A remote cultivation guidance system based on the growth stage of an infant

Abstract

This invention relates to the field of infant and toddler development guidance technology, and discloses a remote development guidance system based on infant and toddler growth stages. The system includes: identifying abnormal signals in a multi-source development data set to obtain feeding abnormality signals, sleep disturbance signals, motor support signals, environmental disturbance signals, and execution deviation signals; mapping the stage migration identification results to determine the target analysis range corresponding to the current growth stage; and associating and matching the feeding abnormality signals, sleep disturbance signals, motor support signals, environmental disturbance signals, execution deviation signals, and target analysis range to obtain a conflict-related feature set. This invention employs a multi-source development feature joint analysis and stage migration identification technology to achieve the technical effect of dynamically determining the infant's current growth stage and stage migration buffer state, thus matching the guidance content with the actual developmental ability.

Description

Technical Field

[0001] This invention relates to the field of infant and toddler development guidance technology, specifically a remote development guidance system based on infant and toddler growth stages. Background Technology

[0002] In the context of remote home-based parenting guidance for infants and toddlers transitioning from milk feeding to complementary food and experiencing rapid changes in motor skills and circadian rhythms, families continuously upload feeding records, sleep records, motor performance, and care feedback data. However, remote parenting guidance systems need to determine the infant's growth stage under remote conditions and output matching parenting guidance content. Existing technologies are not good at identifying whether an infant or toddler is in a stable growth stage or a stage transition state, and they cannot clarify the influence relationship between multiple sources of behavioral changes, making it difficult to form stage-appropriate and actionable guidance results.

[0003] Currently, existing technologies typically push pre-set feeding suggestions, sleep schedules, and interactive guidance content based on infants' monthly age information or age ranges selected by caregivers. They also analyze basic growth indicators or individual behavior records and provide reference suggestions to caregivers through online consultations or recorded feedback.

[0004] Existing stage identification methods are too static, typically dividing infants into stages based on age in months. They fail to integrate and analyze multiple features, such as feeding refusal, hand-feeding movements, sitting stability, cup-drinking cooperation, circadian rhythm stability, and caregiving responsiveness. This makes it impossible to determine whether an infant has entered the next developmental stage or is transitioning between stages. In practical applications, this can lead to situations where developmental abilities have changed but guidance remains at the previous stage, or where the infant has reached the corresponding age range but their actual abilities do not yet match the guidance content. This results in inconsistencies between the guidance outcomes and the infant's actual developmental status. Therefore, to address these issues, this invention proposes a remote developmental guidance system based on infant developmental stages. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a remote nurturing guidance system based on the growth stages of infants and young children, thereby at least partially solving the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a remote nurturing guidance system based on the growth stages of infants and young children, comprising:

[0007] The data acquisition module obtains multi-source nurturing characteristic data of infants and young children, resulting in a multi-source nurturing data set;

[0008] The stage migration identification module performs joint calculations on the multi-source cultivation data set to obtain the stage migration identification result.

[0009] The conflict correlation analysis module performs correlation analysis on the multi-source breeding data set and the stage migration identification results to obtain coupled conflict data;

[0010] The conflict propagation modeling module constructs a coupled conflict model, performs conflict propagation analysis on the coupled conflict data, and obtains the dominant conflict factor and conflict ranking results.

[0011] The minimum intervention guidance generation module filters and combines the dominant conflict factors and conflict ranking results to obtain the minimum intervention guidance package;

[0012] The remote intervention window matching module performs time window matching on the minimum intervention guidance package to obtain the remote intervention window;

[0013] The review and determination module performs feedback analysis and threshold determination on the remote intervention window to obtain the remote review and determination result.

[0014] Preferably, the multi-source cultivation data set is jointly calculated to obtain the stage migration identification result, including:

[0015] Feature extraction was performed on the multi-source breeding data set to obtain age-matching features, feeding behavior features, motor milestone features, and circadian rhythm features;

[0016] The age-matching features, feeding behavior features, motor milestone features, and circadian rhythm features are weighted and fused to obtain a set of stage adaptation scores.

[0017] The set of adaptation scores for each stage is compared and analyzed to obtain the highest and second-highest adaptation scores.

[0018] Threshold determination is performed based on the difference between the highest and second-highest fit scores to obtain the stage migration recognition result.

[0019] Preferably, the multi-source cultivation data set and the stage migration identification results are correlated and analyzed to obtain coupled conflict data, including:

[0020] Anomaly signals are identified from the multi-source breeding data set to obtain feeding anomaly signals, sleep disturbance signals, motion support signals, environmental disturbance signals, and execution deviation signals.

[0021] The stage migration identification results are used to perform stage mapping to determine the target analysis range corresponding to the current growth stage;

[0022] The feeding abnormality signal, sleep disturbance signal, action support signal, environmental disturbance signal, execution deviation signal, and target analysis range are correlated and matched to obtain a conflict correlation feature set;

[0023] A temporal correlation analysis was performed on the set of conflict-related features to obtain the sequential influence relationship between each conflict-related feature.

[0024] Based on the aforementioned sequential influence relationship, the conflict-related feature set is integrated to obtain coupled conflict data.

[0025] Preferably, a coupled conflict model is constructed, and conflict propagation analysis is performed on the coupled conflict data to obtain the dominant conflict factor and conflict ranking results, including:

[0026] The coupled conflict data is input into the coupled conflict model to determine the propagation relationship between each conflict-related feature.

[0027] Based on the aforementioned propagation correlation, the propagation path analysis is performed on the coupled conflict data to obtain the conflict propagation path corresponding to each conflict correlation feature;

[0028] The propagation intensity of the conflict propagation path is calculated to obtain the conflict impact value corresponding to each conflict association feature;

[0029] By comparing and analyzing the conflict impact values ​​corresponding to each conflict-related feature, the conflict-related feature with the highest conflict impact value is determined as the dominant conflict factor.

[0030] The conflict impact values ​​corresponding to each conflict-related feature are sorted to obtain the conflict ranking result.

[0031] Preferably, the dominant conflict factors and conflict ranking results are screened and combined to obtain a minimal intervention guidance package, including:

[0032] The dominant conflict factor is mapped to a set of candidate guidance actions.

[0033] Based on the conflict sorting results, the candidate guidance action set is prioritized to obtain the target guidance action set;

[0034] Perform action correlation analysis on the target guidance action set, remove duplicate and conflicting actions, and obtain action combination set;

[0035] By imposing a constraint on the number of actions on the set of action combinations, a minimum intervention guidance package is obtained.

[0036] Preferably, the minimal intervention guidance package is matched with a time window to obtain a remote intervention window, including:

[0037] The action type of the minimum intervention guidance package is analyzed to obtain the execution timing characteristics of each guidance action;

[0038] Temporal features were extracted from the multi-source breeding data set to obtain sleep rhythm temporal features, feeding behavior temporal features, and historical execution time features;

[0039] The execution timing characteristics, sleep rhythm time characteristics, feeding behavior time characteristics, and historical execution time characteristics are matched and analyzed to obtain a set of candidate intervention time windows;

[0040] The suitability of the candidate intervention time window set is evaluated to obtain the target time window;

[0041] The target time window is defined as the remote intervention window.

[0042] Preferably, feedback analysis and threshold determination are performed on the remote intervention window to obtain a remote review determination result, including:

[0043] Obtain the guidance execution feedback data corresponding to the remote intervention window to obtain the execution completion status, execution time deviation, and post-execution improvement status;

[0044] Feedback analysis is performed on the execution completion status, execution time deviation, and post-execution improvement to obtain execution stability parameters;

[0045] A preset review and judgment threshold is established, and the execution stability parameter is compared with the review and judgment threshold to obtain the threshold judgment result.

[0046] Based on the threshold determination result, it is determined whether to trigger remote review, and the remote review determination result is obtained.

[0047] Preferably, the propagation intensity of the conflict propagation path is calculated to obtain the conflict impact value corresponding to each conflict association feature, including:

[0048] Obtain the conflict association features corresponding to each path node in the conflict propagation path and the propagation association relationship corresponding to each path edge;

[0049] Based on the propagation association relationship corresponding to each of the path edges, the path propagation weight of each of the conflict propagation paths is determined;

[0050] Based on the positional order of the conflict association features corresponding to each path node in the conflict propagation path, the path attenuation coefficient of each conflict propagation path is determined.

[0051] The path propagation weight and path attenuation coefficient of each conflict propagation path are weighted and calculated to obtain the path propagation intensity corresponding to each conflict propagation path;

[0052] The propagation intensities of each path corresponding to the same conflict association feature are accumulated and integrated to obtain the conflict impact value corresponding to each conflict association feature.

[0053] Preferably, the set of action combinations is constrained in terms of the number of actions to obtain a minimum intervention guidance package, including:

[0054] Preset action quantity constraints;

[0055] The number of guided actions contained in each action combination set is counted to obtain the action quantity value corresponding to each action combination set;

[0056] The number of each action value and the action number constraint range are matched and judged, and the set of target action combinations that meet the action number constraint range is retained.

[0057] The combination priority of the target action combination set is compared to determine the target action combination set with the highest priority;

[0058] The set of highest priority target action combinations is determined as the minimum intervention guidance package.

[0059] Preferably, the suitability evaluation of the candidate intervention time window set is performed to obtain the target time window, including:

[0060] Obtain the sleep rhythm time characteristics, feeding behavior time characteristics, and historical execution time characteristics corresponding to each of the candidate intervention time windows;

[0061] The sleep rhythm time characteristics, feeding behavior time characteristics, and historical execution time characteristics corresponding to each candidate intervention time window are quantified to obtain time feature parameters;

[0062] The time feature parameters are weighted and fused to obtain the time window adaptation value corresponding to each candidate intervention time window;

[0063] The adaptation values ​​of each time window are compared and analyzed to determine the candidate intervention time window with the highest adaptation value as the target time window;

[0064] The target time window is defined as the remote intervention window.

[0065] This invention provides a remote nurturing guidance system based on the developmental stages of infants and young children. It has the following beneficial effects:

[0066] 1. This invention adopts a multi-source cultivation feature joint analysis and stage migration identification technology to achieve the technical effect of dynamically judging the current growth stage and stage migration buffer state of infants and young children, so as to match the guidance content with the actual developmental ability.

[0067] 2. This invention employs a coupling conflict analysis, minimum intervention guidance package generation, and remote verification closed-loop control technology to achieve the technical effect of identifying the coupling relationship between feeding, sleep, movement, environment, and execution deviation, and to realize the executable and verifiable results of remote guidance. Attached Figure Description

[0068] Figure 1 This is a schematic diagram of the system framework of the present invention. Detailed Implementation

[0069] To enable those skilled in the art to understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort should fall within the scope of protection of the present invention.

[0070] The present invention will now be described in detail with reference to the accompanying drawings:

[0071] Please see the appendix Figure 1 This invention provides a remote nurturing guidance system based on infant and toddler growth stages, primarily applicable to home-based remote nurturing guidance scenarios during the simultaneous occurrence of complementary food introduction, motor skill development, sleep rhythm reorganization, and care method adjustments. Unlike existing technologies that use age-based templates for guidance, this invention is based on multi-source nurturing characteristic data of infants and toddlers. Through stage transition identification, coupling conflict analysis, minimal intervention guidance generation, remote intervention window matching, and feedback verification, a complete remote guidance closed loop is formed.

[0072] In an implementable embodiment, the remote training guidance system includes a data acquisition module, a stage migration identification module, a conflict association analysis module, a conflict propagation modeling module, a minimum intervention guidance generation module, a remote intervention window matching module, and a review and judgment module.

[0073] The data acquisition module is used to obtain multi-source nurturing characteristic data of infants and young children, resulting in a multi-source nurturing data set. This module provides a unified input for stage identification and conflict analysis, avoiding the problems of existing technologies that rely on single record items or single consultation information for analysis.

[0074] The preferred multi-source cultivation characteristic data includes the following:

[0075] Basic information is used to provide basic constraints for individuals, including at least date of birth, gender, current age in months, corrected age in months, recent height and weight records, and previous stage labels. For premature infants, corrected age in months is preferred as the basis for stage determination.

[0076] Feeding records are used to reflect an infant's and toddler's feeding adaptation and complementary food progress. They should include at least the feeding time, duration, food texture and type, food acceptance, signs of refusal to eat, self-feeding using hands, cup / drink cooperation, and post-feeding reactions. Feeding records are entered manually by caregivers or captured by video equipment, with some characteristics identified through video recognition.

[0077] Sleep records are used to reflect the stability of an infant's circadian rhythm and include at least the time of falling asleep, the time of waking up, the number of nighttime awakenings, the distribution of daytime naps, the duration of each sleep episode, and the fluctuations in sleep onset over several consecutive days.

[0078] Motion video data is used to identify whether infants and toddlers currently have the motor ability to undertake the next stage of feeding or training tasks. The identification objects include at least sitting stability, grasping movements, hand transfer movements, cup drinking movements, crawling movements, assisted standing movements, and hand-mouth coordination movements.

[0079] Environmental data is used to identify external factors that interfere with an infant’s feeding, sleep or training, including at least noise levels, light conditions, temperature and humidity conditions, stability of the feeding environment, and changes in the care environment.

[0080] The execution feedback data is used to reflect the actual implementation of the system's recommendations in the home setting, including at least whether the recommendations were implemented, the time of completion of implementation, the execution deviation, the actual adjustment actions, subjective difficulty feedback, and post-implementation improvement feedback.

[0081] After collecting the raw data, the remote cultivation guidance system performs the following preprocessing steps in sequence:

[0082] Time is processed uniformly, and feeding records, sleep records, video records, and execution feedback are all converted onto a unified timeline to ensure that time-series correlation analysis can be carried out.

[0083] Outlier handling involves removing, correcting, or marking data as low-confidence if it is clearly illogical or exceeds a reasonable range. For example, if the feeding time is abnormally long, the number of nighttime awakenings is abnormally zero, and other data is missing, the remote breeding guidance system will not directly use it as high-confidence input.

[0084] For missing values, data with missing non-critical fields is filled by the most recent valid record or interpolated based on adjacent time periods of the same day; for data with missing critical fields, missing values ​​are marked.

[0085] Feature normalization is performed to ensure that data from different sources and with different dimensions can be jointly calculated. It is preferable to normalize the feature values ​​using the following formula:

[0086] ,

[0087] in, Represents the original feature values. Represents the normalized eigenvalues. This represents the preset minimum value of the feature. This indicates the preset maximum value for this feature.

[0088] After the above processing, a multi-source cultivation data set is obtained that can be directly called by subsequent modules.

[0089] II. The Stage Transition Identification Module is used to perform joint calculations on multi-source nurturing datasets to obtain stage transition identification results. The core purpose of this module is to move away from using age in months as the sole criterion for judgment, and instead combine age in months, feeding behavior, motor skills, and circadian rhythms to determine whether the infant is currently in a stable stage or a stage transition buffer state.

[0090] The following features were extracted from the multi-source breeding dataset:

[0091] Age-matching features are used to represent the degree of conformity between an infant's current or corrected age and a candidate growth stage.

[0092] Feeding behavior characteristics are used to represent the degree of matching between an infant’s current food acceptance, food refusal behavior, texture adaptation status, and independent feeding status and the candidate stage.

[0093] Motor milestone features are used to indicate the degree of match between an infant's current sitting posture, grasping, transferring, standing with support, or drinking from a cup and the candidate stage.

[0094] Circadian rhythm features are used to represent the degree of matching between the current sleep structure and the stability of the wake window in infants and young children and the candidate stages.

[0095] After the above feature extraction, the system performs joint calculations on each feature to obtain the stage adaptation score for the candidate stage. The following formula is preferred:

[0096] ,

[0097] in, Indicates candidate growth stage The stage adaptation score, Indicates age-matching features. Indicates feeding behavior characteristics, Indicates milestone features of action, It indicates the characteristics of the daily routine. , , , This represents the weight parameters for the corresponding feature.

[0098] During the initial deployment phase of the remote cultivation guidance system, all weight parameters were set using an equal-weighting method; after sample accumulation, the weights were adjusted based on the results of manual annotation.

[0099] The system compares the stage fit scores of each candidate stage to obtain the highest and second-highest fit scores. The difference between these two scores determines whether the infant is currently in a stable stage or a stage transition buffer state. The preferred formula for calculating the difference is as follows:

[0100] ,

[0101] in, This represents the difference between the highest fit score and the second highest fit score. Indicates the highest fit score. This indicates the second-highest fit score.

[0102] The remote nurturing guidance system compares the difference with a preset migration threshold. When the difference does not exceed the migration threshold, the infant is determined to be in a stage migration buffer state; when the difference exceeds the migration threshold, the infant is determined to be in a stable stage. To avoid frequent fluctuations in stage results between adjacent stages, it is preferable to require that this determination be consistently true in numerous consecutive observation sub-windows before outputting the final stage migration identification result.

[0103] Instead of directly applying guidance templates based on age, it dynamically identifies stages based on the combined adaptation results of age or corrected age, feeding behavior, motor milestones, and circadian rhythms, thus solving the problem of overly static stage identification.

[0104] III. The Conflict Correlation Analysis module is used to perform correlation analysis on multi-source breeding datasets and stage migration identification results to obtain coupled conflict data. This module's function is to identify the correlations between feeding, sleep, movement, environment, and executive deviations.

[0105] The remote breeding guidance system first identifies the following abnormal signals from a multi-source breeding dataset:

[0106] Feeding abnormality signals, sleep disturbance signals, movement support signals, environmental disturbance signals, and execution deviation signals.

[0107] Among them, feeding abnormality signals are triggered by increased frequency of food refusal, failure to switch textures, persistently low acceptance, or poor cooperation with cup drinking; sleep disturbance signals are triggered by significant fluctuations in sleep onset time, abnormally increased number of nighttime awakenings, or disordered nap distribution; motor support signals are used to determine whether sitting stability, grasping ability, or hand-mouth coordination ability are sufficient to support the current or next stage of training; environmental disturbance signals are used to determine whether factors such as noise, light, and environmental changes constitute interference; and execution deviation signals are used to reflect situations where recommendations are not implemented on time, in the correct amount, or as required.

[0108] Based on the stage transition identification results, the target analysis scope corresponding to the current growth stage is determined. This target analysis scope is not simply limited to an age range, but rather used to define the conflict relationships that are truly worth focusing on at the current stage. For example, when the remote parenting guidance system identifies that the infant is in the transition buffer state of introducing complementary foods, priority is given to analyzing the relationship between food texture, sitting support, awakening window, and performance deviations.

[0109] The remote training guidance system correlates and matches feeding abnormality signals, sleep disturbance signals, movement support signals, environmental disturbance signals, and execution deviation signals with the target analysis range to obtain a set of conflict-related features. Then, a temporal correlation analysis is performed on this set of conflict-related features to determine the sequential influence relationships between the various conflict-related features.

[0110] In feasible embodiments, by comparing the occurrence sequence and adjacent time distances of different abnormal signals, it is determined whether one abnormal signal typically precedes another. If a feature consistently precedes another feature across numerous consecutive observation periods, and the two meet a preset proximity condition in time, a sequential influence relationship is considered to exist between them. Then, based on this sequential influence relationship, the conflict-related feature set is correlated and integrated to obtain coupled conflict data.

[0111] IV. The conflict propagation modeling module is used to construct coupled conflict models and perform conflict propagation analysis on coupled conflict data to obtain dominant conflict factors and conflict ranking results. The role of this module is to identify the true dominant contradictions.

[0112] A graph structure is preferred for constructing the coupled conflict model. In this model, each conflict-related feature corresponds to a node, and the propagation relationship between nodes corresponds to an edge. Propagation weights can be set on the edges to represent the strength of the influence of the previous feature on the next feature.

[0113] For example, when the system recognizes that "late bedtime" usually precedes "decreased food acceptance the next day" and that this relationship is stable across many observation windows, a propagation relationship is established between the node corresponding to "late bedtime" and the node corresponding to "decreased food acceptance".

[0114] After the coupled conflict model is constructed, propagation path analysis is performed on the coupled conflict data to obtain the conflict propagation path corresponding to each conflict-related feature. The propagation path refers to the process in which a certain conflict feature serves as the starting point, and after being influenced by numerous intermediate conflict features, it ultimately triggers a complex chain of problems. For example, "misalignment of the awakening window" may further affect "texture switching failure" through "decreased food acceptance," and then affect "frequent changes in caregiving actions."

[0115] To determine the conflict propagation path, it is necessary to calculate the propagation intensity to quantify the impact of different conflict characteristics on the overall problem chain. Preferably, the path propagation intensity can be calculated first, and then the conflict impact value can be obtained.

[0116] The path propagation strength can be expressed by the following formula: ,

[0117] in, This represents the path propagation strength corresponding to a given conflict propagation path. This represents the path propagation weight of the path. This represents the path attenuation coefficient of the path.

[0118] Among them, the path propagation weight is used to represent the combined influence strength of the various propagation relationships in the path; the path attenuation coefficient is used to represent the gradual attenuation of the direct influence of the initial conflict characteristics on the terminal problem due to the extension of the propagation chain.

[0119] For a given conflict association feature, the propagation strengths of numerous related paths are accumulated and integrated to obtain the conflict impact value corresponding to that feature. The following formula is preferred:

[0120] ,

[0121] in, Indicates the first The conflict impact value corresponding to the conflict association feature. Indicates the relationship with the first The set of conflict propagation paths related to conflict association features. This represents the path propagation strength corresponding to each conflict propagation path in the path set.

[0122] After obtaining the conflict impact value corresponding to each conflict-related feature, the conflict impact values ​​are compared and analyzed. The conflict-related feature with the highest conflict impact value is determined as the dominant conflict factor, and the remaining conflict-related features are ranked according to the size of their conflict impact values.

[0123] The technical idea embodied in this module is that instead of analyzing feeding, sleep, movement, environment, and execution deviations separately, it uses propagation relationships to identify the true root causes and enhance the system's ability to solve complex coupled problems.

[0124] V. The Minimal Intervention Guidance Generation Module is used to filter and combine the dominant conflict factors and conflict ranking results to obtain a minimal intervention guidance package. The purpose of this module is to output a set of guidance actions that are most worthy of priority and have the lowest execution burden at the current stage, rather than outputting a large number of scattered suggestions at once.

[0125] First, guide actions are mapped based on the dominant conflict factor to obtain a set of candidate guide actions. Guide action mapping refers to mapping the dominant conflict factor to several operational actions that may alleviate the conflict. For example, when the dominant conflict factor is a misaligned eating time, it can be mapped to adjusting the attempt time, optimizing the reminder advance, or reducing unnecessary transitional actions; when the dominant conflict factor is insufficient sitting support, it can be mapped to adjusting the support posture, reducing the intensity of the eating task, or fixing the environmental positioning.

[0126] Based on the conflict ranking results, the candidate guidance action set is prioritized to obtain the target guidance action set. Preferably, the priority considers the mitigation effect of the action on the dominant conflict and the family's executive burden. Then, action correlation analysis is performed on the target guidance action set to remove duplicate and conflicting actions, resulting in an action combination set.

[0127] To prevent excessive output actions from causing execution difficulties, a constraint is imposed on the number of actions in the action combination set. Preferably, a preset range for the number of actions is defined, and the number of guiding actions contained in each action combination is matched and judged, retaining the target action combination set that meets the number constraint. Next, the combination priority of the target action combination set is compared, and the combination with the highest priority is determined, which is then identified as the minimum intervention guidance package.

[0128] VI. The Remote Intervention Window Matching Module is used to match the minimum intervention guidance package with a time window to obtain the remote intervention window. This module does not solve the problem of ordinary message reminders, but rather the problem of when the guidance action should be delivered and executed, thereby improving the completion rate and resolution effect.

[0129] First, the action types in the minimal intervention guidance package are analyzed to obtain the timing characteristics of each guided action. For example, some actions are suitable for execution during the stable phase of the lucid window, some are suitable for execution before the start of feeding, and some are suitable for execution during periods of environmental stability and when the caregiver's attention is sufficient.

[0130] Time-related features are extracted from multi-source breeding datasets, including at least sleep rhythm time features, feeding behavior time features, and historical execution time features. Among these:

[0131] Sleep rhythm temporal characteristics are used to represent the stability of the wakefulness window, the regularity of sleep onset time, and the extent of nighttime awakenings.

[0132] Feeding behavior time characteristics are used to indicate the time periods during which infants and young children have higher acceptance of food and less refusal to eat;

[0133] Historical execution time characteristics are used to indicate the time periods during which past recommendations had a higher success rate and smaller time deviations.

[0134] By matching the timing characteristics of interventions with sleep rhythm time characteristics, feeding behavior time characteristics, and historical intervention time characteristics, a set of candidate intervention time windows is obtained. The remote nurturing guidance system does not directly use fixed time periods, but rather seeks numerous alternative time windows based on the type of action and the infant's actual daily rhythm.

[0135] The suitability of the candidate intervention time window set is evaluated to obtain the target time window. Preferably, the sleep rhythm time characteristics, feeding behavior time characteristics, and historical execution time characteristics corresponding to each candidate intervention time window are first quantified to obtain time feature parameters. Then, a weighted fusion calculation is performed to obtain the time window suitability value corresponding to each candidate intervention time window. The following formula is used: ,

[0136] in, This represents the time window adaptation value corresponding to the candidate intervention time window. This represents the quantification value of the temporal characteristics of sleep rhythm. This represents the quantitative value of the time characteristics of feeding behavior. This represents the quantified value of historical execution time characteristics. , , This represents the weight parameters for the corresponding feature.

[0137] The adaptation values ​​of each time window were compared and analyzed, and the candidate intervention time window with the highest adaptation value was selected as the target time window and determined as the remote intervention window.

[0138] The technical idea embodied in this module is that it does not simply remind you at the appropriate time, but determines the optimal remote intervention window by comprehensively matching sleep rhythm, feeding behavior and historical execution patterns, thereby improving the actual feasibility of the guided actions.

[0139] VII. The review and judgment module is used to perform feedback analysis and threshold determination on the remote intervention window to obtain the remote review and judgment result. The main function of this module is to form an execution closed loop after guidance, avoiding the problem of not following up or reviewing after giving suggestions.

[0140] Obtain the guidance execution feedback data corresponding to the remote intervention window. The execution feedback data includes at least the execution completion status, execution time deviation, and post-execution improvement status. Among them, the execution completion status indicates whether the guided action was fully executed; the execution time deviation indicates the degree of deviation between the actual execution time and the target time window; and the post-execution improvement status indicates whether the problems related to the dominant conflict factor have been alleviated.

[0141] Feedback analysis is performed on execution completion status, execution timing deviations, and post-execution improvements to obtain execution stability parameters. The following formula is preferred:

[0142] ,

[0143] in, Indicates the stability parameter. Indicates the degree of completion of the execution. This represents the quantized value of the deviation at the execution time point. This represents the quantitative value of the improvement trend after implementation. , , This represents the weighting coefficient of the corresponding parameter. The higher the execution completion rate, the larger the parameter value; the smaller the deviation at the execution time point, the larger the parameter value; and the more obvious the improvement trend in execution, the larger the parameter value.

[0144] A preset review threshold is set, and the execution stability parameter is compared with the review threshold. When the execution stability parameter is lower than the review threshold, the automatic guidance is deemed insufficient, and remote review needs to be triggered; when the execution stability parameter is not lower than the review threshold, the current automatic guidance is deemed acceptable, and a remote review result is obtained.

[0145] Example 1:

[0146] This embodiment is used to verify whether the present invention can accurately identify the current growth stage and stage transition buffer state of infants and young children by jointly calculating age-matching features, feeding behavior features, motor milestone features and circadian rhythm features according to the static age division method.

[0147] In this embodiment, a total of infant samples were selected from the stages of complementary food introduction and motor development changes. For example, full-term infants For example, premature infants Example. Continuous collection of all samples. The multi-source data collection included basic information, age in months or corrected age in months, feeding records, sleep records, home activity videos, environmental data, and performance feedback data. Feeding records were created for each infant on average during the observation period. Items, sleep records Group, action video Section, environmental data collection records Item, Execution Feedback Record strip.

[0148] First, the multi-source breeding feature data mentioned above undergoes time unification, outlier labeling, missing value processing, and normalization to obtain a multi-source breeding dataset. Then, age-matching features, feeding behavior features, motor milestone features, and circadian rhythm features are extracted for each sample, and the stage adaptation score corresponding to the candidate stage is calculated according to the stage adaptation score model. For each sample, the highest adaptation score, the second highest adaptation score, and the difference between the two are output, and stage migration is determined based on a preset migration threshold.

[0149] In this embodiment, the manual annotation group consists of... The panel of three experts with experience in assessing infant and toddler development used manual annotation criteria, including changes in the infant's acceptance of food, texture adaptation, sitting stability, hand-mouth coordination, and circadian rhythm stability over a continuous observation period. If at least two of the three annotators agreed, that result was used as the baseline for manual annotation.

[0150] To conduct comparative verification, all samples were assessed using the following two methods at each stage:

[0151] The first method is the multi-source joint computing method of the present invention;

[0152] The second method is a static determination method that directly corresponds to a preset stage template based on the age in months.

[0153] The statistical results are as follows:

[0154] Overall accuracy of stage identification:

[0155] The number of samples whose identification results using the method of this invention are consistent with those of manual annotation is For example, the overall accuracy rate is The number of samples whose identification results based on the static age-based classification method are consistent with the manually labeled results is: For example, the overall accuracy rate is .

[0156] Stage transition buffer state identification accuracy:

[0157] exist In the example sample, 46 samples were manually labeled as being in the stage transition buffer state. Of these, the samples correctly identified as being in the stage transition buffer state by the method of this invention are: For example, the recognition accuracy rate is The samples correctly identified by the static age-based classification method are: For example, the recognition accuracy rate is .

[0158] Phase handover error scenarios:

[0159] exist Within a continuous observation period, this study examines unreasonable fluctuations in the results between adjacent stages. Specifically, the sample exhibiting stage switching errors in this invention's method is: For example, the false handover rate is The samples that exhibited incorrect phase switching when using the static age-based segmentation method are: For example, the false handover rate is .

[0160] Identification of samples from premature infants:

[0161] In 18 premature infants: the accuracy rate of identification using the corrected age combined with feeding, motor and daily routine characteristics was [percentage missing]. The accuracy rate of identification based on the static classification method according to actual age is [missing information]. .

[0162] The results above show that the present invention obtains the stage migration identification results by jointly calculating the multi-source cultivation features, which can significantly improve the stage identification accuracy, especially the ability to identify the stage migration buffer state, and reduce the occurrence rate of stage mis-switching.

[0163] Example 2:

[0164] This embodiment is used to verify whether the present invention can accurately locate the dominant conflict factor and generate a minimal intervention guidance package based on the identification of the coupling relationship between feeding, sleep, movement, environment and executive deviation.

[0165] In this embodiment, a total of infant samples exhibiting complex problem manifestations were selected. Example. Complex problems include, but are not limited to, the simultaneous presence of at least two of the following: decreased food acceptance, significant food refusal, increased sleep disturbances, insufficient postural stability, frequent environmental disturbances, and significant deviation from recommended implementation. Data was collected continuously from each sample. The system collects multi-source cultivation characteristic data and performs stage migration identification and conflict analysis after the data is collected.

[0166] Will The sample was divided into an experimental group and a control group, and each group had... example.

[0167] Among them: the experimental group adopted the method of the present invention, that is, firstly, abnormal signal identification was performed, then stage mapping, correlation matching, temporal correlation analysis and coupling conflict propagation analysis were performed to obtain the dominant conflict factor and conflict ranking results, and finally the minimum intervention guidance package was generated.

[0168] The control group adopted a separate suggestion approach, that is, suggestions were given for feeding problems, sleep problems and motor problems respectively, without coupling conflict propagation analysis or motor quantity constraints.

[0169] In the experimental group, the distribution of the high-frequency dominant conflict factors is as follows:

[0170] Awake window misalignment: example;

[0171] The texture of the food does not match the current movement support: example;

[0172] Insufficient postural support: example;

[0173] Unstable feeding environment: example;

[0174] Execution bias leads to distorted observations: example.

[0175] Guided actions are mapped based on the dominant conflict factor, and priority filtering and action quantity constraints are performed by combining conflict ranking results. In the minimum intervention guidance package generated by the experimental group, the number of output guided actions for each sample is [number missing]. Xiang Zhi Item, average The recommended number of items for the control group is [number]. Xiang Zhi Item, average item.

[0176] exist During the 14-day guidance implementation period, the implementation status and improvement of the experimental group and the control group were statistically analyzed, and the results are as follows:

[0177] Average number of guided movements in the experimental group: item;

[0178] Average number of guided movements in the control group: item.

[0179] The completion rate of guidance is calculated as "the number of guidance actions actually completed / the number of guidance actions that should be completed":

[0180] The average completion rate of the experimental group was ;

[0181] The average completion rate of the control group was .

[0182] A five-level difficulty rating system is used, with higher scores indicating greater difficulty in execution.

[0183] The average subjective difficulty score of the experimental group was point;

[0184] The average subjective difficulty score for the control group was point.

[0185] The improvement rate of the dominant problem is calculated as "the percentage of samples where the degree of abnormality corresponding to the dominant conflict factor decreased after the execution cycle ends":

[0186] The improvement rate of the main problem in the experimental group was ;

[0187] The improvement rate of the main problem in the control group was .

[0188] The fragmentation rate is defined as the percentage of samples where "suggestions are duplicated, conflicting, or lack prioritization":

[0189] The fragmentation rate of the experimental group was ;

[0190] The fragmentation rate of the control group was .

[0191] A typical sample from the experimental group was selected. The following problem chain was identified in the TianData database:

[0192] "Continuously delayed sleep onset time → decreased morning receptivity the following day → failure to switch feeding methods → frequent changes in feeding methods by caregivers." Through coupled conflict propagation analysis, "misaligned wakefulness window" was identified as the dominant conflict factor, and a guidance package was output. The measures include maintaining a stable morning wakefulness window, consistent times for introducing complementary foods, and observing the effects of a single texture. (Execution) After that, the sample's food acceptance rate increased from Upgraded to The frequency of food refusal decreased from an average of one day The next drop to Second-rate.

[0193] This invention obtains the dominant conflict factors and conflict ranking results through coupled conflict propagation analysis, and generates a minimum intervention guidance package by screening and combining them. This can significantly reduce the number of suggestions, improve the completion rate, reduce the difficulty of implementation, and improve the improvement effect of the main problems.

[0194] Example 3:

[0195] This embodiment is used to verify whether the present invention can improve the execution stability of remote guidance actions by matching the minimum intervention guidance package with a time window, outputting a remote intervention window, and performing feedback analysis and threshold determination on the remote intervention window, and can promptly trigger remote review when the automatic guidance effect is insufficient, thus forming an effective closed loop.

[0196] In this embodiment, a total of infant samples that have completed the identification of dominant conflict factors and generated minimal intervention guidance packages were selected. example. Will The sample was randomly divided into an experimental group and a control group, with 40 cases in each group. Among them:

[0197] The experimental group used the method of the present invention, which is to perform time window matching based on the action type analysis of the minimum intervention guidance package, sleep rhythm time characteristics, feeding behavior time characteristics and historical execution time characteristics to obtain the remote intervention window, and to perform feedback analysis and threshold determination after execution.

[0198] The control group uses a fixed-time push method, that is, pushes guidance content at the same fixed time, without time window adaptation, execution stability calculation and review trigger analysis.

[0199] In the experimental group, a set of candidate intervention time windows was first generated for each sample. The average of all samples formed the candidate time windows, and the target time window was selected based on the time window fit value. The average distribution of the target time windows is as follows:

[0200] Morning stable wakefulness window percentage ;

[0201] The percentage of the first complementary food window in the morning ;

[0202] Recovery window percentage after afternoon nap ;

[0203] The percentage of fixed care windows in the evening .

[0204] In continuous Within the daily guidance period, record the execution completion rate, execution time deviation, continuous execution stability, and review triggering status. The statistical results are as follows:

[0205] The average completion rate of the experimental group was ;

[0206] The average completion rate of the control group was .

[0207] Execution time deviation is expressed as the average deviation in minutes between the actual execution time and the center time of the target time period:

[0208] The average execution time deviation of the experimental group was minute;

[0209] The average execution time deviation of the control group was minute.

[0210] Continuous execution stability rate is calculated based on "continuous" Statistics on the percentage of samples that completed the instructed actions as required more than once:

[0211] The continuous execution stability rate of the experimental group was ;

[0212] The stability rate of continuous execution in the control group was .

[0213] In the experimental group In the example sample, execution stability parameters are calculated based on execution completion status, execution time deviation, and post-execution improvement, and compared with a preset review threshold. Ultimately:

[0214] The sample that triggered remote review was For example, percentage ;

[0215] The samples that did not trigger remote review are For example, percentage .

[0216] For triggering remote review For example, after further statistical analysis of the sample and verification, the stage identification results, dominant conflict factors, or remote intervention windows were readjusted. Examples will follow. The completion rate within one day increased by more than On average, from Upgraded to The average deviation at the execution time point is from Minutes down to minute.

[0217] A typical sample from the experimental group was selected to generate candidate intervention time windows, with the target time window being "before the first stable awakening window in the morning". Minute reminder, after waking up Perform the instructed actions within minutes.

[0218] Before execution The completion rate of this sample under fixed time period push was [percentage missing]. ;

[0219] After switching to the time window matching method of this invention, the subsequent The daily completion rate has increased to The execution time deviation is due to the average Minutes down to Minutes, corresponding to a decrease in the degree of anomaly related to the dominant conflict factor. .

[0220] This invention significantly improves the execution completion rate, reduces execution timing deviation, and enhances continuous execution stability by matching the minimum intervention guidance package to a time window and outputting a remote intervention window. At the same time, by performing feedback analysis and threshold determination on the remote intervention window, it can promptly identify situations where automatic guidance is insufficient and trigger remote review, forming an effective closed loop.

[0221] Example 1 verifies the technical effectiveness of the present invention in terms of the accuracy of stage migration identification.

[0222] Example 2 verifies the technical effectiveness of the present invention in coupling conflict resolution, dominant conflict localization, and minimal intervention guidance generation.

[0223] Example 3 verifies the technical effectiveness of the present invention in remote intervention window matching, execution feedback analysis, and remote verification closed-loop control.

[0224] Example 1 shows that by jointly calculating age-matching features, feeding behavior features, motor milestone features, and circadian rhythm features, the present invention can accurately identify the current growth stage and stage transition buffer state of infants and young children according to the static age division method, and maintain high stability during stage switching. This indicates that the present invention has good accuracy and reliability in stage transition identification.

[0225] Example 2 shows that by performing coupled conflict propagation analysis on feeding, sleep, movement, environment, and executive deviation, the present invention can effectively identify the dominant conflict factors and conflict ranking results. Furthermore, based on this, it generates a minimal intervention guidance package with fewer movements, lower executive burden, and better improvement effect, demonstrating that the present invention has strong pertinence and practicality in identifying dominant problems and generating guidance strategies.

[0226] Example 3 shows that by matching the minimum intervention guidance package with a time window to determine the remote intervention window, and combining execution feedback analysis and threshold determination to trigger remote review, the present invention can significantly improve the completion rate of guided actions, reduce execution time deviation, and enhance the stability of continuous execution. Moreover, it can promptly correct when the automatic guidance effect is insufficient, indicating that the present invention has good stability and continuous optimization capability in remote guidance execution closed-loop control.

[0227] Embodiments of the present invention have been presented and described. It will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A remote nurturing guidance system based on the growth stages of infants and young children, characterized in that, include: The data acquisition module obtains multi-source nurturing characteristic data of infants and young children, resulting in a multi-source nurturing data set; The stage migration identification module performs joint calculations on the multi-source cultivation data set to obtain the stage migration identification result. The conflict correlation analysis module performs correlation analysis on the multi-source breeding data set and the stage migration identification results to obtain coupled conflict data; The conflict propagation modeling module constructs a coupled conflict model, performs conflict propagation analysis on the coupled conflict data, and obtains the dominant conflict factor and conflict ranking results. The minimum intervention guidance generation module filters and combines the dominant conflict factors and conflict ranking results to obtain the minimum intervention guidance package; The remote intervention window matching module performs time window matching on the minimum intervention guidance package to obtain the remote intervention window; The review and determination module performs feedback analysis and threshold determination on the remote intervention window to obtain the remote review and determination result.

2. The remote cultivation guidance system according to claim 1, characterized in that, The multi-source cultivation dataset is jointly calculated to obtain the stage migration identification results, including: Feature extraction was performed on the multi-source breeding data set to obtain age-matching features, feeding behavior features, motor milestone features, and circadian rhythm features; The age-matching features, feeding behavior features, motor milestone features, and circadian rhythm features are weighted and fused to obtain a set of stage adaptation scores. The set of adaptation scores for each stage is compared and analyzed to obtain the highest and second-highest adaptation scores. Threshold determination is performed based on the difference between the highest and second-highest fit scores to obtain the stage migration recognition result.

3. The remote cultivation guidance system according to claim 1, characterized in that, The multi-source cultivation data set and the stage migration identification results are correlated and analyzed to obtain coupled conflict data, including: Anomaly signals are identified from the multi-source breeding data set to obtain feeding anomaly signals, sleep disturbance signals, motion support signals, environmental disturbance signals, and execution deviation signals. The stage migration identification results are used to perform stage mapping to determine the target analysis range corresponding to the current growth stage; The feeding abnormality signal, sleep disturbance signal, action support signal, environmental disturbance signal, execution deviation signal, and target analysis range are correlated and matched to obtain a conflict correlation feature set; A temporal correlation analysis was performed on the set of conflict-related features to obtain the sequential influence relationship between each conflict-related feature. Based on the aforementioned sequential influence relationship, the conflict-related feature set is integrated to obtain coupled conflict data.

4. The remote cultivation guidance system according to claim 1, characterized in that, A coupled conflict model is constructed, and conflict propagation analysis is performed on the coupled conflict data to obtain the dominant conflict factors and conflict ranking results, including: The coupled conflict data is input into the coupled conflict model to determine the propagation relationship between each conflict-related feature. Based on the aforementioned propagation correlation, the propagation path analysis is performed on the coupled conflict data to obtain the conflict propagation path corresponding to each conflict correlation feature; The propagation intensity of the conflict propagation path is calculated to obtain the conflict impact value corresponding to each conflict association feature; By comparing and analyzing the conflict impact values ​​corresponding to each conflict-related feature, the conflict-related feature with the highest conflict impact value is determined as the dominant conflict factor. The conflict impact values ​​corresponding to each conflict-related feature are sorted to obtain the conflict ranking result.

5. The remote cultivation guidance system according to claim 1, characterized in that, The dominant conflict factors and conflict ranking results are screened and combined to obtain a minimal intervention guidance package, including: The dominant conflict factor is mapped to a set of candidate guidance actions. Based on the conflict sorting results, the candidate guidance action set is prioritized to obtain the target guidance action set; Perform action correlation analysis on the target guidance action set, remove duplicate and conflicting actions, and obtain action combination set; By imposing a constraint on the number of actions on the set of action combinations, a minimum intervention guidance package is obtained.

6. The remote cultivation guidance system according to claim 1, characterized in that, The minimal intervention guidance package is matched with a time window to obtain a remote intervention window, including: The action type of the minimum intervention guidance package is analyzed to obtain the execution timing characteristics of each guidance action; Temporal features were extracted from the multi-source breeding data set to obtain sleep rhythm temporal features, feeding behavior temporal features, and historical execution time features; The execution timing characteristics, sleep rhythm time characteristics, feeding behavior time characteristics, and historical execution time characteristics are matched and analyzed to obtain a set of candidate intervention time windows; The suitability of the candidate intervention time window set is evaluated to obtain the target time window; The target time window is defined as the remote intervention window.

7. The remote cultivation guidance system according to claim 1, characterized in that, Feedback analysis and threshold determination are performed on the remote intervention window to obtain the remote review determination result, including: Obtain the guidance execution feedback data corresponding to the remote intervention window to obtain the execution completion status, execution time deviation, and post-execution improvement status; Feedback analysis is performed on the execution completion status, execution time deviation, and post-execution improvement to obtain execution stability parameters; A preset review and judgment threshold is established, and the execution stability parameter is compared with the review and judgment threshold to obtain the threshold judgment result. Based on the threshold determination result, it is determined whether to trigger remote review, and the remote review determination result is obtained.

8. The remote cultivation guidance system according to claim 4, characterized in that, The propagation intensity of the conflict propagation path is calculated to obtain the conflict impact value corresponding to each conflict association feature, including: Obtain the conflict association features corresponding to each path node in the conflict propagation path and the propagation association relationship corresponding to each path edge; Based on the propagation association relationship corresponding to each of the path edges, the path propagation weight of each of the conflict propagation paths is determined; Based on the positional order of the conflict association features corresponding to each path node in the conflict propagation path, the path attenuation coefficient of each conflict propagation path is determined. The path propagation weight and path attenuation coefficient of each conflict propagation path are weighted and calculated to obtain the path propagation intensity corresponding to each conflict propagation path; The propagation intensities of each path corresponding to the same conflict association feature are accumulated and integrated to obtain the conflict impact value corresponding to each conflict association feature.

9. The remote cultivation guidance system according to claim 5, characterized in that, By applying a constraint on the number of actions to the set of action combinations, a minimum intervention guidance package is obtained, including: Preset action quantity constraints; The number of guided actions contained in each action combination set is counted to obtain the action quantity value corresponding to each action combination set; The number of each action value and the action number constraint range are matched and judged, and the set of target action combinations that meet the action number constraint range is retained. The combination priority of the target action combination set is compared to determine the target action combination set with the highest priority; The set of highest priority target action combinations is determined as the minimum intervention guidance package.

10. The remote cultivation guidance system according to claim 6, characterized in that, The suitability of the candidate intervention time window set is evaluated to obtain the target time window, including: Obtain the sleep rhythm time characteristics, feeding behavior time characteristics, and historical execution time characteristics corresponding to each of the candidate intervention time windows; The sleep rhythm time characteristics, feeding behavior time characteristics, and historical execution time characteristics corresponding to each candidate intervention time window are quantified to obtain time feature parameters; The time feature parameters are weighted and fused to obtain the time window adaptation value corresponding to each candidate intervention time window; The adaptation values ​​of each time window are compared and analyzed to determine the candidate intervention time window with the highest adaptation value as the target time window; The target time window is defined as the remote intervention window.

Images