolic law of multiple organs in different age segments, the metabolic dynamics of obese people in different age groups are accurately captured, and for 18-35 year-old young people and 36-50 year-old middle-aged people, the phase angle atlas is extracted from the 24-hour metabolic data after exercise, combined with the rhythm out-of-step quantity of multiple organ metabolism, a multi-dimensional evaluation system covering metabolic fluctuations and organ coordination is constructed, the metabolic data is calibrated by the motion coupling correction value, the motion intensity difference interference is eliminated, the differences in the coordination mode of metabolic indicators in different age groups are revealed through dynamic correlation node analysis, the characteristics of young people's fast metabolic recovery and middle-aged people's multiple organ metabolic rhythm more prone to out-of-step are fully considered in the prescription adjustment, and the problem that the traditional method ignores the age-specific metabolic fluctuations is solved.

[0058] By integrating multi-source data through deep learning technology, the metabolic resonance factor and the target direction quantity realize accurate regulation of the metabolic state, from the analysis of the metabolic network elasticity of the real-time metabolic physiological matrix to the evaluation of the organ phase matching degree by the rhythm out-of-step quantity, and the dynamic distribution of multi-parameter weight by the attention mechanism layer, so that the intensity, time length and time period of the exercise prescription can be dynamically adjusted based on the quantified metabolic state, and the risk of disconnection between the prescription and the actual metabolic condition is reduced.

[0059] Through age adaptation and dynamic fine-tuning, the effectiveness and safety of obesity intervention are improved, the motion type is selected based on the age segment motion adaptation coefficient, the implementation time is determined in combination with the resonance sensitive period, the intensity and time length are fine-tuned through organ signal coupling correction value, and a customized scheme of time period-intensity-time length is formed, wherein, for young people, the metabolic sensitive period is concentrated and the elastic regulation interval is wide, so that the exercise prescription can focus on high-intensity short-time exercise, and for middle-aged people, the multiple organ metabolic rhythm is prone to out-of-step, so that the exercise prescription can choose gentle exercise with short rhythm remodeling time effect, and match the metabolic stable period for implementation, this precise adaptation not only maximizes the regulation effect of exercise on metabolism, but also reduces the risk of exercise caused by metabolic fluctuations, providing a more targeted intervention method for obese people in different age groups, and ensuring the weight loss effect of obese people's exercise. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1is a deep learning-based age segment personalized obesity prescription generation method of the present application. DETAILED DESCRIPTION

[0061] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0062] Reference Figure 1 The deep learning-based age segment personalized obesity prescription generation method comprises the following steps.

[0063] Step 1: Real-time acquisition of metabolic data of target objects in different age groups within 24 hours after exercise, analysis of the metabolic data, and obtaining metabolic phase angle maps of different age groups. The target objects are obese people, the different age groups are young people aged 18-35 and middle-aged people aged 36-50, and the metabolic data includes resting metabolic rate, body temperature, respiratory rate, blood glucose, heart rate, blood ketone body level, etc.

[0064] Step 2: Real-time acquisition of whole-body multi-organ metabolic data of target objects in different age groups within 24 hours after exercise, analysis of the multi-organ metabolic data, and generation of rhythm out-of-step quantity. The multi-organ metabolic data includes liver metabolic data, fat metabolic data, muscle metabolic data, cortisol, insulin, etc.

[0065] Step 3: Extraction of the metabolic phase angle map to generate a real-time metabolic physiological matrix, combination analysis of the real-time metabolic physiological matrix and the rhythm out-of-step quantity, and obtaining a metabolic resonance factor.

[0066] Step 4: Regulation based on the metabolic resonance factor to generate a target direction quantity.

[0067] Step 5: Calculation of the target direction quantity to generate a final exercise prescription.

[0068] By real-time collection of resting metabolic rate, blood glucose and other metabolic data of obese people in young age group (18-35 years old) and middle age group (36-50 years old) after exercise, as well as multi-organ metabolic data such as liver and fat, metabolic phase angle maps and rhythm out-of-step quantity are constructed, and then metabolic resonance factor and target direction quantity are generated, and finally personalized exercise prescription is formed. The metabolic fluctuation difference within 24 hours after exercise of people in different age groups is fully considered, the disconnection problem caused by adjusting the prescription based on the data of the previous day or a unified template is avoided, the exercise prescription is more in line with the actual metabolic condition, the weight loss effect of exercise therapy can be more accurately played, and the risk of obesity-related chronic diseases is reduced.

[0069] In one case of the embodiment, the metabolic data is analyzed to obtain metabolic phase angle maps of different age groups, including:

[0070] The exercise intensity of the target object of different ages in the previous day is obtained, the preprocessed metabolic data is calibrated based on the exercise intensity, and an exercise coupling correction value is generated, specifically including: calculating the mean value of the exercise intensity, calculating the average value of each data in the metabolic data when the mean value of the exercise intensity is taken as a reference value, and taking it as a standard reference value; the target object exercise intensity is divided by the reference value to obtain a proportion coefficient; the proportion coefficient is multiplied by the standard reference value of each data in the metabolic data to obtain the theoretical metabolic value under the corresponding exercise intensity; the value of each data in the preprocessed metabolic data is subtracted from the theoretical metabolic value to obtain the deviation of each data, and then all the deviations are added together, and the sum is the exercise coupling correction value;

[0071] Based on the exercise coupling correction value, the 24-hour metabolic data is divided into a plurality of continuous short time sequence segments, the mutation degree of metabolic indicators between adjacent segments is calculated, and a metabolic time sequence mutation index is generated, specifically including: dividing the 24-hour metabolic data into 24 continuous short time sequence segments at an interval of 1 hour, calculating the mean difference of each metabolic indicator (the data of the metabolic data is the metabolic indicator) in the adjacent two segments, and then dividing the mean difference by the time length of the two segments to obtain the instantaneous change rate of each indicator. The sum of the absolute values of the sum of the instantaneous change rates of all indicators, the metabolic time sequence mutation index can be obtained;

[0072] According to the metabolic time sequence mutation index, the dynamic correlation nodes between metabolic indicators are constructed for target objects of different age groups, and a metabolic coordination feature matrix of different age groups is generated. Specifically, for the metabolic indicators such as resting metabolic rate and body temperature of the target object, the values of each indicator in the 24-hour short time sequence segment are taken, and the correlation degree between each two indicators is calculated one by one by taking the metabolic time sequence mutation index as the weight. First, the metabolic time sequence mutation index is multiplied by the mean difference of the values of the two indicators, then the sum of the products is calculated, and then the sum is divided by the product of the standard deviation of the two indicators and the sample size (the effective data quantity of the target object in the 24-hour short time sequence segment of a certain group of metabolic indicators), to obtain the correlation degree between the two indicators. The correlation degrees are input into the graph neural network layer of deep learning, and the graph neural network layer compares the correlation degrees one by one by a fixed threshold (which can be set according to actual conditions), and retains the indicator pairs that exceed the threshold to determine them as dynamic correlation nodes. The dynamic correlation nodes are sorted in descending order of correlation degree, and the correlation degrees of the nodes are added in turn according to the order. Each time a node is added, the cumulative value is recorded once. The different cumulative values are associated with the corresponding 24-hour short time sequence segment, and the correlation degree sorting result of each node is taken as the row index of the matrix (such as the first row corresponding to the first sorted node), and the time point of the 24-hour short time sequence segment is taken as the column index. Each element value in the matrix integrates three parts of information through coding: the correlation degree sorting sequence number of the corresponding node, the cumulative value change amount of the node in the corresponding time segment, and the time point information, so as to form a metabolic coordination feature matrix of different age groups.

[0073] The metabolic coordination feature matrix of different age groups is analyzed to generate a metabolic phase shift amount. Specifically, the cumulative value change amount of each time point in each row of the metabolic coordination feature matrix of different age groups is extracted to form a 24-hour time sequence of the node, and a metabolic time sequence belonging to the age group standard is obtained. The dot product of the 24-hour time sequence of the node (24 numerical values) and the corresponding position numerical values of the standard metabolic time sequence (24 reference numerical values) of the same age group is obtained by multiplying and summing the corresponding position numerical values. The square root of the sum of the squares of all numerical values of the node time sequence is calculated to obtain the length of the node time sequence. The square root of the sum of the squares of all numerical values of the standard metabolic time sequence is calculated to obtain the length of the standard metabolic time sequence. The similarity is obtained by dividing the dot product by the product of the two lengths. The single node phase deviation is obtained by subtracting the similarity from 1. The node correlation degree sorting sequence number is taken as the weight of the single node phase deviation (the first sorting weight is 1, the second sorting weight is 0.8, and the weight decreases by 0.2 in turn). The sum of the products of all single node deviations and their corresponding weights is obtained, and then the sum is divided by the total number of nodes to obtain the metabolic phase shift amount.

[0074] The metabolic phase angle atlas of different age groups is constructed by fusing and analyzing the metabolic coordination feature matrix of the age segment and the metabolic phase shift amount, specifically including: multiplying each element value in the metabolic coordination feature matrix of the age segment with the metabolic phase shift amount of the corresponding time point node to obtain the fusion value of each element; arranging the fusion values of each time point in 24 hours in time sequence, calculating the difference value of the fusion values of adjacent time points, and then taking the sum of the absolute values of all difference values as the overall fluctuation amount; dividing the fusion value of each time point by the overall fluctuation amount to obtain the normalized time sequence characteristic value; setting the threshold range corresponding to the phase angle of 0-360 degrees (1 degree corresponds to 0.003-0.006, 2 degrees correspond to 0.006-0.009, and the range increases by 0.003 for each increase of 1 degree), comparing the time sequence characteristic value of each time point with the threshold range, and matching the corresponding phase angle (for example, if the time sequence characteristic value falls within the 0.003-0.006 interval corresponding to 1 degree, the phase angle of 1 degree is matched), and after matching the corresponding phase angle for each time point in 24 hours according to the rule, arranging these phase angles in time sequence, taking time as the horizontal axis and phase angle (0-360 degrees) as the vertical axis, marking the phase angle corresponding to each time point in the coordinate system, and connecting the marking points of adjacent time points with a polyline to form a continuous curve pattern, which is the metabolic phase angle atlas.

[0075] The metabolic data is calibrated by the motion coupling correction value, the interference of the motion intensity difference on the metabolic indicators is effectively eliminated, the data is more in line with the true metabolic state of the obese population in different age groups, the theoretical metabolic value is calculated based on the average motion intensity, and then the correction value is generated through the sum of the deviation amounts, so that the metabolic data for subsequent analysis is not affected by the motion intensity fluctuation, and the motion prescription generated based on the data can more objectively reflect the metabolic characteristics of people in different age groups.

[0076] By dividing the short time sequence segment, calculating the mutation index, and constructing the dynamic correlation node, the metabolic fluctuation rule of people in different age groups within 24 hours after exercise is accurately captured, the fusion and analysis of the metabolic coordination feature matrix of the age segment and the metabolic phase shift amount, and the differences in the metabolic index correlation mode and time sequence phase of obese people in youth and middle age are intuitively presented, avoiding the neglect of age-specific metabolic characteristics by a unified template, so that the next day's exercise prescription adjustment is more in line with the actual metabolic status of each age group, and the obesity intervention effect is further improved.

[0077] In one case of the embodiment, the multi-organ metabolic data is analyzed to generate a rhythm step-out amount, including:

[0078] The core index set of multi-organ metabolism data is screened, and the core index set of organ metabolism is generated, specifically including: calculating the time sequence fluctuation amplitude (the difference between the maximum value and the minimum value of the index within 24 hours) of each index (the index is liver metabolism data, fat metabolism data, muscle metabolism data, etc.) of the pre-processed multi-organ metabolism data, and retaining the index whose time sequence fluctuation amplitude is greater than 1.2 times the average value of the time sequence fluctuation amplitudes of all indexes; calculate the Spearman rank correlation coefficient between the remaining indexes: for any two indexes, there are 24 ranks (the 24 values of the index are sorted in size, the smallest value corresponds to rank 1, the second smallest corresponds to rank 2, and so on to the largest, which corresponds to rank 24), calculate the difference value of the rank at each corresponding position (two indexes at the same time point), then square each of these difference values and add them together to get the sum of squares; multiply 6 by the sum of squares, divide by the sample size (24, indicating the number of samples collected, because there is one time point per hour, corresponding to 24 times of collection), then multiply by (the square of the sample size minus 1), get the product result, let 1 minus the product result, the value obtained is the Spearman rank correlation coefficient of the two indexes; if the absolute value of the Spearman rank correlation coefficient is greater than 0.7, compare the time sequence fluctuation amplitudes of the two indexes, remove the index with the smaller value from the remaining indexes, and only retain the index with the larger fluctuation amplitude; when all the indexes with excessive Spearman rank correlation coefficients are processed, the remaining indexes constitute the core index set of organ metabolism.

[0079] Based on the organ metabolism core index set, the interaction intensity of different organ metabolism signals in the time sequence change is calculated, and the organ metabolism interaction coefficient is generated, specifically including: for the indexes in the organ metabolism core index set, the time sequence data (1 value per hour) of each index in 24 hours is input into the attention mechanism layer of deep learning in time sequence to calculate the covariation rate of two organ signals at each time point: for example, taking liver metabolism data and fat metabolism data as an example, the mean values of the liver data and fat data in 24 hours time sequence data are calculated; for each time point (a total of 24), the liver data value at the time point is subtracted from the mean value of the liver data to obtain the liver deviation; the fat data value at the time point is subtracted from the mean value of the fat data to obtain the fat deviation; then the liver deviation is multiplied by the fat deviation to obtain the product at the time point, and finally the products of the 24 time points are added and divided by 24 to obtain the covariation rate; the covariation rate is filled into the weight matrix of the attention mechanism layer, and the interaction score of each pair of organs is obtained through linear transformation of the fully connected layer: wherein the weight matrix of the attention mechanism layer is a square matrix, and the rows and columns correspond to each organ (for example, row 1 is liver and row 2 is fat, column 1 is liver and column 2 is fat), the covariation rate of liver and fat is filled into the first row and the second column and the second row and the first column of the matrix, and the covariation rate of other organ pairs is filled in according to the rule to form a complete matrix; the fully connected layer sets a unique weight (for example, the liver-fat pair corresponds to the weight a) and a bias value (for example, the liver-fat pair corresponds to the bias value b) for each organ pair, multiplies the covariation rate of the organ pair in the attention matrix by a, and adds the bias value b to obtain the interaction score of the organ pair; each interaction score is divided by the sum of all organ pair interaction scores to obtain the organ metabolism interaction coefficient;

[0080] Based on the organ metabolism interaction coefficient, the phase matching degree of the metabolism signals of liver, adipose tissue and muscle in the periodic change is analyzed, and the organ metabolism phase difference value is generated;

[0081] According to the organ metabolism phase difference value, the metabolism signal phase misplacement degree of liver-adipose tissue-muscle as a whole is calculated, and the rhythm step loss amount is generated, specifically including: for the organ metabolism phase difference values of liver-fat, liver-muscle and fat-muscle, the absolute deviation of each organ metabolism phase difference value from the standard phase difference (0 degree) is calculated; the three absolute deviations are weighted and summed with the organ metabolism interaction coefficient as the weight, and then divided by 3 to obtain the mean value, which is the metabolism signal phase misplacement degree of liver-adipose tissue-muscle as a whole, i.e. the rhythm step loss amount.

[0082] The scientific screening of the organ metabolism core index set improves the analysis efficiency and accuracy of multi-organ metabolism data. High sensitivity indicators are selected according to the timing fluctuation amplitude, and redundant indicators are removed through the Spearman rank correlation coefficient. The core indicators with more significant fluctuations are retained to ensure that the subsequent analysis focuses on key indicators that can truly reflect the metabolic state, making the analysis of multi-organ metabolic synergy more targeted and allowing exercise prescriptions to better fit the metabolic characteristics of obese people of different ages.

[0083] The organ metabolism interaction coefficient is calculated through the attention mechanism layer calculator, and the rhythm step-out amount is generated by combining the phase difference value. The dynamic correlation and phase matching of multi-organ metabolism such as liver, fat, and muscle are comprehensively captured, reflecting the strength difference of metabolic interaction between different organs, making up for the shortcomings of traditional methods that ignore the multi-organ coordinated metabolism rule. The final exercise prescription adjustment fully considers the multi-organ metabolic rhythm characteristics of different age groups within 24 hours after exercise, further reducing the gap between the final exercise prescription and the actual metabolic status, and ensuring the intervention effect of obese people's exercise.

[0084] In one case of the embodiment, based on the organ metabolism interaction coefficient, the phase matching degree of the metabolic signals of liver, adipose tissue, and muscle in the periodic change is analyzed to generate organ metabolism phase difference values, including:

[0085] According to the organ metabolism interaction coefficient, the rhythm characteristics of the metabolic signals of liver, adipose tissue, and muscle within a 24-hour cycle are extracted to generate organ metabolism cycle feature points, specifically including: according to the organ metabolism interaction coefficient, the 24-hour metabolic signals of liver, fat, and muscle are calculated using the sliding window method to calculate the mean value of the signals in each window: the window size is 3 hours, starting from 0 and sliding 1 hour every hour, resulting in 22 windows; the mean value of the metabolic signal values in each window is calculated; for organ pairs with an organ metabolism interaction coefficient greater than 0.5, the maximum and minimum values in the 22 window mean values are found, and the corresponding window start time points are the time points of the maximum and minimum mean values; the difference between the mean values of adjacent windows is calculated and divided by 1 hour (sliding interval) to obtain the mean value change rate; the three time points of the maximum and minimum signal mean values and the maximum mean value change rate are taken as the organ metabolism cycle feature points;

[0086] Based on the organ metabolic cycle feature points, the time difference of the main cycle peak time of any two organ metabolic signals is calculated, and the organ metabolic phase difference value is generated, specifically including: for any two organs (such as liver and fat), the time point with the maximum signal mean value in the respective metabolic cycle feature point is taken as the main cycle peak time; the time difference of the two time points of the two organs (such as the liver peak at 8 o'clock and the fat peak at 10 o'clock, then the time difference is 2 hours); the time difference is multiplied by 15 (because 24 hours corresponds to 360 degrees, and each hour corresponds to 15 degrees), and the obtained value is the organ metabolic phase difference value of the two organs;

[0087] The organ metabolic phase difference value is analyzed to generate an organ metabolic phase difference value, specifically including: subtracting 180 (representing 180 degrees) from the organ metabolic phase difference value to obtain a deviation; taking the organ metabolic interaction coefficient of the corresponding organ pair as the weight, multiplying the deviation by (1 minus the weight) and then dividing by 180 to obtain the organ metabolic phase difference value.

[0088] By extracting organ metabolic cycle feature points, calculating phase difference values and generating difference values, the phase matching degree of liver, fat and muscle metabolic signals is quantified, and the sliding window method combined with organ metabolic interaction coefficient is used to screen feature points, focusing on the key time nodes of high interaction intensity organ pairs, ensuring the pertinence of feature capture, and converting the time difference into phase difference value, reflecting the time dislocation of the metabolic rhythm between organs, improving the matching degree of the final exercise prescription and the actual metabolic condition, and ensuring the effect of weight loss of obese personnel through exercise.

[0089] In one case of the embodiment, the metabolic phase angle graph is extracted to generate a real-time metabolic physiological matrix, including:

[0090] The metabolic phase angle graph of different age groups is dynamically analyzed in time sequence, and the phase change within 24 hours is disassembled to generate a metabolic phase function unit set, specifically including: dividing 24 hours into 12 short periods at intervals of 2 hours for the metabolic phase angle graph of different age groups, calculating the fluctuation amplitude (maximum value minus minimum value) of the phase angle in each short period, and sorting the fluctuation amplitudes from large to small; the periods in the top 30% of the fluctuation amplitude sorting are taken as active units, the middle 40% of the periods are taken as stable units, and the last 30% of the periods are taken as resting units, each unit includes the corresponding period and the numerical range of the phase angle in the period, and the active units, stable units and resting units are combined to obtain the metabolic phase function unit set;

[0091] The metabolic signal transduction effects between different phase units in the metabolic phase functional unit set are analyzed to generate a phase unit linkage coefficient, specifically including: for active, stable and resting units in the metabolic phase functional unit set, calculating the overlap period of adjacent units (such as the active units 12-14 hours and the stable units 14-16 hours, then the overlap period is 14 hours), calculating the absolute value of the phase angle difference value of the two units at the overlap time, dividing the absolute value by 360 (representing 360 degrees), obtaining a basic coefficient, multiplying the fluctuation amplitude average of the two units by the basic coefficient, and the phase unit linkage coefficient of the two units is obtained;

[0092] Based on the phase unit linkage coefficient, the activity of the dynamic correlation node is analyzed to generate a phase control factor, specifically including: calculating the activity of the dynamic correlation node in each phase unit (the node correlation degree multiplied by the fluctuation amplitude of the unit, and the activity is obtained); taking the phase unit linkage coefficient as the weight, weighting and summing the activity of the nodes in the same group, and then dividing by the total number of nodes to obtain the calculation result, taking the absolute value of the difference between the calculation result and the respective standard value, and the phase control factor is obtained;

[0093] Based on the phase control factor, the metabolic phase functional unit set and the phase unit linkage coefficient are reorganized to form a real-time metabolic physiological matrix, specifically including: taking the phase control factor as the weight, multiplying the function information (type (active, stable, resting), time period and phase range (such as 200-300 degrees)) of each unit in the metabolic phase functional unit set with the corresponding phase unit linkage coefficient to obtain the product result; the nodes in the dynamic correlation node are sorted from high to low according to the activity, and for the top 3 nodes, their serial numbers (such as 1, 2, 3) are recorded; when constructing the matrix, the row index is set as the unit type, and the column index is set as the time period corresponding to each unit (such as 12-14 hours for active units); the element value is integrated with the product result and the node serial number (such as the product result is 0.06, the node serial number is 1, 2, 3, and the element value is 0.06_1_2_3); to form a real-time metabolic physiological matrix.

[0094] By disassembling the metabolic phase angle atlas to generate a functional unit set, calculating the linkage coefficient and reorganizing the matrix, the metabolic dynamic characteristics of obese people in different age groups within 24 hours after exercise are captured. The 24 hours are divided into active, stable and resting units, and the metabolic state differences are distinguished by combining the fluctuation amplitude, avoiding the limitations of traditional static analysis. At the same time, the phase unit linkage coefficient quantifies the signal transmission effect of adjacent units, understands the continuous change of metabolic state, and highlights the metabolic regulation core of each age group. The final metabolic physiological matrix integrates the unit function, linkage strength and key node information, makes up for the defects of traditional methods in paying insufficient attention to the details of metabolic timing fluctuations, and makes the exercise prescription accurately match the real-time metabolic state of people in different age groups, improving the adaptability and effectiveness of obesity intervention.

[0095] In one case of the embodiment, the real-time metabolic physiological matrix and the rhythm out-of-step quantity are combined and analyzed to obtain a metabolic resonance factor, including:

[0096] The connection density and regulation path of each node in the real-time metabolic physiological matrix are analyzed to obtain a metabolic network elasticity coefficient.

[0097] The metabolic network elasticity coefficient and the rhythm out-of-step quantity are calculated to generate a metabolic-organ dynamic response rate, specifically including: calculating the absolute difference between the metabolic network elasticity coefficient and the rhythm out-of-step quantity, subtracting the absolute difference from 1, and multiplying the organ metabolic interaction coefficient of the corresponding organ pair to obtain the metabolic-organ dynamic response rate.

[0098] Based on the metabolic-organ dynamic response rate, the interaction between the metabolic network elasticity coefficient and the rhythm out-of-step quantity is calibrated to generate a metabolic resonance factor, specifically including: taking the metabolic-organ dynamic response rate as a weight, adding the metabolic network elasticity coefficient and the rhythm out-of-step quantity, and multiplying the weight to obtain the metabolic resonance factor.

[0099] By combining the real-time metabolic physiological matrix and the rhythm out-of-step quantity to generate the metabolic resonance factor, the metabolic network elasticity coefficient obtained by analyzing the connection density and regulation path of the nodes in the matrix reflects the dynamic regulation ability of the metabolic system. The metabolic-organ dynamic response rate calculated by combining the metabolic network elasticity coefficient and the rhythm out-of-step quantity, and the metabolic resonance factor finally generated by taking the response rate as the weight, integrate the metabolic dynamic characteristics and organ rhythm out-of-step information, so that the final exercise prescription adjustment can be based on the overall metabolic state to fully adapt to the metabolic fluctuations and organ rhythm characteristics within 24 hours after exercise for people in different age groups.

[0100] In one case of the embodiment, the connection density and regulation path of each node in the real-time metabolic physiological matrix are analyzed to obtain a metabolic network elasticity coefficient, including:

[0101] The connection strength between each node in the real-time metabolic physiological matrix is analyzed to generate an age-section node connection weight, specifically including: calculating the mean value of the product (here, the product is the connection strength) of the values of any two nodes in the real-time metabolic physiological matrix; and then dividing the mean value by the sum of the fluctuation amplitudes of the cells where the two nodes are located, to obtain the connection weight, which is the age-section node connection weight;

[0102] Based on the age-section node connection weight, parallel regulation paths existing in the real-time metabolic physiological matrix are identified to generate a metabolic regulation path redundancy, specifically including: for the real-time metabolic physiological matrix, the number of parallel paths with the same starting and ending nodes among the connected nodes in the real-time metabolic physiological matrix is counted respectively, the sum of the age-section node connection weights of each path is calculated, and the smallest sum is taken as a reference; dividing the sum of all paths by the product of the reference and the number of paths to obtain the metabolic regulation path redundancy;

[0103] Based on the metabolic regulation path redundancy, the fluctuation range of the node connection strength is calculated to generate a node connection fault tolerance coefficient, specifically including: calculating the difference between the maximum value and the minimum value of the node connection strength in the real-time metabolic physiological matrix to obtain the fluctuation range, multiplying the metabolic regulation path redundancy by the fluctuation range, and then dividing by the mean value of the node connection strength to obtain the node connection fault tolerance coefficient;

[0104] The age-section node connection weight, the metabolic regulation path redundancy, and the node connection fault tolerance coefficient are comprehensively calculated to obtain a metabolic network elasticity coefficient, specifically including: multiplying the age-section node connection weight by 0.4, multiplying the metabolic regulation path redundancy by 0.3, multiplying the node connection fault tolerance coefficient by 0.3, and adding the three product results to obtain the metabolic network elasticity coefficient; wherein the age-section node connection weight directly reflects the metabolic basic connection strength of the target object of different ages, which is the core basis of regulation, so the weight is the highest, which is 0.4; the metabolic regulation path redundancy reflects the compensatory ability during regulation, and the node connection fault tolerance coefficient reflects the stability during regulation, both of which are equally important for maintaining metabolic elasticity, so the weight of both is 0.3, which meets the actual needs of obesity population age-section metabolic regulation.

[0105] By calculating the age-section node connection weight, the basic connection strength difference between metabolic nodes of different age groups of obese people is accurately captured, and the ratio of the mean value of the product of the node values to the sum of the fluctuation amplitudes of the cells is taken as the weight, which reflects the actual strength of the node connection, and also combines the fluctuation characteristics of the metabolic cells, avoiding the neglect of age specificity in the unified calculation method, making the metabolic network analysis more in line with the metabolic characteristics of young and middle-aged people.

[0106] The metabolic network elasticity coefficient is generated by the age segment node connection weight, metabolic regulation path redundancy and node connection fault tolerance coefficient, which comprehensively reflects the regulation ability of the metabolic system. By giving reasonable weights to the three, the basic role of the core connection strength is highlighted, and the compensation ability and stability are also considered, which is consistent with the actual metabolic regulation of obese people in different ages. The dynamic elasticity of the metabolic network is quantified, and the differences in metabolic network resilience within 24 hours after exercise are fully considered in the exercise prescription adjustment, further narrowing the gap between the prescription and the actual metabolic status.

[0107] In one case of the embodiment, based on the metabolic resonance factor, a target direction quantity is generated, including:

[0108] Based on the metabolic resonance factor, the resonance sensitive points of the metabolic resonance of the target object in different ages are analyzed, and an age segment resonance sensitive map is generated. Specifically, the metabolic resonance factor is corresponded to each time point in 24 hours, the absolute difference between the metabolic resonance factor value and the resonance reference value (0.5) is calculated, the time points with a difference less than 0.1 are taken as resonance sensitive points, and the resonance sensitive points are labeled in time sequence. All resonance sensitive points are connected into a line, with time as the horizontal axis and factor value as the vertical axis, forming an age segment resonance sensitive map.

[0109] The dynamic correlation between the metabolic resonance factor and the metabolic network elasticity coefficient is analyzed, and a metabolic elasticity regulation interval is generated.

[0110] According to the metabolic resonance factor and the organ metabolic interaction coefficient, the optimal coupling strength of each organ metabolic signal is calculated, and an organ signal coupling correction value is generated. Specifically, the metabolic resonance factor is multiplied by the organ metabolic interaction coefficient to obtain the optimal coupling strength, and then the difference between the optimal coupling strength and the standard coupling strength (0.5) is calculated. The absolute value of the difference is taken and multiplied by 0.8 to obtain the organ signal coupling correction value.

[0111] According to the metabolic rhythm stability reflected by the metabolic resonance factor, a rhythm remodeling time effectiveness parameter is generated. Specifically, the difference between the maximum value and the minimum value of the metabolic resonance factor within 24 hours is calculated, and then the difference is divided by 0.5 (the reference fluctuation value) and multiplied by 12 hours (the basic time effectiveness) to obtain the rhythm remodeling time effectiveness parameter. The larger the value of the rhythm remodeling time effectiveness parameter, the longer the time required for rhythm remodeling.

[0112] The analysis of the age segment resonance sensitive map, the metabolic elasticity regulation interval, the organ signal coupling correction value and the rhythm remodeling time effect parameter generates a targeted direction quantity, specifically including: for the proportion of the number of resonance sensitive points in the age segment resonance sensitive map, the width of the metabolic elasticity regulation interval, the organ signal coupling correction value, and the rhythm remodeling time effect parameter, input the four data into the deep learning attention mechanism layer, multiply the four values by the row vectors of the learnable parameter matrix respectively, sum them up, and normalize the sum result to obtain the weights of the four; then add the four values multiplied by their corresponding weights to generate the targeted direction quantity; wherein the learnable parameter matrix is a 4x4 numerical matrix optimized in the deep learning model training.

[0113] The age segment resonance sensitive map generated by analyzing the metabolic resonance factor accurately locates the key sensitive time points of metabolism in different age groups of obese people. The sensitive points are screened by the difference between the metabolic resonance factor and the reference value. The active period of metabolic resonance within 24 hours is understood. The problem of ambiguous identification of metabolic sensitive period in traditional methods is avoided. The exercise intervention for young and middle-aged obese people can focus on the period with the strongest metabolic resonance effect. The specificity of exercise on metabolic regulation is improved, and the defect that static indicators cannot reflect the metabolic timing sensitivity characteristics is remedied.

[0114] By integrating multi-dimensional parameters such as the age segment resonance sensitive map and the metabolic elasticity regulation interval, and generating a targeted direction quantity through the attention mechanism layer, comprehensive and accurate control of metabolic regulation is achieved. The attention mechanism gives dynamic weights to each parameter, highlighting core factors such as resonance sensitive points and organ coupling strength, while also considering metabolic elasticity and rhythm remodeling time effect, meeting the differentiated needs of metabolic regulation in different age groups, integrating metabolic dynamic characteristics and organ coordination rules, and allowing exercise prescriptions to accurately match the metabolic regulation needs of different age groups.

[0115] In one case of the present embodiment, the dynamic correlation between the metabolic resonance factor and the metabolic network elasticity coefficient is analyzed to generate a metabolic elasticity regulation interval, including:

[0116] A dynamic response curve of the metabolic resonance factor and the metabolic network elasticity coefficient is constructed to capture the corresponding fluctuation law of the elasticity coefficient when the metabolic resonance state changes, and a resonance-elasticity linkage characteristic value is generated. Specifically, the metabolic resonance factor for 24 hours is taken as the horizontal axis, the metabolic network elasticity coefficient is taken as the vertical axis, the points are labeled in turn and connected to obtain the dynamic response curve; the absolute value of the slope of adjacent points in the dynamic response curve is calculated, and the mean value of the absolute value of the slope is calculated. Multiply the mean value by the product of the metabolic resonance factor and the metabolic network elasticity coefficient to obtain the resonance-elasticity linkage characteristic value;

[0117] The resonance-elasticity linkage characteristic value and the metabolic network elasticity coefficient are calculated to generate an age segment elasticity critical value, specifically including: calculating the sum of the maximum value and the minimum value of the metabolic network elasticity coefficient in 24 hours, dividing the sum by 2 to obtain a coefficient benchmark; multiplying the resonance-elasticity linkage characteristic value by 0.5 and then adding the coefficient benchmark to obtain the age segment elasticity critical value;

[0118] According to the resonance-elasticity linkage characteristic value and the age segment elasticity critical value, the regulation range of the metabolism of the target object is calculated when the metabolic resonance factor is at different levels to generate a metabolic elasticity regulation interval, specifically including: dividing the metabolic resonance factor in 24 hours into three groups of low (metabolic resonance factor ≤ 0.3), medium (metabolic resonance factor between 0.3 and 0.7), and high (metabolic resonance factor ≥ 0.7) at intervals of 0.2; wherein the metabolic elasticity regulation interval of the low group is: the lower limit of the interval is obtained by subtracting twice the resonance-elasticity linkage characteristic value from the age segment elasticity critical value, and the upper limit of the interval is obtained by adding the resonance-elasticity linkage characteristic value to the age segment elasticity critical value; the metabolic elasticity regulation interval of the medium group is: the upper limit (addition) and the lower limit (subtraction) of the interval are obtained by adding or subtracting the resonance-elasticity linkage characteristic value from the age segment elasticity critical value; the metabolic elasticity regulation interval of the high group is: the upper limit of the interval is obtained by adding twice the resonance-elasticity linkage characteristic value to the age segment elasticity critical value, and the lower limit of the interval is obtained by subtracting the resonance-elasticity linkage characteristic value from the age segment elasticity critical value; the metabolic elasticity regulation interval of each group is obtained.

[0119] By constructing a dynamic response curve of the metabolic resonance factor and the metabolic network elasticity coefficient, combined with the age segment elasticity critical value, the metabolic elasticity regulation interval under different metabolic resonance levels is accurately divided, the dynamic response curve captures the fluctuation correlation of the two, and the resonance-elasticity linkage characteristic value quantifies the correlation strength, finally generating differentiated intervals for low, medium and high metabolic resonance levels, clearly defining the metabolic regulation range in each state, so that the exercise prescription can accurately adjust within the reasonable interval according to the metabolic resonance state of the population in different age groups within 24 hours after exercise, ensuring that the intervention is neither excessive nor ineffective.

[0120] In one case of the embodiment, a target direction quantity is calculated to generate a final exercise prescription, including:

[0121] The targeted direction quantity is extracted, and key features directly related to the exercise intervention are screened to generate exercise intervention core feature values, specifically including: the resonance sensitive point number ratio of the targeted direction quantity, the metabolic elasticity regulation interval width, wherein the resonance sensitive point number ratio is used to reflect the proportion of the metabolic sensitive period, and the resonance sensitive point number ratio is multiplied by 0.6 (intensity coefficient) to obtain the exercise intensity benchmark value; the metabolic elasticity regulation interval width is used to reflect the metabolic adjustable space, and the metabolic elasticity regulation interval width is multiplied by 0.5 (time length coefficient) to obtain the exercise time length benchmark value; the exercise intensity benchmark value and the exercise time length benchmark value are added to obtain the exercise intervention core feature value of the integrated intensity and time length benchmark.

[0122] The exercise intervention core feature value and the age segment resonance sensitive map are calculated to analyze the metabolic response differences of the target object to different exercise types, and an age segment exercise adaptation coefficient is generated, specifically including: calculating the mean value of the resonance sensitive points in the age segment resonance sensitive map, multiplying the mean value by the exercise intervention core feature value to obtain the exercise response benchmark value; setting the standard response values of different exercise types (active, stable, and resting) as 0.5, 0.3, and 0.2, respectively, multiplying the different standard response values by the exercise response benchmark value to obtain the adaptation values of each exercise type, and the adaptation values of each exercise type are the age segment exercise adaptation coefficients;

[0123] Based on the age segment exercise adaptation coefficient, the metabolic elasticity regulation interval and the rhythm remodeling time efficiency parameter are fused to calculate the optimal combination range of exercise intensity, time length and implementation period, and generate an exercise parameter coordination scheme, specifically including: taking the highest value of the age segment exercise adaptation coefficient as a weight, multiplying each value of the metabolic elasticity regulation interval by the weight to obtain an intensity range; multiplying the rhythm remodeling time efficiency parameter by the weight to obtain a time length; taking the period of the resonance sensitive point as the implementation period, integrating the exercise intensity range, exercise time length, and implementation period in the format of intensity range-time length-period to obtain the exercise parameter coordination scheme;

[0124] According to the exercise parameter coordination scheme, the targeted direction quantity is analyzed to form a final exercise prescription, specifically including: calculating the median value of the intensity range in the exercise parameter coordination scheme, and taking the time length and the period as the basic parameters, multiplying the median value and the time length by the organ signal coupling correction value to obtain the adjusted intensity and time length; combining the period, the adjusted intensity, and the time length to form a "period-intensity-time length" prescription, which is the final exercise prescription. The final exercise prescription is an individualized obesity exercise prescription for different ages.

[0125] The final exercise prescription is generated by extracting the core characteristic value of exercise intervention, calculating the age section exercise adaptation coefficient and other steps, realizing the individualized and accurate intervention on the obese population of different age groups. The core characteristic value of exercise intervention integrates the strength and duration benchmark, analyzes the exercise response difference combined with the age section resonance sensitive map, provides a reference for exercise type selection, and determines the optimal exercise intensity, duration and time period by fusing the metabolic elasticity regulation interval and rhythm remodeling time efficiency parameter of the exercise parameter coordination scheme. The customized prescription of time period-intensity-duration is formed by fine tuning through organ signal coupling correction value, fully considering the metabolic fluctuation of different age groups within 24 hours after exercise, avoiding the disadvantages of traditional template prescription, making the exercise prescription highly matched with the actual metabolic condition, and improving the pertinence of obesity intervention.

[0126] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A deep learning-based age-segmented personalized obesity prescription generation method, characterized in that, The method comprises the following steps: Step 1, real-time acquisition of metabolic data of target objects of different ages within 24 hours after exercise, analysis of the metabolic data, and obtaining metabolic phase angle atlas of different ages: Obtain the exercise intensity of the target objects of different ages the day before, calibrate the preprocessed metabolic data based on the exercise intensity, and generate exercise coupling correction value; Based on the exercise coupling correction value, the 24-hour metabolic data is divided into multiple continuous short time sequence segments, the mutation degree of metabolic indicators between adjacent segments is calculated, and a metabolic time sequence mutation index is generated; According to the metabolic time sequence mutation index, the dynamic correlation nodes between metabolic indicators are constructed for target objects of different ages, and a segmented metabolic coordination feature matrix is generated; Analyzing the segmented metabolic coordination feature matrix generates a metabolic phase shift; Fusion analysis of the segmented metabolic coordination feature matrix and the metabolic phase shift constructs the metabolic phase angle atlas of different ages; The target objects are obese people; Step 2, real-time acquisition of multi-organ metabolic data of target objects of different ages within 24 hours after exercise, analysis of the multi-organ metabolic data, and generation of rhythm step-out quantity: Core index screening is performed on the preprocessed multi-organ metabolic data to generate an organ metabolic core index set; Based on the organ metabolic core index set, the interaction intensity of different organ metabolic signals in time sequence change is calculated to generate an organ metabolic interaction coefficient; Based on the organ metabolic interaction coefficient, the phase matching degree of metabolic signals of liver, adipose tissue and muscle in periodic change is analyzed to generate an organ metabolic phase difference value; According to the organ metabolic phase difference value, the phase dislocation degree of the metabolic signals of liver-adipose tissue-muscle as a whole is calculated to generate a rhythm step-out quantity; Step 3, extracting the metabolic phase angle atlas to generate a real-time metabolic physiological matrix: Dynamic time sequence analysis is performed on the metabolic phase angle atlas of different ages, the phase change within 24 hours is decomposed, and a metabolic phase function unit set is generated; The metabolic signal conduction effect between different phase units in the metabolic phase function unit set is analyzed to generate a phase unit linkage coefficient; Based on the phase unit linkage coefficient, the activity of the dynamic correlation node is analyzed to generate a segmented phase master factor; Based on the segmented phase master factor, the metabolic phase function unit set and the phase unit linkage coefficient are reorganized to form a real-time metabolic physiological matrix; Combination analysis of the real-time metabolic physiological matrix and the rhythm step-out quantity generates a metabolic resonance factor: Analyze the connection density and regulation path of each node in the real-time metabolic physiological matrix to obtain a metabolic network elasticity coefficient; The metabolic-organ dynamic response rate is generated by calculating the metabolic network elasticity coefficient and the rhythm step-out quantity; Based on the metabolic-organ dynamic response rate, the interaction between the metabolic network elasticity coefficient and the rhythm step-out quantity is calibrated to generate a metabolic resonance factor; Step 4, based on the metabolic resonance factor, the target direction quantity is generated: Based on the metabolic resonance factor, the resonance sensitive points of the metabolism of target objects of different ages are analyzed to generate a segmented resonance sensitive map; Analyzing the dynamic correlation between metabolic resonance factors and metabolic network elasticity coefficients to generate metabolic elasticity regulation intervals; According to the metabolic resonance factors and organ metabolic interaction coefficients, the optimal coupling strength of each organ metabolic signal is calculated to generate organ signal coupling correction values; According to the metabolic rhythm stability reflected by the metabolic resonance factors, the rhythm remodeling time-effect parameters are generated; Analyzing the age-segment resonance sensitivity map, metabolic elasticity regulation interval, organ signal coupling correction value and rhythm remodeling time-effect parameter to generate the target direction quantity; Step 5, calculate the target direction quantity to generate the final exercise prescription: Extract the key feature values directly related to exercise intervention from the target direction quantity to generate the exercise intervention core feature values; Calculate the exercise intervention core feature values and the age-segment resonance sensitivity map to analyze the metabolic response differences of the target object to different exercise types and generate the age-segment exercise adaptation coefficient; Based on the age-segment exercise adaptation coefficient, the metabolic elasticity regulation interval and the rhythm remodeling time-effect parameter are fused to calculate the optimal combination range of exercise intensity, duration and implementation period to generate the exercise parameter coordination scheme; According to the exercise parameter coordination scheme, the target direction quantity is analyzed to form the final exercise prescription. 2.The deep learning-based age-segmented personalized obesity prescription generation method of claim 1, wherein, Based on the organ metabolic interaction coefficient, the phase matching degree of the metabolic signals of liver, adipose tissue and muscle in the periodic change is analyzed to generate the organ metabolic phase difference value, including: According to the organ metabolic interaction coefficient, the rhythm characteristics of the metabolic signals of liver, adipose tissue and muscle in a 24-hour cycle are extracted to generate organ metabolic periodic feature points; Based on the organ metabolic periodic feature points, the time difference between the peak time of the main cycle of any two organ metabolic signals is calculated to generate the organ metabolic phase difference value; The organ metabolic phase difference value is analyzed to generate the organ metabolic phase difference value. 3.The deep learning-based age-segmented personalized obesity prescription generation method of claim 1, wherein, Analyze the connection density and regulation path of each node in the real-time metabolic physiological matrix to obtain the metabolic network elasticity coefficient, including: Analyze the connection strength between each node in the real-time metabolic physiological matrix to generate the age-segment node connection weight; Based on the age-segment node connection weight, identify the parallel regulation paths existing in the real-time metabolic physiological matrix to generate the metabolic regulation path redundancy; Based on the metabolic regulation path redundancy, calculate the fluctuation range of the node connection strength to generate the node connection fault tolerance coefficient; Comprehensively calculate the age-segment node connection weight, metabolic regulation path redundancy and node connection fault tolerance coefficient to obtain the metabolic network elasticity coefficient. 4.The deep learning-based age-segmented personalized obesity prescription generation method of claim 1, wherein, Analyze the dynamic correlation between metabolic resonance factors and metabolic network elasticity coefficients to generate metabolic elasticity regulation intervals, including: Construct the dynamic response curve of metabolic resonance factors and metabolic network elasticity coefficients to capture the corresponding fluctuation law of elasticity coefficients when the metabolic resonance state changes to generate resonance-elasticity linkage feature values; Calculate the resonance-elasticity linkage feature values and the metabolic network elasticity coefficients to generate the age-segment elasticity critical value; According to the resonance-elasticity linkage feature values and the age-segment elasticity critical value, calculate the regulation amplitude range of the target object metabolism when the metabolic resonance factor is at different levels to generate the metabolic elasticity regulation interval.

Images