Real-time monitoring system for child stress in conjunction with behavioral data analysis

Abstract

The present application relates to the technical field of children stress monitoring, and particularly relates to a children stress real-time monitoring system combined with behavior data analysis. The system is used for: collecting motion data and physiological data corresponding to target children and candidate children, and extracting motion state vectors and physiological state vectors, and performing first clustering processing on the motion state vectors and the physiological state vectors respectively; performing second clustering processing on motion condition sequences and physiological condition sequences respectively to obtain general motion types and general physiological types; determining representative physiological parameters corresponding to each general motion type; determining reference children according to the representative physiological parameters; training a stress prediction model; collecting motion data and physiological data of target children, inputting the motion data and the physiological data into the trained stress prediction model, and outputting stress levels of the target children. The present application can improve the accuracy of monitoring stress levels of children.

Description

Technical Field

[0001] This invention relates to the technical field of child stress monitoring, and specifically to a real-time child stress monitoring system that combines behavioral data analysis. Background Technology

[0002] Childhood is a critical period for an individual's mental health development. However, increasing academic competition, complex family environments, and social pressures have led to unprecedented psychological challenges for children. Prolonged stress can not only damage children's cognitive and learning abilities but also trigger serious mental health problems such as anxiety and depression. Therefore, early identification and timely intervention of children's stress has become an important issue in the public and educational fields.

[0003] With the development of information technology, especially the maturity of wearable devices, IoT sensors and artificial intelligence, it is possible to monitor children's behavioral data and physiological conditions in real time. By combining these physiological and behavioral data with daily behavior patterns and past stress levels, children's stress levels can be detected. However, in reality, because children's minds are not fully mature, their behavioral patterns do not have obvious trend models, making it difficult to assess and monitor their stress performance based on a single child's behavioral pattern.

[0004] In other words, current technology has low accuracy in monitoring children's stress levels. Summary of the Invention

[0005] The purpose of this invention is to provide a real-time stress monitoring system for children that combines behavioral data analysis, in order to solve the technical problem of low accuracy in monitoring children's stress levels in the prior art.

[0006] In a first aspect, one embodiment of the present invention provides a real-time stress monitoring system for children that incorporates behavioral data analysis, the system comprising: The data acquisition module is used to collect the movement and physiological data of the target children and candidate children, and to build a reference database based on the movement data, physiological data and historical stress level data; The data preprocessing module extracts motion state vectors and physiological state vectors from the reference database, and performs a first clustering process on the motion state vectors and physiological state vectors to obtain motion state type, physiological state type, motion type descriptor, and physiological type descriptor. A second clustering process is then performed on the motion sequence composed of motion type descriptors and the physiological sequence composed of physiological type descriptors to obtain a general motion type and a general physiological type. Based on the general motion type, representative physiological parameters corresponding to each general motion type are determined. The reference child screening module is used to determine reference children based on representative physiological parameters; The model training module is used to train the model based on the corresponding motion sequence, physiological sequence and historical stress level data of the target child and the reference child, combined with training weights, to obtain a trained stress prediction model. The monitoring module is used to collect the target child's movement and physiological data within a preset time period, input them into the trained stress prediction model, and output the target child's stress level.

[0007] In some embodiments, the motion state vector and the physiological state vector are subjected to a first clustering process to obtain motion state type, physiological state type, motion type descriptor, and physiological type descriptor, including: The motion state vector and physiological state vector corresponding to each child are subjected to the first clustering process to obtain at least one motion state type and at least one physiological state type corresponding to each child. For any motion state type, the motion category descriptor corresponding to the motion state type is determined based on the statistical characteristics of all motion state vectors in the motion state type. For any physiological state type, the physiological category descriptor corresponding to the physiological state type is determined based on the statistical characteristics of all physiological state vectors in the physiological state type.

[0008] In some embodiments, before performing a second clustering process on the motion sequence and the physiological sequence, the method further includes: The accuracy of motion description is determined by the cosine similarity between the motion state vector at any given time and the motion type descriptor corresponding to the motion state type at that time. The accuracy of physiological description is determined by the cosine similarity between the physiological state vector at any given time and the physiological category descriptor corresponding to the physiological state type at that time. The accuracy of time description is determined by multiplying the accuracy of motion description and the accuracy of physiological description at any given time. The overall description accuracy is obtained by averaging the description accuracy over all moments within a preset time period.

[0009] In some embodiments, before performing a second clustering process on the sequence of motion conditions composed of motion type descriptors and the sequence of physiological conditions composed of physiological type descriptors to obtain a general motion type and a general physiological type, the method further includes: For any child, the movement type descriptors and physiological type descriptors corresponding to each moment within the preset time period are arranged in chronological order to determine the child's corresponding movement status sequence and physiological status sequence; For any two children, determine the similarity between their motion state vectors and the similarity between their physiological state vectors at the same time. The motion difference is determined based on the similarity of motion vectors and the accuracy of motion description, and the physiological difference is determined based on the similarity of physiological vectors and the accuracy of physiological description.

[0010] In some embodiments, a second clustering process is performed on the sequence of motion conditions composed of motion type descriptors and the sequence of physiological conditions composed of physiological type descriptors to obtain a general motion type and a general physiological type, including: Using motion difference and physiological difference as distance inputs, temporal matching is performed on motion sequence and physiological sequence to obtain motion difference distance and physiological difference distance. Using the distance between movement differences and the distance between physiological differences as distance metrics, a second clustering process is performed on the movement sequence and the physiological sequence to obtain the general movement type and the general physiological type.

[0011] In some embodiments, representative physiological parameters corresponding to each general exercise type are determined based on the general exercise type, including: For any given general movement type, obtain the physiological sequence of each child under that general movement type; Based on the physiological condition sequence, representative physiological parameters corresponding to each general movement type are determined.

[0012] In some embodiments, determining a reference child based on representative physiological parameters includes: The developmental consistency between target children and candidate children is determined based on representative physiological parameters and the total number of common motor types. Reference children are determined based on developmental concordance; developmental concordance is used to characterize the similarity of representative physiological parameters between target children and candidate children in the same general movement type.

[0013] In some embodiments, determining a reference child based on developmental congruence includes: If the developmental consistency is greater than or equal to the preset consistency threshold, the candidate child will be identified as the reference child. If the developmental consistency is less than the preset consistency threshold, the candidate child will be identified as a non-reference child.

[0014] In some embodiments, a well-trained stress prediction model is obtained by training a model based on the motion sequence, physiological sequence, and historical stress level data corresponding to the target child and the reference child, combined with training weights. This model includes: Extract limb movement vectors from the reference database, and determine the importance of each limb movement based on the limb movement vectors; The training reference weights are determined based on developmental consistency and the importance of the movement; The training weights are determined based on the training reference weights and the overall descriptive accuracy. The stress prediction model is trained by using the motion sequence, physiological sequence, and historical stress level data of the target child and the reference child as training samples, adjusting the contribution of each sample through training weights, and obtaining a well-trained stress prediction model.

[0015] Secondly, another embodiment of the present invention provides a method for real-time monitoring of children's stress by combining behavioral data analysis, the method comprising: Collect motor and physiological data of target and candidate children, and build a reference database based on motor data, physiological data and historical stress level data; Motion state vectors and physiological state vectors are extracted from the reference database. A first clustering process is then performed on both vectors to obtain motion state type, physiological state type, motion type descriptor, and physiological type descriptor. A second clustering process is then performed on the motion sequence composed of motion type descriptors and the physiological sequence composed of physiological type descriptors to obtain general motion type and general physiological type. Based on the general motion type, representative physiological parameters corresponding to each general motion type are determined. The reference child is determined based on representative physiological parameters; Based on the movement sequence, physiological sequence and historical stress level data of the target child and the reference child, the model is trained by combining the training weights to obtain a well-trained stress prediction model. Collect the target child's movement and physiological data within a preset time period, input them into a trained stress prediction model, and output the target child's stress level.

[0016] Thirdly, in another embodiment of the present invention, an electronic device is provided, including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method described in the second aspect above.

[0017] Fourthly, in another embodiment of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the method described in the second aspect above.

[0018] The present invention has the following beneficial effects: This invention constructs a reference database by collecting movement and physiological data from target and candidate children and combining it with historical stress levels. The extracted movement and physiological state vectors are then subjected to a first clustering process to obtain a unique movement / physiological state type and corresponding descriptor for each child. Subsequently, the descriptors are arranged chronologically to form a sequence and subjected to a second clustering process to classify common movement / physiological types across children. Representative physiological conditions corresponding to each common movement type are then selected to eliminate interference from children's daily movement on physiological indicators and avoid misjudgments of stress due to fluctuations in exercise volume. Simultaneously, developmental consistency is calculated based on representative physiological conditions to select reference children with similar developmental levels to the target children. Training weights are calculated by combining developmental consistency, movement importance, and data accuracy. A stress prediction model is trained using the movement and physiological state sequences of the target and reference children, along with historical stress level data. This ensures that the training samples better match the physiological and behavioral characteristics of the target children, significantly improving the model's prediction accuracy. Finally, by collecting data from the target children within a preset time period in real time and inputting it into the model, the accuracy of monitoring children's stress levels is improved. Attached Figure Description

[0019] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of a real-time child stress monitoring system that combines behavioral data analysis, provided by an embodiment of the present invention. Figure 2 This is a flowchart illustrating a method for real-time monitoring of children's stress that combines behavioral data analysis, provided by an embodiment of the present invention. Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0021] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the real-time child stress monitoring system based on behavioral data analysis proposed in this invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0023] The specific solution of the real-time child stress monitoring system combining behavioral data analysis provided by the present invention will be described in detail below with reference to the accompanying drawings.

[0024] In one embodiment, the present invention provides a real-time child stress monitoring system 100 that incorporates behavioral data analysis, such as... Figure 1 As shown, the system 100 includes: The data acquisition module 101 is used to collect the movement data and physiological data of the target children and candidate children, and to build a reference database based on the movement data, physiological data and historical stress level data.

[0025] Among them, the target children refer to specific children who need to be monitored for stress in real time. They are the ultimate target of the monitoring system, and their stress levels are output in real time through the trained model.

[0026] Candidate children refer to a large number of other children, besides the target children, who participate in data collection and model training. Their data are used to screen reference children whose development is consistent with that of the target children, thereby expanding the model training sample.

[0027] It should be noted that the stress monitoring work conducted on the target children and candidate children in this invention is carried out with the permission of the children's guardians.

[0028] Motion data refers to data on limb movements and motion states collected by accelerometers and gyroscopes integrated into wristbands worn on children's limbs, with a sampling frequency of 1 time per second.

[0029] Physiological data refers to data related to the child's physical physiological state collected by the blood oxygen and heart rate sensors integrated in the wristband, with a sampling frequency of 1 time per second.

[0030] Historical stress level data refers to the stress level records obtained through comprehensive evaluation at different times in the past for both target and candidate children. It is the core output label data during the model training phase.

[0031] The reference database refers to a multi-dimensional database formed by integrating the movement data, physiological data, and historical stress level data of target children and candidate children, which provides data support for subsequent data preprocessing, reference child selection, and model training.

[0032] It should be noted that, in the embodiments of the present invention, the data acquisition module can simultaneously collect real-time motion and physiological data (1 time / second sampling) of the target child and the candidate child through a specific wristband device, and integrate the historical stress level data of both after comprehensive evaluation, and finally construct a multi-dimensional reference database to provide basic data for the subsequent construction of the stress monitoring model and real-time monitoring of the system.

[0033] The data preprocessing module 102 is used to extract motion state vectors and physiological state vectors from the reference database, and perform a first clustering process on the motion state vectors and physiological state vectors to obtain motion state type, physiological state type, motion type descriptor, and physiological type descriptor; perform a second clustering process on the motion situation sequence composed of motion type descriptors and the physiological situation sequence composed of physiological type descriptors to obtain general motion type and general physiological type; and determine the representative physiological parameters corresponding to each general motion type based on the general motion type.

[0034] Among them, the motion state vector ( This refers to a vector that represents a child's overall motion state at a certain moment, constructed using various data collected by motion sensors (accelerometer, gyroscope) as dimensions (e.g., one vector per second, reflecting information such as the amplitude and speed of limb movement at that moment).

[0035] Physiological state vector ( This refers to a vector that represents a child's physiological state at a certain moment, constructed using various data collected by physiological sensors (blood oxygen, heart rate) as dimensions (e.g., one vector per second, reflecting the body's functional indicators at that moment).

[0036] Limb motion vectors ( ( ) refers to the vector constructed for each limb part (such as left hand, right foot) of each child, used to represent the movement characteristics of a single limb part, which is different from the motion state vector that reflects the overall motion state.

[0037] The first clustering process refers to the clustering operation performed on the motion state vector and physiological state vector of a single child, using the density-based spatial clustering of applications with noise (DBSCAN) algorithm (minimum number of points). =5, radius pass (Distance graph determined), the distance metric is 1 - normalized cosine similarity, used to classify the individual state type of a child.

[0038] Movement state type refers to the specific categories (such as sitting, walking, jumping, etc.) that a child's overall movement state is divided into after the first clustering. Each category represents a typical movement pattern.

[0039] Physiological state type refers to the specific categories (such as resting, mild activity, stress, etc.) into which the physiological state of an individual child is divided after the first clustering. Each category represents a typical physiological response pattern.

[0040] For example, the motion type descriptor refers to the mean vector of all motion state vectors in the cluster corresponding to each motion state type, which is used to characterize the core features of the motion state and realize a standardized description of the motion.

[0041] For example, the physiological category descriptor refers to the mean vector of all physiological state vectors in the cluster corresponding to each physiological state type, which is used to characterize the core features of the physiological state and realize a standardized description of the physiological response of that type.

[0042] A movement sequence refers to a sequence formed by arranging the movement type descriptors corresponding to each second of a single child within a certain time period (e.g., 5 minutes) in chronological order (e.g., 300 descriptors in 5 minutes), which is used to characterize the changes in the child's movement patterns within that time period.

[0043] Physiological condition sequence refers to a sequence formed by arranging the physiological type descriptors corresponding to each second of a single child within a preset time period (e.g., 5 minutes) in chronological order (e.g., 300 descriptors in 5 minutes), used to characterize the changes in the child's physiological response within that time period.

[0044] For example, the second clustering process refers to the clustering operation performed on the motion sequence and physiological sequence of all children (target children + candidate children), using the DBSCAN algorithm and the distance metric is the Dynamic Time Warping (DTW) difference distance, which is used to classify the common state types across children.

[0045] General movement types refer to several common categories (such as general sitting, general walking, etc.) that are divided into for all children's movement sequences after the second clustering, so as to unify the movement patterns of different children.

[0046] The general physiological type refers to several common categories (such as general resting, general stress, etc.) that are divided into for all children’s physiological condition sequences after the second clustering, so as to unify the physiological response patterns of different children.

[0047] For example, the representative physiological parameter refers to the physiological sequence that best represents the physiological response of a general type of exercise (obtained by filtering through DTW distance mean + negative correlation normalization).

[0048] It should be noted that, in the embodiments of the present invention, motion state vectors, physiological state vectors, and limb movement vectors are extracted from the reference database. First, the motion / physiological state vectors of a single child are clustered for the first time using the DBSCAN algorithm to obtain the child's unique motion / physiological state type and corresponding descriptor. Then, the descriptors are arranged chronologically according to a preset time period to construct a sequence of motion / physiological conditions for all children. A second clustering is performed using the DBSCAN algorithm to obtain a universal motion / physiological type for all children. Finally, based on the universal motion type, representative physiological condition sequences corresponding to each universal motion type are selected to provide data support for subsequent selection of reference children and model training.

[0049] Furthermore, the motion state vector and physiological state vector are subjected to a first clustering process to obtain motion state type, physiological state type, motion type descriptor, and physiological type descriptor, including: The motion state vector and physiological state vector corresponding to each child are subjected to the first clustering process to obtain at least one motion state type and at least one physiological state type corresponding to each child.

[0050] The term "each child" refers to both the target children who need stress monitoring and the candidate children who participate in data collection and are used to screen reference children; these are the subjects of the first clustering process.

[0051] The first clustering process refers to the independent clustering operation performed on the motion state vector and physiological state vector of a single child, using the DBSCAN algorithm, with the distance metric being 1 - normalized cosine similarity, and the clustering parameter being the minimum number of points (…). =5), radius ( pass (Determined by distance map method), used to classify individual children's specific motor and physiological states into categories.

[0052] It should be noted that, in the embodiments of the present invention, for each individual among the target child and candidate children, the motion state vector and physiological state vector collected and constructed in real time are independently clustered using the DBSCAN clustering algorithm (specifying distance metrics and parameters), and finally each child is classified into at least one exclusive motion state type and at least one exclusive physiological state type, thereby realizing the structured classification of the motion and physiological characteristics of a single child.

[0053] For any motion state type, the motion category descriptor corresponding to the motion state type is determined based on the statistical characteristics of all motion state vectors in the motion state type.

[0054] Among them, any type of movement state refers to a specific category (such as sitting, walking, jumping, etc.) that a single child's overall movement state is divided into after the first clustering process. Each category represents a typical movement pattern of the child and is the product of movement state vector clustering.

[0055] It should be noted that, in the embodiments of the present invention, in a cluster corresponding to a certain type of motion state, the central tendency of all vectors in the cluster can be obtained based on the average value of all motion state vectors in each dimension, which can reflect the core characteristics of this type of motion state.

[0056] A motion type descriptor is a vector constructed based on the statistical characteristics (such as mean and standard deviation) of all motion state vectors corresponding to a certain motion state type. It is a standardized representation of the motion state type and is used to uniformly describe this type of motion state.

[0057] It should be noted that the statistical features of all motion state vectors can be the feature mean vector of high-frequency sampled data (such as 100Hz) within a short time window, which preserves the main motion pattern features of that period.

[0058] Furthermore, for each specific movement state type obtained from the first clustering of a single child (target child or candidate child), the mean value of each dimension of all movement state vectors under that type is calculated, and this mean value vector is determined as the movement type descriptor corresponding to that movement state type, thereby achieving a standardized representation of that type of movement state.

[0059] For any physiological state type, the physiological category descriptor corresponding to the physiological state type is determined based on the statistical characteristics of all physiological state vectors in the physiological state type.

[0060] Among them, the physiological category descriptor refers to a vector constructed with the statistical characteristics of all physiological state vectors corresponding to a certain physiological state type as the core. It is a standardized representation of the physiological state type and is used to uniformly describe the physiological response of that type.

[0061] Furthermore, for each specific physiological state type obtained from the first clustering of a single child (target child or candidate child), the mean of each dimension of all physiological state vectors under that type is calculated, and this mean vector is determined as the physiological category descriptor corresponding to that physiological state type, thereby achieving a standardized representation of that type of physiological state.

[0062] Furthermore, before performing a second clustering process on the motion sequence and the physiological sequence, the following steps are also included: The accuracy of motion description is determined by the cosine similarity between the motion state vector at any given time and the motion type descriptor corresponding to the motion state type at that time.

[0063] After the first DBSCAN clustering, the motion state vector may be classified into different motion categories, such as sitting, walking, jumping, etc.

[0064] Cosine similarity refers to the degree of fit between the motion state vector and the core features of the corresponding type at a given moment by calculating the cosine of the angle between the motion state vector and the motion type descriptor. The value range is [-1, 1], and the closer the value is to 1, the higher the similarity.

[0065] The accuracy of motion description is a parameter determined by the cosine similarity between the motion state vector at that moment and the corresponding motion type descriptor. It can reflect the degree to which the motion state at that moment is accurately described by its respective motion type.

[0066] Furthermore, for a child's motion state vector at any given moment, the motion state type to which the vector belongs is first determined, then the cosine similarity between this motion state vector and the corresponding motion type descriptor is calculated, and finally the cosine similarity value is determined as the accuracy of the motion description at that moment, thereby quantifying the degree to which the motion state at that moment is accurately described by its type.

[0067] The accuracy of physiological description is determined by the cosine similarity between the physiological state vector at any given time and the physiological category descriptor corresponding to the physiological state type at that time.

[0068] Cosine similarity is an index used to measure the similarity between two vectors. Its value ranges from -1 to 1. The closer the value is to 1, the higher the similarity between the two vectors.

[0069] Physiological description accuracy refers to the quantified value of the similarity between the physiological state vector at a certain moment and the corresponding physiological category descriptor, which is obtained by calculating the cosine similarity between the two.

[0070] For example, the accuracy of the physiological description at a given moment can be determined by calculating the cosine similarity between the child's physiological state vector at a certain moment and the descriptor corresponding to the physiological state type to which the vector belongs.

[0071] The accuracy of the time description is determined by multiplying the accuracy of the motion description and the accuracy of the physiological description at any given time.

[0072] The accuracy of motion description refers to the cosine similarity between a child's motion state vector at a given moment and the motion type descriptor of the corresponding motion cluster. The motion state vector is a multi-dimensional vector constructed based on motion data collected by accelerometers, gyroscopes, etc.; the motion type descriptor is a statistical feature of all motion state vectors within that motion cluster, representing the typical characteristics of that type of motion. The closer the value is to 1, the higher the match between the child's motion state and the corresponding motion type at that moment.

[0073] Physiological description accuracy refers to the cosine similarity between a child's physiological state vector at a given moment and the physiological type descriptor of the corresponding physiological cluster. The physiological state vector is a multi-dimensional vector constructed based on physiological sensor data such as blood oxygenation and heart rate; the physiological type descriptor is the statistical characteristic of all physiological state vectors within that physiological cluster, representing the typical features of that type of physiological state. The closer the value is to 1, the higher the match between the child's physiological state and its corresponding physiological type at that moment.

[0074] Accuracy of time description refers to a comprehensive quantitative index of the precision and reliability of the description of a child's combined movement and physiological state at a certain moment. It is obtained by multiplying the accuracy of the movement description and the accuracy of the physiological description at that moment.

[0075] The overall description accuracy is obtained by averaging the description accuracy over all moments within a preset time period.

[0076] The preset time period refers to a fixed time unit (e.g., 5 minutes) for dividing children's motor and physiological data into time-series segments, which is used to unify the data dimensions for subsequent sequence construction, accuracy calculation, model training, and real-time monitoring.

[0077] The accuracy of time description refers to the product of the accuracy of the child's motor description and the accuracy of the physiological description within a certain second. It is a parameter that quantifies the degree to which the motor and physiological states at that moment are accurately represented by the corresponding type of descriptor.

[0078] For example, the arithmetic mean of the accuracy of the description of 300 moments within a preset time period (5 minutes = 300 seconds) can reflect the overall accuracy of the data description within that period.

[0079] Overall description accuracy refers to the average description accuracy of all moments within a preset time period. It is a comprehensive indicator that measures the accuracy of children's movement and physiological state after clustering by descriptors during that period, and provides a reliable basis for subsequent training weight calculation.

[0080] It should be noted that, for all moments within a preset time period, the overall description accuracy for that time period is obtained by calculating the arithmetic mean. This quantifies the comprehensive accuracy of the movement and physiological state represented by the cluster descriptor within that time period, providing a basis for data reliability compensation for the subsequent calculation of training weights.

[0081] Furthermore, before performing a second clustering process on the sequence of motion conditions composed of motion type descriptors and the sequence of physiological conditions composed of physiological type descriptors to obtain the general motion type and the general physiological type, the following steps are also included: For any child, the movement type descriptors and physiological type descriptors corresponding to each moment within a preset time period are arranged in chronological order to determine the child's corresponding movement status sequence and physiological status sequence.

[0082] In this process, the descriptors corresponding to each moment within a preset time period are arranged sequentially according to the chronological order, resulting in the motion sequence and the physiological sequence, respectively.

[0083] A movement sequence refers to a sequence of movement type descriptors arranged chronologically within a preset time period (e.g., 5 minutes), used to reflect the changes in a child's movement status during that period.

[0084] A physiological condition sequence refers to a sequence formed by arranging physiological type descriptors in chronological order at various times within a preset time period, which is used to reflect the changes in the child's physiological state during that period.

[0085] For example, in an embodiment of the present invention, for any child, a preset time period of 5 minutes is used, and the movement type descriptors and physiological type descriptors of each moment in the period are sorted in chronological order to obtain the movement status sequence and physiological status sequence of the child for the corresponding time period.

[0086] For any two children, determine the similarity between their motion state vectors and the similarity between their physiological state vectors at the same time.

[0087] Among them, motion vector similarity refers to the index obtained by calculating the positive correlation normalization value of the cosine similarity of the motion state vectors of any two children at the same moment. The closer the value is to 1, the higher the matching degree of the motion states of the two children at that moment.

[0088] Physiological vector similarity refers to the index obtained by calculating the positive correlation normalization value of the cosine similarity of the physiological state vectors of any two children at the same moment. The closer the value is to 1, the higher the matching degree of the physiological states of the two children at that moment.

[0089] It should be noted that, in the embodiments of the present invention, for any two children, the matching degree of their motion state vectors (motion vector similarity) and the matching degree of their physiological state vectors (physiological vector similarity) at the same moment are calculated respectively, so as to quantify the similarity between the two children's motion and physiological state at that moment, and provide data for subsequent screening of reference children and calculation of developmental consistency.

[0090] The motion difference is determined based on the similarity of motion vectors and the accuracy of motion description, and the physiological difference is determined based on the similarity of physiological vectors and the accuracy of physiological description.

[0091] Among them, the motion difference degree is used to measure the degree of difference in the motion state of two children at the same time. It is obtained by multiplying the motion vector similarity and the motion description accuracy, and then performing negative correlation normalization. The larger the value, the more significant the difference in the motion state of the two children at that time.

[0092] Physiological difference is used to measure the degree of difference in the physiological state of two children at the same time. It is obtained by multiplying the physiological vector similarity and the accuracy of the physiological description, and then performing negative correlation normalization. The larger the value, the more significant the difference in the physiological state of the two children at that time.

[0093] For example, for data from any two children at the same time, the similarity of motion vectors is multiplied by the accuracy of motion description, and the similarity of physiological vectors is multiplied by the accuracy of physiological description. The two product results are then negatively correlated and normalized to obtain the motion difference degree used to measure the difference in motion state between the two children at that time, and the physiological difference degree used to measure the difference in physiological state. This provides a distance metric for subsequent DTW time-series matching and clustering of motion / physiological condition sequences.

[0094] Furthermore, a second clustering process is performed on the motion sequence composed of motion type descriptors and the physiological sequence composed of physiological type descriptors, respectively, to obtain general motion types and general physiological types, including: Using motion difference and physiological difference as distance inputs, temporal matching is performed on motion sequence and physiological sequence to obtain motion difference distance and physiological difference distance.

[0095] Temporal matching refers to the use of the DTW algorithm to align and match the movement and physiological sequences of different children along the time dimension. This algorithm can effectively handle the problem of inconsistent sequence lengths and achieve time-level sequence difference comparison.

[0096] The motion difference distance refers to the overall difference value of motion status at the time interval level obtained after temporal matching of the motion sequences of two children, and is used to measure the similarity of the motion patterns of the two children.

[0097] Physiological difference distance refers to the overall difference value of physiological state at the time period level obtained after temporal matching of the physiological condition sequences of two children, which is used to measure the similarity of physiological response patterns between the two children.

[0098] It should be noted that, in the embodiments of the present invention, the motion difference degree and the physiological difference degree are used as the basis for distance calculation, and the DTW algorithm is used to perform time-series matching of the motion sequence and physiological sequence of the two children, and finally obtain the motion difference distance and physiological difference distance that can reflect the time-level difference between the two.

[0099] Using the distance between movement differences and the distance between physiological differences as distance metrics, a second clustering process is performed on the movement sequence and the physiological sequence to obtain the general movement type and the general physiological type.

[0100] Among them, distance metric refers to the quantitative basis used to measure the similarity between two sequences.

[0101] It should be noted that, in the embodiments of the present invention, the first clustering refers to clustering the motion and physiological state vectors of a single child to classify the child's motion / physiological state type; the second clustering refers to clustering the motion / physiological state sequences of multiple children to extract the general state type at the group level, that is, to solve the problem of non-standardization of children's action definitions through dual clustering.

[0102] For example, a general movement type refers to a category of movement patterns that are common to the group, extracted from the movement sequences of multiple children through a second clustering process, such as walking, jumping, sitting still, etc.

[0103] For example, the general physiological type refers to the category of physiological response patterns that have common characteristics of the group, which are extracted from the physiological condition sequences of multiple children through a second clustering process, such as the heart rate rise after exercise and the blood oxygen stabilization state at rest.

[0104] It should be noted that, in the embodiments of the present invention, the movement difference distance and physiological difference distance, which measure the similarity of different children's movement and physiological time sequence patterns, are used as the judgment criteria. Cluster analysis is performed on the movement and physiological sequences of multiple children to extract general movement types and general physiological types applicable to the group, providing a classification basis for subsequent screening of reference children and calculation of developmental consistency.

[0105] Furthermore, based on the general movement types, representative physiological parameters corresponding to each general movement type are determined, including: For any given general movement type, obtain the physiological sequence of each child under that general movement type.

[0106] Among them, any general movement type refers to the common movement category across children after the sequence of all children's movement patterns are clustered by the second DBSCAN, such as general sitting, general walking, general jumping, etc., which realizes the unification and comparability of different children's movement patterns.

[0107] It should be noted that, in the embodiments of the present invention, for each common movement type across children obtained by the second clustering, the physiological condition sequence of all children (target children + candidate children) under the common movement type is collected, which provides a data basis for subsequent calculation of the DTW distance mean and screening of representative physiological conditions of the common movement type.

[0108] Based on the physiological condition sequence, representative physiological parameters corresponding to each general movement type are determined.

[0109] For example, the representative physiological parameter refers to the physiological sequence with the largest negative correlation normalized value of the mean DTW distance to other physiological sequence sequences under a certain general movement type, which can represent the typical physiological response characteristics of children under that movement type.

[0110] It should be noted that, in the embodiments of the present invention, for each general movement type obtained by clustering, the most representative physiological condition sequence under that type is selected to determine the representative physiological parameters corresponding to that movement type.

[0111] Reference child screening module 103 is used to determine reference children based on representative physiological parameters.

[0112] In this context, reference children refer to those whose developmental similarity to the target children exceeds a preset similarity threshold. Reference children exhibit physiological response patterns similar to the target children under the same exercise load. Introducing reference children aims to eliminate the influence of individual physiological differences (such as physical fitness and basal metabolic rate) on the model baseline. By selecting children with similar physiological response characteristics under the same exercise load as reference samples, the data distribution of different children can be effectively aligned, thereby expanding the training samples through transfer learning and improving the model's generalization ability.

[0113] It should be noted that the reference child screening module calculates and screens children whose developmental consistency meets the standards by comparing the representative physiological parameters of the target child with those of other children in various types of sports, and identifies them as reference children.

[0114] Furthermore, reference children are determined based on representative physiological parameters, including: The developmental consistency between the target child and the candidate child is determined based on representative physiological parameters and the total number of common movement types.

[0115] Among them, the representative physiological parameters refer to the most representative physiological sequence selected under a certain general exercise type, which is used to characterize the typical physiological response pattern corresponding to that general exercise type.

[0116] For example, developmental consistency can be expressed as: ; in, Children The degree of physical developmental consistency with the target child; This indicates the total number of common movement types possessed by the target child; Children and target children The The DTW distance between representative physiological parameters corresponding to different types of exercise; Used for negative correlation normalization.

[0117] For example, the negative correlation normalization function It can be represented as: ; in, It is a very small positive number, used to prevent the denominator from being zero.

[0118] For example, the total number of common movement types refers to the total number of common movement categories across children after the movement sequence of all children is clustered by the second DBSCAN, denoted as A.

[0119] Target children refer to specific children who require real-time stress monitoring; candidate children refer to a large number of other children, besides the target children, who participate in data collection and model training, and whose data are used to screen reference children whose development is consistent with that of the target children.

[0120] It should be noted that if a certain general movement type is missing in any child's data, the DTW distance for that type can be set to a preset maximum value (e.g., 1.0).

[0121] It should be noted that, in the embodiments of the present invention, for the target child and each candidate child, based on the representative physiological conditions of both under all general movement types, the DTW distance of the representative physiological conditions under each general movement type is calculated and negatively normalized, and then the average value is taken in combination with the total number of general movement types to finally obtain the developmental consistency degree used to characterize the similarity of the developmental levels of the two.

[0122] Reference children are determined based on developmental concordance; developmental concordance is used to characterize the similarity of representative physiological parameters between target children and candidate children in the same general movement type.

[0123] Developmental similarity refers to a quantitative indicator used to measure the degree of similarity in physical development between the target child and the candidate child. It is obtained by calculating the mean DTW distance, which represents physiological parameters, for both children under various common movement types, and then normalizing it for negative correlation. A higher value indicates a higher degree of developmental similarity.

[0124] General movement types refer to typical movement patterns identified by clustering the movement sequences of numerous children using the DBSCAN clustering algorithm based on the distance of movement differences between children, such as walking and jumping.

[0125] It should be noted that, in the embodiments of the present invention, based on the index that measures the similarity of physiological parameters between the target child and the candidate child under various general movement types, candidate children whose developmental consistency meets the threshold are selected and determined as reference children.

[0126] Furthermore, reference children are identified based on developmental consistency, including: If the developmental consistency is greater than or equal to the preset consistency threshold, the candidate child will be identified as the reference child.

[0127] The preset consistency threshold refers to a critical value used to determine whether children's developmental levels are similar. For example, the preset consistency threshold can be 0.6. This threshold can filter out candidate children whose developmental matching degree meets the target child, ensuring the effectiveness of subsequent model training data.

[0128] It should be noted that the preset consistency threshold (0.6) can be an empirical value derived from historical datasets. In practical applications, the median or a specific quantile (such as the Top 30%) of the developmental consistency distribution can be selected as the dynamic threshold based on the sample distribution of the current dataset.

[0129] It should be noted that, in the embodiments of the present invention, by calculating the developmental consistency between the target child and the candidate child, the candidate children whose developmental matching degree reaches the preset consistency threshold are screened out and determined as reference children, so as to provide a high-matching reference sample for the subsequent training of the stress prediction model.

[0130] If the developmental consistency is less than the preset consistency threshold, the candidate child will be identified as a non-reference child.

[0131] Among them, non-reference children refer to candidate children whose developmental consistency is lower than the preset consistency threshold. These children have a large difference in developmental level from the target children, and their motor-physiological response patterns are not of reference value for stress monitoring of the target children and will not participate in subsequent model training.

[0132] It should be noted that, in the embodiments of the present invention, if the developmental consistency between the candidate child and the target child is lower than a preset consistency threshold, the candidate child is determined to have a large difference in developmental level from the target child, and is identified as a non-reference child and excluded from the data source for subsequent stress prediction model training, thereby ensuring that the training data matches the developmental characteristics of the target child.

[0133] The model training module 104 is used to train the model based on the corresponding motion sequence, physiological sequence and historical stress level data of the target child and the reference child, combined with training weights, to obtain a trained stress prediction model.

[0134] The historical stress level data refers to the stress level data of the target child and the reference children obtained through comprehensive assessment in the past. This data serves as the output label for model training, providing a basis for model learning.

[0135] For example, training weights refer to quantified values ​​used to adjust the importance of training data. Determined by multiplying the training reference weights by the overall accuracy of the time period description, they can compensate for errors caused by cluster approximation.

[0136] For example, the stress prediction model can be a model trained using a hybrid model of Long Short-Term Memory Transformer (LSTM-Transformer). This model can output the corresponding stress level based on the child's real-time movement and physiological sequences.

[0137] It should be noted that, in the embodiments of the present invention, the model training module takes the motion sequence and physiological sequence of the target child and the reference child as input, and the corresponding historical stress level data as output label, and combines the training weights to train the LSTM-Transformer hybrid model, and finally obtains a model that can predict the stress level of children.

[0138] For example, in the process of obtaining historical stress level data, multiple experts in the fields of child psychology and education can be selected first to provide the experts with de-identified assessment materials, including video clips of children's behavior, physiological indicator fluctuation curves, and scene descriptions, while concealing information that may lead to subjective judgments, such as children's names and ages. Based on a unified assessment dimension (degree of behavioral abnormality, magnitude of physiological deviation, and scene suitability), the experts independently score each potential stress-related period (1-10 points) and fill in the scoring basis.

[0139] Then, calculate the Kendall's coefficient of harmony (W) for all expert scores to test the consistency of the expert evaluation results: if W is greater than or equal to a preset threshold (e.g., 0.7 (an exemplary empirical value, which can be determined in actual applications through grid search or hyperparameter optimization based on the validation set)), it indicates high consistency; if W is less than the preset threshold (e.g., 0.7), organize experts to discuss the samples with large differences, clarify the evaluation criteria, and re-score until the consistency standard is met.

[0140] For example, weights are assigned to different experts, with the weighting coefficient determined based on the expert's years of experience, professional qualifications, and past assessment accuracy. For instance, experts with more than 10 years of experience have a weight of 0.3, while those with 5-10 years have a weight of 0.2. The original stress value (historical stress level data) is calculated using a weighted average formula.

[0141] The original pressure value can be expressed as: ; in, For the first Scores from several experts To correspond to the expert weights, and .

[0142] Finally, the weighted and fused final score is associated with the corresponding motion sequence and physiological sequence as the label value for model training.

[0143] It should be noted that the expert-based labeling process is mainly used in the offline training phase of the stress prediction model to obtain a well-trained model. During the real-time monitoring phase, the stress prediction model can directly output stress levels or be fine-tuned online based on limited user feedback (such as parental confirmation), without relying on real-time expert scoring.

[0144] For example, under low stress (quantitative value 0-3), the motor performance is characterized by relaxed and moderate limb movements, with spontaneous playful actions such as fiddling with stationery or gently swinging the legs. The type of motor state is mostly sitting relaxation or light activity. The physiological performance is characterized by a stable heart rate within the normal range and stable blood oxygen saturation. The physiological state type is resting state. The behavioral characteristics are a pleasant mood, willingness to actively communicate with peers or parents, longer attention span, and the ability to quickly adjust one's mindset when encountering minor problems.

[0145] Under moderate stress (quantitative value 3-6), the motor manifestations include increased frequency of limb movements, repetitive small movements such as frequent hair scratching, nail biting, and pen twirling, with the movement state type mostly being intermittent small movements; the physiological manifestations include a slightly elevated heart rate compared to the low-pressure state, with a larger fluctuation range, and the physiological state type is mild stress state; the behavioral characteristics are easy to be distracted, slower response when communicating with others, and brief hesitation when facing tasks, but still able to complete basic tasks.

[0146] Under high stress (quantitative value 6-10), the motor manifestations include stiff limb movements or obvious agitation, such as restlessness, pacing, and forcefully slamming the table. The types of motor states are mostly vigorous small movements or agitated activities. The physiological manifestations include a significant increase in heart rate and slight fluctuations in blood oxygen saturation. The physiological state type is moderate to severe stress. The behavioral characteristics include irritability or depression, refusal to communicate with others, severe inattention, inability to complete routine tasks, and some children may exhibit crying, avoidance and other behaviors.

[0147] It should be noted that the range of quantized values ​​can be set according to different algorithms, and is not limited in the embodiments of the present invention.

[0148] Furthermore, based on the movement sequences, physiological sequences, and historical stress level data of the target and reference children, and combined with training weights, the model is trained to obtain a well-trained stress prediction model, including: Limb movement vectors are extracted from the reference database, and the importance of each limb movement is determined based on the limb movement vectors.

[0149] Among them, limb motion vectors refer to multi-dimensional vectors constructed based on real-time data collected by motion sensors such as accelerometers and gyroscopes for each part of a child's limbs. Each dimension corresponds to monitoring data from a different motion sensor and is used to quantitatively describe the motion state of a single limb part at a specific moment.

[0150] Action importance refers to a quantitative indicator of the degree of difference between a child's current action and a historical action sequence, used to determine the reference value of that action for stress monitoring. The greater the difference, the higher the action importance value, indicating that the action is more likely to be affected by current emotional stress and contributes more to subsequent model training.

[0151] For example, the importance of an action can be expressed as: ; in, Indicates the body part of the target child at the current moment. The importance of the action at that location; Indicates the target child Body parts in the current time period Sequence of limb movements And the past Body movement sequence during the sliding window period DTW distance between; Indicates the current time period Compared with the past The time interval between sliding windows (the interval between the center time points of two time periods); Used for negative correlation normalization; Used for positive correlation normalization; This indicates the total number of sliding windows within a reference time period (e.g., preset to 3 minutes), where To prevent the smoothing coefficient from being zero in the denominator (e.g.: ).

[0152] It should be noted that calculating the difference through DTW actually determines the deviation of the timing pattern, which can be used to identify the turning point when switching from a normal state to a stress state.

[0153] For example, the positive correlation normalization function It can be represented as: ; in, It is a very small positive number, used to prevent the denominator from being zero.

[0154] It should be noted that, in the embodiments of the present invention, based on the quantification carrier (limb movement vector) of the movement of each part of the child's limbs, the importance index of each limb movement is calculated and determined by comparing the degree of difference between the current limb movement sequence and the historical limb movement sequence, so as to provide a basis for subsequent selection of key limb movements and calculation of training weights.

[0155] Training reference weights are determined based on developmental consistency and the importance of the movement.

[0156] For example, the importance of a movement refers to an indicator used to measure the difference between a child's current movement and historical movements of a certain limb. It can be obtained by using a sliding window with a length of 5 seconds and a step length of 4 seconds to calculate the DTW distance between the current limb movement sequence and the historical sequence within the past 3 minutes. The result is obtained by weighted normalization based on the time interval. The larger the value, the more special the limb movement is and the more likely it is to be related to stress.

[0157] Training reference weights refer to weight parameters calculated by combining developmental consistency and motor importance. They are used to measure the reference value of data from children at a certain time period for model training. The weights are obtained by multiplying the two data and then normalizing them with positive correlation. The higher the value, the higher the reference priority of the data at that time period.

[0158] For example, the training reference weights can be represented as: ; in, See reference for children's first Training reference weights for important limb movements and related data corresponding to different time periods; This indicates the set of data Corresponding reference children and target children Developmental consistency between them; This indicates the reference child during this period. The importance of a limb movement corresponds to the importance of the movement in that limb part. Used for positive correlation normalization.

[0159] It should be noted that, in the embodiments of the present invention, for a certain period of data of the reference child, the developmental consistency between the child and the target child is multiplied by the importance of the important limb movements corresponding to the important limb movements in that period, and then subjected to positive correlation normalization processing to finally determine the training reference weight of the data in that period. This highlights the role of reference data with high developmental similarity and significant movement features in model training and improves the model monitoring accuracy.

[0160] For example, important limb movements can be the movements of the limb parts corresponding to the maximum importance of each limb movement.

[0161] The training weights are determined based on the training reference weights and the overall descriptive accuracy.

[0162] The training weights refer to the final quantified values ​​used to adjust the priority of training data weights. They are obtained by multiplying the training reference weights by the overall descriptive accuracy, and can increase the proportion of important and reliable data in model training.

[0163] For example, the final training weights used for model training can be determined by multiplying the training reference weights, which measure the importance of the reference data, by the overall descriptive accuracy that assesses the reliability of data monitoring.

[0164] The stress prediction model is trained by using the motion sequence, physiological sequence, and historical stress level data of the target child and the reference child as training samples, adjusting the contribution of each sample through training weights, and obtaining a well-trained stress prediction model.

[0165] The training sample refers to a dataset consisting of the movement and physiological sequences (input features) of the target children and reference children, as well as historical stress level data (output labels) for the corresponding time periods.

[0166] It should be noted that, in the embodiments of the present invention, during the training of the stress prediction model, the shorter sequences in the motion sequence and physiological sequence are interpolated to ensure that the same type of input sequences are of equal length, and the training weight is the product of the training reference weight and the overall description accuracy of the corresponding time period.

[0167] It should be noted that, in the embodiments of the present invention, the movement and physiological sequences of the target child and the reference child are used as input features, and the historical stress level data of the corresponding time period is used as output labels to construct training samples; then training weights are introduced to adjust the contribution of different samples, and the LSTM-Transformer hybrid model is trained to finally obtain a model that can predict children's stress levels in real time.

[0168] The monitoring module 105 is used to collect the target child's movement and physiological data within a preset time period, input them into the trained stress prediction model, and output the target child's stress level.

[0169] The stress level is the result of the model output, used to quantify the psychological stress level of the target children within a preset time period, so as to intuitively indicate to medical staff that there are children with high psychological stress risk. However, it should be emphasized that the numerical value of the stress level of the target children does not directly determine whether subsequent stress intervention work will be carried out. Whether stress intervention work will be carried out for the target children and the intensity of stress intervention need to be determined by medical staff based on their own experience and their actual observation results of the target children.

[0170] It should be noted that, in the embodiments of the present invention, the monitoring module collects the target child's movement data and physiological data within a preset time period, inputs them into a stress prediction model that has been trained, and finally outputs the target child's current stress level, thereby realizing real-time dynamic monitoring of the child's stress.

[0171] In summary, this invention constructs a reference database by collecting movement and physiological data from target and candidate children and combining this data with historical stress levels. The extracted movement and physiological state vectors are then subjected to a first clustering process to obtain a unique movement / physiological state type and corresponding descriptor for each child. Subsequently, the descriptors are arranged chronologically to form a sequence and subjected to a second clustering process to classify common movement / physiological types across children. Representative physiological conditions corresponding to each common movement type are then selected to eliminate interference from children's daily movement on physiological indicators and avoid misjudgments of stress due to fluctuations in exercise volume. Simultaneously, developmental consistency is calculated based on representative physiological conditions to select reference children with similar developmental levels to the target children. Training weights are calculated by combining developmental consistency, movement importance, and data accuracy. The stress prediction model is trained using the movement and physiological state sequences of the target and reference children, along with historical stress level data, making the training samples more closely match the physiological and behavioral characteristics of the target children and significantly improving the model's prediction accuracy. Finally, by collecting data from the target children within a preset time period in real time and inputting it into the model, the accuracy of monitoring children's stress levels is improved.

[0172] It should be noted that the system provided in the above embodiments is only an example of the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above.

[0173] In one embodiment, the present invention also provides a method for real-time monitoring of children's stress by combining behavioral data analysis, such as... Figure 2 As shown, the method includes: Step S1: Collect the motor and physiological data of the target children and candidate children, and build a reference database based on the motor data, physiological data and historical stress level data; Step S2: Extract motion state vectors and physiological state vectors from the reference database, and perform a first clustering process on the motion state vectors and physiological state vectors respectively to obtain motion state type, physiological state type, motion type descriptor, and physiological type descriptor; perform a second clustering process on the motion situation sequence composed of motion type descriptors and the physiological situation sequence composed of physiological type descriptors respectively to obtain general motion type and general physiological type; determine the representative physiological parameters corresponding to each general motion type based on the general motion type. Step S3: Determine the reference child based on representative physiological parameters; Step S4: Based on the movement sequence, physiological sequence, and historical stress level data of the target child and the reference child, and combined with the training weights, train the model to obtain a trained stress prediction model. Step S5: Collect the target child's exercise and physiological data within a preset time period, input them into the trained stress prediction model, and output the target child's stress level.

[0174] The above embodiments provide a child stress real-time monitoring system that combines behavioral data analysis and a child stress real-time monitoring method that combines behavioral data analysis, which belong to the same concept. The specific implementation process is detailed in the system embodiments and will not be repeated here.

[0175] This invention also provides an electronic device. Please refer to [link to relevant documentation]. Figure 3 The electronic device may include a processor 301, a memory 302, and a program 3021 stored in the memory 302 and capable of running on the processor 301.

[0176] When program 3021 is executed by processor 301, it can achieve the following: Figure 2 Any steps in the corresponding method embodiments and the achievement of the same beneficial effects will not be repeated here.

[0177] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by hardware related to program instructions, and the program can be stored in a readable medium.

[0178] This invention also provides a readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described functions. Figure 2 Any step in the corresponding method embodiment can achieve the same technical effect, and will not be repeated here to avoid repetition.

[0179] The computer-readable storage medium of this invention can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. For example, a computer-readable storage medium can be an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0180] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0181] The program code contained on the storage medium can be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0182] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or terminal. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0183] This invention also provides a computer program product that, when run on a computer, causes the computer to perform the aforementioned steps to achieve the real-time monitoring method for children's stress that combines behavioral data analysis provided in the above embodiments.

[0184] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0185] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A real-time stress monitoring system for children that combines behavioral data analysis, characterized in that, The system includes: The data acquisition module is used to collect the movement data and physiological data of the target children and candidate children, and to build a reference database based on the movement data, the physiological data and historical stress level data; The data preprocessing module is used to extract motion state vectors and physiological state vectors from the reference database, and perform a first clustering process on the motion state vectors and the physiological state vectors respectively to obtain motion state type, physiological state type, motion type descriptor, and physiological type descriptor; perform a second clustering process on the motion situation sequence composed of the motion type descriptor and the physiological situation sequence composed of the physiological type descriptor respectively to obtain general motion type and general physiological type; and determine the representative physiological parameters corresponding to each general motion type based on the general motion type. The reference child screening module is used to determine reference children based on the representative physiological parameters. The model training module is used to train the model based on the movement sequence, physiological sequence and historical stress level data of the target child and the reference child, combined with training weights, to obtain a trained stress prediction model. The monitoring module is used to collect the target child's movement and physiological data within a preset time period, input them into the trained stress prediction model, and output the target child's stress level.

2. The real-time child stress monitoring system combining behavioral data analysis according to claim 1, characterized in that, The first clustering process is performed on the motion state vector and the physiological state vector respectively to obtain motion state type, physiological state type, motion type descriptor, and physiological type descriptor, including: The motion state vector and the physiological state vector corresponding to each child are subjected to a first clustering process to obtain at least one motion state type and at least one physiological state type corresponding to each child. For any motion state type, the motion type descriptor corresponding to the motion state type is determined based on the statistical characteristics of all motion state vectors in the motion state type. For any given physiological state type, the physiological category descriptor corresponding to the physiological state type is determined based on the statistical characteristics of all physiological state vectors in the physiological state type.

3. The real-time child stress monitoring system combining behavioral data analysis according to claim 2, characterized in that, Before performing the second clustering process on the motion sequence and the physiological sequence, the method further includes: The accuracy of motion description is determined by the cosine similarity between the motion state vector at any given time and the motion type descriptor corresponding to the motion state type at that time. The accuracy of physiological description is determined by the cosine similarity between the physiological state vector at any given time and the physiological category descriptor corresponding to the physiological state type at that time. The accuracy of time description is determined by multiplying the accuracy of motion description and the accuracy of physiological description at any given time. The overall description accuracy is obtained by averaging the description accuracy over all moments within the preset time period.

4. The real-time child stress monitoring system combining behavioral data analysis according to claim 3, characterized in that, Before performing a second clustering process on the sequence of motion conditions composed of the motion type descriptors and the sequence of physiological conditions composed of the physiological type descriptors to obtain the general motion type and the general physiological type, the process further includes: For any child, the movement type descriptors and physiological type descriptors corresponding to each moment within the preset time period are arranged in chronological order to determine the movement status sequence and physiological status sequence corresponding to the child; For any two children, determine the similarity between the motion vectors of the two children at the same time and the similarity between the physiological vectors of the two children. Based on the motion vector similarity and the motion description accuracy, the motion difference is determined, and based on the physiological vector similarity and the physiological description accuracy, the physiological difference is determined.

5. The real-time child stress monitoring system combining behavioral data analysis according to claim 4, characterized in that, The second clustering process is performed on the sequence of movement conditions composed of the movement type descriptors and the sequence of physiological conditions composed of the physiological type descriptors, respectively, to obtain a general movement type and a general physiological type, including: Using the motion difference degree and the physiological difference degree as distance inputs, temporal matching is performed on the motion sequence and the physiological sequence to obtain the motion difference distance and the physiological difference distance. Using the motion difference distance and the physiological difference distance as distance metrics, a second clustering process is performed on the motion sequence and the physiological sequence to obtain a general motion type and a general physiological type.

6. The real-time child stress monitoring system combining behavioral data analysis according to claim 1, characterized in that, The step of determining representative physiological parameters corresponding to each of the general movement types based on the general movement types includes: For any given general movement type, obtain the physiological sequence of each child under that general movement type; Based on the physiological condition sequence, representative physiological parameters corresponding to each of the general movement types are determined.

7. The real-time child stress monitoring system combining behavioral data analysis according to claim 1, characterized in that, The process of determining the reference child based on the representative physiological parameters includes: Based on the representative physiological parameters and the total number of common movement types, the developmental consistency between the target child and the candidate child is determined; Reference children are determined based on the developmental concordance; the developmental concordance is used to characterize the similarity of the representative physiological parameters between the target child and the candidate child in the same general movement type.

8. The real-time child stress monitoring system combining behavioral data analysis according to claim 7, characterized in that, The determination of reference children based on the developmental consistency includes: If the developmental consistency is greater than or equal to a preset consistency threshold, the candidate child is identified as a reference child. If the developmental consistency is less than the preset consistency threshold, the candidate child is identified as a non-reference child.

9. The real-time child stress monitoring system combining behavioral data analysis according to claim 1, characterized in that, The step of training a model based on the movement sequence, physiological sequence, and historical stress level data of the target child and the reference child, combined with the training weights, to obtain a trained stress prediction model includes: Limb movement vectors are extracted from the reference database, and the importance of each limb movement is determined based on the limb movement vectors. The training reference weights are determined based on developmental consistency and the importance of the aforementioned movements; The training weights are determined based on the training reference weights and the overall descriptive accuracy. The exercise sequence, physiological sequence, and historical stress level data corresponding to the target child and the reference child are used as training samples. The contribution of each sample is adjusted by the training weights to train the stress prediction model and obtain a trained stress prediction model.

10. The real-time child stress monitoring system combining behavioral data analysis according to claim 1, characterized in that, The system also includes: The clustering module is used to perform a first clustering process on the motion state vector and the physiological state vector respectively to obtain at least one motion state type and at least one physiological state type for each child. The determination module is used to determine, for any motion state type, a motion category descriptor corresponding to the motion state type based on the statistical characteristics of all motion state vectors in the motion state type; and to determine, for any physiological state type, a physiological category descriptor corresponding to the physiological state type based on the statistical characteristics of all physiological state vectors in the physiological state type.

Images