Infant care quality evaluation system and method based on data analysis

Abstract

The application relates to the field of activity evaluation and discloses a baby and infant care quality evaluation system and method based on data analysis, which comprises a behavior statistical unit, a daily activity video of a baby and infant in a care place is acquired by the behavior statistical unit, behavior recognition is performed on the baby and infant according to the daily activity video, the behavior characteristics of the baby and infant are obtained, and the behavior statistics of the care place are performed according to the behavior characteristics of the baby and infant, so that the behavior distribution of the baby and infant in the care place is obtained; and an activity evaluation unit, the quality evaluation unit is used for extracting the activity track of the care place according to the behavior distribution of the baby and infant, so that the activity track of the baby and infant in the care place is obtained. The behavior statistical unit can acquire the daily activity video of the baby and infant, perform behavior recognition on the daily activity video, accurately capture the behavior characteristics of the baby and infant in the care process, and in-depth understand the behavior distribution of the baby and infant by performing statistical analysis on the behavior characteristics.

Description

Technical Field

[0001] This invention relates to the field of activity assessment technology, specifically to a data analysis-based system and method for assessing the quality of infant and toddler care. Background Technology

[0002] Existing systems suffer from the following drawbacks: Traditional systems often rely on manual observation and simple recording, lacking automated behavior recognition technology. This makes the identification and analysis of infant and toddler behavior incomplete and inaccurate, easily overlooking certain details or key behavioral characteristics, leading to an inaccurate grasp of infants' and toddlers' needs. Behavioral analysis relies heavily on subjective judgment, which may be biased. Furthermore, traditional systems mostly rely on simple activity records and observations, failing to evaluate the quality and effectiveness of infants' and toddlers' activities through in-depth data analysis. They typically lack comprehensive analysis of the types and trajectories of infants' and toddlers' activities, making it impossible to know whether infants and toddlers have participated in beneficial and age-appropriate activities. Moreover, traditional systems often ignore the impact of the childcare environment on infants' and toddlers' behavior. Environmental adaptability analysis often lacks specific quantitative basis, easily overlooking potential problems in the environment (such as temperature, humidity, and light), which may affect the quality of infants' and toddlers' activities and health. In addition, traditional systems usually rely on manual intervention to adjust the childcare environment, failing to achieve automatic optimization based on data. For example, environmental parameters such as temperature, light, and humidity often require manual control, and these adjustments are static, making it difficult to adjust in real time according to the needs of infants and toddlers' activities. Finally, traditional systems often rely on only a single monitoring dimension, lacking a comprehensive quality assessment mechanism. The quality of infants' and toddlers' growth and activities can usually only be assessed from individual perspectives, lacking holistic and systematic quality feedback.

[0003] Therefore, this application proposes a data analysis-based infant and toddler care quality assessment system and method to address the aforementioned problems. Summary of the Invention

[0004] To achieve the above objectives, the present invention provides the following technical solution: a data analysis-based infant and toddler childcare quality assessment system, comprising:

[0005] A behavior statistics unit is used to acquire videos of the daily activities of infants and toddlers in the childcare center; to perform behavior recognition on the daily activity videos to obtain the behavioral characteristics of the infants and toddlers; and to perform behavior statistics on the childcare center based on the behavioral characteristics of the infants and toddlers to obtain the behavioral distribution of the infants and toddlers in the childcare center.

[0006] An activity assessment unit is included, wherein the quality assessment unit is used to extract activity trajectories of the childcare center based on the infant and toddler behavior distribution to obtain the activity trajectories of the infants and toddlers within the childcare center; and to assess the activity quality of the infants and toddlers based on the activity trajectories to obtain the activity quality of the infants and toddlers.

[0007] An environmental identification unit is used to classify the childcare center according to the childcare demand cycle based on the quality of the infants' activities, so as to obtain a childcare demand cycle benchmark; acquire real-time environmental data of the childcare center; and identify the environmental adaptation of the infants' activities based on the real-time environmental data, so as to obtain an environmental adaptation influence coefficient.

[0008] The comprehensive evaluation unit is used to correct the childcare quality environment based on the environmental adaptability influence coefficient to obtain the childcare environment quality; and to map the childcare service quality based on the childcare environment quality and the infant activity quality to obtain the comprehensive childcare quality.

[0009] Preferably, behavioral recognition is performed on the infants and toddlers based on the daily activity videos to obtain their behavioral characteristics, including:

[0010] The daily activity video is divided into video frames to obtain a daily activity image sequence; the action segmentation network is used to identify the boundaries of infant limb movements through the daily activity image sequence to obtain action video clips; wherein, the action video clips include the action start frame, duration and spatial occupancy information;

[0011] The motion video clips are localized using a 3D human pose estimation model to obtain a pose sequence; wherein, the pose sequence includes the 3D coordinates and motion trajectories of multiple key human joints.

[0012] The preset infant basic behavior pattern library is matched with the posture sequence to obtain the infant behavior type label; the action video clip, the posture sequence and the behavior type label are fused to obtain the infant behavior characteristics; wherein, the infant behavior characteristics include action intensity, behavior frequency and interaction object.

[0013] Preferably, behavioral statistics are performed on the childcare center based on the infants' behavioral characteristics to obtain the distribution of infant behaviors within the childcare center, including:

[0014] Threshold cleaning is performed on the infant and toddler behavioral characteristics to obtain a set of effective behavioral fragments;

[0015] A behavior pattern topology graph is constructed based on the set of effective behavior fragments; the action chain is clustered using the behavior pattern topology graph through a graph convolutional network to obtain a topology feature matrix; wherein, the topology feature matrix includes action connection relationships and behavior transformation probabilities;

[0016] The continuous monitoring period is divided into time windows of preset duration using time series binning technology. Within each time window, the node activation frequency of the behavior pattern topology map is counted to obtain a behavior pattern heatmap. The behavior pattern heatmap has a periodicity marked by timestamps.

[0017] The childcare center is divided into interactive units of a preset size, and the dwell time ratio of behavioral patterns in each interactive unit is counted to obtain a spatial behavior density cloud map.

[0018] The behavioral pattern heatmap and the spatial behavioral density cloud map are fused to obtain the distribution of infant and toddler behaviors.

[0019] Preferably, the activity trajectory of the childcare center is extracted based on the infant behavior distribution to obtain the activity trajectory of the infants within the childcare center, including:

[0020] The childcare center is divided into functional areas, and a gridded spatial topology map is generated by combining the spatial behavior density cloud map of the infant behavior distribution; wherein, the gridded spatial topology map is labeled with behavior preference tags;

[0021] According to the dynamic time warping algorithm, the spatiotemporal neighborhood matching of continuous trajectory points of infants and young children is performed through the gridded spatial topology map to obtain a spatiotemporal trajectory primitive sequence; wherein, the spatiotemporal trajectory primitive sequence includes behavioral pattern transfer links;

[0022] The spatiotemporal trajectory primitive sequence is context-aware encoded using a graph attention network to obtain a motivation trajectory map; wherein, the motivation trajectory map includes typical trajectory motivation patterns and behavioral intention markers;

[0023] The motivation trajectory map and the environmental data of the childcare center are multimodally aligned to obtain a spatiotemporal trajectory correlation matrix; wherein, the spatiotemporal trajectory correlation matrix includes the intensity of environmental interaction; wherein, the environmental data includes toy distribution and caregiver location;

[0024] Feature extraction is performed on the spatiotemporal trajectory correlation matrix using a spatiotemporal convolutional network to obtain the infant activity trajectory; wherein, the infant activity trajectory includes spatial movement rate, behavior transformation probability, and environmental response delay.

[0025] Preferably, the activity quality of the infant is assessed based on the infant's activity trajectory to obtain the quality of the infant's activity, including:

[0026] The activity trajectory of the infant is analyzed for features to obtain an activity intensity fingerprint; wherein, the activity intensity fingerprint includes movement speed, trajectory curvature and spatial residence heat.

[0027] The activity intensity fingerprint and the distribution of teaching aids in the childcare center are spatiotemporally aligned to obtain a behavioral sequence map; wherein the behavioral sequence map is marked with cognitive stage markers; wherein the cognitive stage markers include observation, manipulation and creation;

[0028] The frequency and duration of interactions between infants and toddlers and entities are statistically analyzed based on the behavioral sequence map to obtain a cognitive engagement matrix; wherein, the cognitive engagement matrix includes social initiative, operational focus, and exploration depth; the entities include peers, caregivers, and teaching aids;

[0029] The activity intensity fingerprint, the behavior sequence map, and the cognitive participation matrix are fused to obtain the activity quality of infants and toddlers.

[0030] Preferably, the childcare center is classified according to the quality of the infants' and toddlers' activities to obtain a childcare demand cycle benchmark, including:

[0031] The demand patterns of infants and toddlers are classified according to the quality of their activities using an unsupervised clustering algorithm to obtain a baseline of periodic behavior patterns; wherein the baseline of periodic behavior patterns is labeled with demand type; the demand patterns include high-intensity physical activity, deep cognitive exploration, and socio-emotional connection.

[0032] The periodic behavior pattern baseline and the childcare center's schedule data are spatiotemporally aligned to obtain the matching deviation between the standard process and the periodic behavior pattern baseline, and a demand-supply coupling degree heat map is generated based on the matching deviation.

[0033] The demand-supply coupling heatmap is used to perform time-series prediction based on the Long Short-Term Memory Network to obtain the demand response baseline curve; wherein, the demand response baseline curve includes the demand surge warning threshold, the supply redundancy critical point, and the intervention response window period;

[0034] The demand response baseline curve and the standardized demand cycle template are fused together to obtain the childcare demand cycle baseline; wherein, the childcare demand cycle baseline includes real-time demand intensity and cyclical fluctuation pattern.

[0035] Preferably, environmental adaptation identification is performed on the quality of the infant's activities based on the real-time environmental data to obtain an environmental adaptation influence coefficient, including:

[0036] The real-time environmental data and the quality of infant activity are synchronized to obtain a spatiotemporally aligned data chain; wherein the spatiotemporally aligned data chain is labeled with an environment tag.

[0037] Semantic association is performed on the behavior intensity fingerprint, the cognitive participation matrix, and the real-time environment data in the spatiotemporal aligned data chain to obtain a feature fusion tensor; wherein, the feature fusion tensor includes environment and behavior interaction patterns;

[0038] The adaptation curve between the real-time environmental data and the quality of the infant's activities was determined based on the infant physiological tolerance database.

[0039] Preferably, the method of identifying environmental adaptation based on the real-time environmental data to obtain an environmental adaptation influence coefficient further includes:

[0040] Structural learning is performed on the feature fusion tensor to obtain causal links;

[0041] The causal link includes identifying sudden temperature changes, surges in exercise volume, and exceeding fatigue limits, and generating an environmental impact topology map with the direction of action based on the causal link;

[0042] The environmental impact topology map and the adaptation curve are weighted and fused to obtain the environmental fitness impact coefficient; wherein the environmental fitness impact coefficient includes stimulus intensity, tolerance threshold and compensation requirement.

[0043] Preferably, the childcare quality environment is corrected based on the environmental adaptability influence coefficient to obtain the childcare environment quality, including:

[0044] The environmental adaptability impact coefficient and the childcare demand cycle benchmark are synchronized to obtain the demand cycle spectrum; wherein the demand cycle spectrum is labeled with environmental intervention.

[0045] An environmental sensitivity analysis is performed on the demand period spectrum based on an adaptive gated cyclic network to obtain an environmentally coupled demand correction curve; wherein, the environmentally coupled demand correction curve includes an environmental demand attenuation mode.

[0046] Counterfactual inference is performed on the environmentally coupled demand correction curve to obtain the environmental correction weight matrix; the childcare quality environment is corrected based on the physiological tolerance characteristics of infants and young children and the environmental correction weight matrix to obtain the childcare environment quality.

[0047] The data-driven method for assessing the quality of infant and toddler care, applicable to the aforementioned data-driven infant and toddler care quality assessment system, includes:

[0048] The system acquires videos of the daily activities of infants and toddlers in a childcare center; it performs behavioral recognition on the videos to obtain behavioral characteristics of the infants and toddlers; and it performs behavioral statistics on the childcare center based on these behavioral characteristics to obtain the behavioral distribution of the infants and toddlers in the childcare center.

[0049] Based on the distribution of infant and toddler behavior, the activity trajectory of the childcare center is extracted to obtain the activity trajectory of the infants and toddlers within the childcare center; based on the activity trajectory of the infants and toddlers, the activity quality of the infants and toddlers is assessed to obtain the activity quality of the infants and toddlers.

[0050] The childcare center is classified into childcare demand cycles based on the quality of infant and toddler activities to obtain a childcare demand cycle benchmark; real-time environmental data of the childcare center is obtained; environmental adaptation identification of the quality of infant and toddler activities is performed based on the real-time environmental data to obtain an environmental adaptation influence coefficient.

[0051] The childcare environment quality is corrected based on the environmental adaptability influence coefficient to obtain the childcare demand cycle benchmark; the childcare service quality is mapped based on the childcare environment quality and the infant activity quality to obtain the comprehensive childcare quality.

[0052] Compared with the prior art, the beneficial effects of the present invention are:

[0053] (1) The present invention can acquire videos of infants' daily activities and identify their behavior through the behavior statistics unit, accurately capture the behavioral characteristics of infants in the childcare process, and gain a deeper understanding of the behavioral distribution of infants by statistical analysis of these behavioral characteristics, such as the types of activities, rest time, and frequency of interaction of infants. This can provide a scientific basis for optimizing the childcare environment and ensure that infants receive appropriate attention and care during the childcare process.

[0054] (2) This invention assesses the quality of an infant’s activities by analyzing the distribution of their behavior and activity trajectories. This assessment not only helps monitor the level of an infant’s activity, but also reflects whether the infant has participated in beneficial and age-appropriate activities. This provides data support for childcare institutions, enabling them to adjust activity arrangements and ensure that infants and young children achieve good development in terms of physical, cognitive and emotional aspects.

[0055] (3) The present invention can perform environmental adaptability analysis through real-time environmental data and the quality of infants' activities by using the environmental identification unit, and identify the degree of influence of the childcare environment on the quality of infants' activities. This can reveal the adaptive influence of the childcare environment on infants' behavior and activities, help identify potential environmental problems, and make timely adjustments to ensure the comfort and healthy development of infants.

[0056] (4) This invention uses the environmental adaptability influence coefficient to modify the childcare demand cycle benchmark through a comprehensive evaluation unit, thereby further optimizing the quality of the childcare environment. In this way, the childcare environment can be adjusted in real time according to the actual needs of infants and young children, such as adjusting the temperature or light intensity in a timely manner, to ensure that infants and young children grow up in the best environment. Through this data-driven dynamic optimization, the quality of the childcare environment can be greatly improved, ensuring that infants and young children receive more efficient and targeted care. Attached Figure Description

[0057] Figure 1 This is a schematic diagram of the overall system architecture in one embodiment of the present invention;

[0058] Figure 2 This is a schematic diagram of the overall method's steps in one embodiment of the present invention.

[0059] In the diagram: 1. Behavioral statistics unit; 2. Activity evaluation unit; 3. Environmental identification unit; 4. Comprehensive evaluation unit. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] Example 1, please refer to Figure 1 This invention provides a technical solution: a data analysis-based infant and toddler childcare quality assessment system, comprising:

[0062] Behavior statistics unit 1 is used to acquire videos of the daily activities of infants and toddlers in the childcare center; to identify the behavior of infants and toddlers based on the videos of daily activities to obtain the behavioral characteristics of infants and toddlers; and to conduct behavioral statistics on the childcare center based on the behavioral characteristics of infants and toddlers to obtain the behavioral distribution of infants and toddlers in the childcare center.

[0063] Activity assessment unit 2 and quality assessment unit 2 are used to extract activity trajectories of the childcare center based on the distribution of infant and toddler behavior, so as to obtain the activity trajectories of infants and toddlers in the childcare center; and to conduct activity quality assessment of infants and toddlers based on the activity trajectories, so as to obtain the activity quality of infants and toddlers.

[0064] The environmental identification unit 3 is used to classify the childcare center according to the childcare demand cycle based on the quality of infants' and toddlers' activities, so as to obtain the childcare demand cycle benchmark; to obtain the real-time environmental data of the childcare center; and to identify the environmental adaptation of the quality of infants' and toddlers' activities based on the real-time environmental data, so as to obtain the environmental adaptation influence coefficient.

[0065] Comprehensive assessment unit 4 is used to correct the childcare quality environment based on the childcare demand cycle benchmark according to the environmental adaptability influence coefficient, so as to obtain the childcare environment quality; and to map the childcare service quality based on the childcare environment quality and the quality of infant and toddler activities, so as to obtain the comprehensive childcare quality.

[0066] In an optional embodiment, infant behavior recognition is performed based on videos of daily activities to obtain infant behavioral characteristics, including:

[0067] Video frames of daily activity videos are divided to obtain daily activity image sequences; based on the action segmentation network, the boundaries of infant and toddler limb movements are identified through the daily activity image sequences to obtain action video segments; wherein, the action video segments include the action start frame, duration and spatial occupancy information;

[0068] The motion video clips are localized by locating the joint coordinates based on the 3D human pose estimation model to obtain the pose sequence; the pose sequence includes the 3D coordinates and motion trajectory of multiple key human joints.

[0069] The pre-defined database of basic infant and toddler behavior patterns is matched with posture sequences to obtain infant and toddler behavior type labels; the action video clips, posture sequences and behavior type labels are fused to obtain infant and toddler behavior characteristics; among which, infant and toddler behavior characteristics include action intensity, behavior frequency and interaction objects.

[0070] It's important to note that breaking down daily activity videos into continuous image frames (or image sequences) is the first step in in-depth video analysis. Dividing the video into frames according to time sequence yields individual image frames; these frames form the basis for subsequent behavioral analysis, with each frame representing a static image of the infant at a specific moment. In this way, the video is transformed into processable image data, facilitating the analysis of actions and postures. The boundaries of infants' limb movements are identified, allowing for precise extraction of the start and end points of actions. Using an action segmentation network, the starting point, duration, and spatial positioning of infants' limb movements are identified through analysis of the image sequence. Specifically, action segmentation... The cut network can detect the boundaries of actions, such as an infant's movement from sitting to standing, or from reaching to grasping. The output information for each action video segment includes: action start frame: the time point at which the action begins; duration: the duration of the action; spatial occupancy information: the position or posture change of the action in space. Through a 3D pose estimation model, the 3D coordinates of each joint of the infant's body are accurately obtained, and their movement trajectory is tracked. Applying 3D human pose estimation technology, the joint information of the infant (such as the coordinates of the head, arms, legs, etc.) is extracted from the action video segments. This joint information accurately describes how different parts of the infant's body move in space. Movement; posture sequences include: 3D coordinates of key human joints: such as the positions of the head, shoulders, wrists, etc. in 3D space; movement trajectory: the movement trajectory of each joint over time; by matching the infant's posture sequence with a preset behavior pattern library, the system identifies the specific behavior type of the infant; the system will compare and match the posture sequence extracted from the video with a preset "basic infant behavior pattern library"; this library contains various common infant behavior patterns (such as crawling, standing, walking, etc.); by comparing posture sequences, the system can identify the specific behavior type that the infant is performing; each action segment is marked as a specific behavior type. Tags such as "crawling," "eating," and "playing" are used to comprehensively describe the behavioral characteristics of infants and toddlers by fusing action video clips, posture sequences, and behavior type tags. The results from the above steps are integrated to extract comprehensive characteristics of infant and toddler behavior, including: Action intensity: describing the activity level of the infant's behavior; for example, some behaviors (such as running) may have higher action intensity, while other behaviors (such as remaining still) may have lower intensity; Behavior frequency: counting the number of times a certain behavior occurs in an infant or toddler; this helps to understand the infant's activity patterns over a certain period of time; Interaction objects: identifying whether the infant or toddler interacts with other people or objects, helping to analyze their social behavior and interaction frequency.

[0071] In an optional embodiment, behavioral statistics are performed on the childcare center based on the behavioral characteristics of infants and toddlers to obtain the distribution of infant and toddler behaviors within the childcare center, including:

[0072] Threshold cleaning is performed on the behavioral characteristics of infants and young children to obtain a set of effective behavioral fragments;

[0073] A behavior pattern topology graph is constructed based on the set of effective behavior fragments; the action chain is clustered using the behavior pattern topology graph through a graph convolutional network to obtain a topology feature matrix; the topology feature matrix includes action connection relationships and behavior transformation probabilities.

[0074] The continuous monitoring period is divided into time windows of preset duration using time series binning technology. Within each time window, the node activation frequency of the behavior pattern topology map is counted to obtain a behavior pattern heatmap. The behavior pattern heatmap has a periodicity marked by timestamps.

[0075] The childcare center was divided into interactive units of a preset size, and the dwell time of behavioral patterns in each interactive unit was counted to obtain a spatial behavior density cloud map.

[0076] The behavioral pattern heatmap and spatial behavioral density cloud map are fused to obtain the distribution of infant and toddler behaviors.

[0077] It is important to note that the process involves filtering out valid behavioral fragments from a large amount of infant and toddler behavior data, removing noise or invalid data; applying threshold cleaning techniques to infant and toddler behavioral characteristics (e.g., setting an activity intensity threshold, ignoring behaviors below this threshold) to filter out those behavioral fragments with practical significance; this step ensures that the data upon which subsequent analysis is based is valid and accurate; a set of valid behavioral fragments is defined as containing real and valid behavioral data fragments of infants and toddlers; a structured graph describing the relationships between infant and toddler behavioral patterns is constructed for subsequent behavioral pattern clustering and analysis; based on the data in the set of valid behavioral fragments, the infant and toddler's action chains (i.e., a continuous sequence of behaviors) are analyzed; these behavioral patterns are then analyzed through... A topological graph is constructed to represent infant and toddler behavioral patterns, where nodes represent different behavioral patterns and edges represent the transition relationships between these patterns. The topological graph displays structured information about infant and toddler behavioral patterns, facilitating subsequent pattern clustering and analysis. A Graph Convolutional Network (GCN) is used to cluster these behavioral patterns, discovering and identifying common behavioral patterns or action chains in infants and toddlers. Using the constructed topological graph as input, the GCN automatically identifies relationships between behavioral patterns and clusters similar patterns. The GCN effectively recognizes and classifies patterns by considering the graph structure and the relationships between nodes. A topological feature matrix is ​​generated, including: action connection relationships: describing the transitions between different behavioral patterns. The system analyzes behavioral patterns, such as the transition from crawling to standing in infants; behavioral transition probability, representing the probability that an infant will transition from one behavioral pattern to another within a certain period; it analyzes the periodicity and frequency of infant behavioral patterns using time series data to create heatmaps showing the distribution of these patterns across different time periods; it divides continuous monitoring periods into preset time windows (time windows), such as 5 minutes or 10 minutes per window; it counts the activation frequency of nodes within each time window, i.e., the number of times an infant performs a particular behavioral pattern within that window; and it creates a behavioral pattern heatmap that shows the frequency distribution of behavioral patterns across different time periods. Furthermore, the data is timestamped to show the periodic changes in behavior; based on infants' behavioral patterns and location data, the behavioral distribution density in various spatial areas within the childcare center is analyzed; the childcare center is divided into several interactive units of preset sizes, each representing a spatial area; the percentage of dwell time is calculated: within each interactive unit, the percentage of dwell time for each infant's behavioral pattern is calculated, i.e., the proportion of time an infant spends in that spatial area under a certain behavioral pattern out of the total time; a spatial behavioral density cloud map is generated, which presents the behavioral density in different spatial areas, revealing the hot and cold zones of infants' behavior within the childcare center; behavioral data from both temporal and spatial dimensions are integrated to comprehensively display the behavioral distribution of infants within the childcare center;The behavioral pattern heatmap (showing behavioral distribution over time) obtained in the first two steps is merged with the spatial behavioral density cloud map (showing behavioral distribution over space); this yields a comprehensive behavioral distribution map. The final infant and toddler behavioral distribution map displays the behavioral distribution of infants and toddlers within the childcare facility, including the temporal periodicity and spatial distribution characteristics of behavioral patterns.

[0078] In an optional embodiment, the activity trajectory of the childcare center is extracted based on the distribution of infant and toddler behavior to obtain the activity trajectory of the infants and toddlers within the childcare center, including:

[0079] The childcare center is divided into functional areas, and a gridded spatial topology map is generated by combining the spatial behavior density cloud map of infant and toddler behavior distribution; the gridded spatial topology map is labeled with behavior preference tags.

[0080] Based on the dynamic time warping algorithm, spatiotemporal neighborhood matching of continuous trajectory points of infants and young children is performed through a gridded spatial topology map to obtain a spatiotemporal trajectory primitive sequence; wherein, the spatiotemporal trajectory primitive sequence includes behavioral pattern transfer links;

[0081] The spatiotemporal trajectory primitive sequence is context-aware encoded using a graph attention network to obtain a motivation trajectory map; the motivation trajectory map includes typical trajectory motivation patterns and behavioral intention labels.

[0082] The motivation trajectory map and the environmental data of the childcare center were multimodally aligned to obtain the spatiotemporal trajectory correlation matrix; the spatiotemporal trajectory correlation matrix includes the intensity of environmental interaction; the environmental data includes toy distribution and caregiver location.

[0083] Feature extraction is performed on the spatiotemporal trajectory correlation matrix using a spatiotemporal convolutional network to obtain the activity trajectory of infants and toddlers; the activity trajectory of infants and toddlers includes spatial movement rate, behavior transformation probability and environmental response delay.

[0084] It should be noted that, based on the behavioral distribution of infants and toddlers, the childcare center is divided into different functional areas, and a gridded spatial topology map is generated to mark the behavioral preferences of different areas. The space of the childcare center is divided into several functional areas, which may include rest areas, play areas, and interaction areas. Each area is defined according to the behavioral patterns of infants and toddlers to ensure that the area is suitable for their activities. Combined with the spatial behavioral density cloud map of the infant and toddler behavioral distribution, this analyzes which areas have high infant and toddler activity and which areas have low activity. This information helps to identify the behavioral preferences of infants and toddlers, thereby influencing the functional area division within the childcare center. A gridded spatial topology map is generated, where each grid represents a spatial unit, marked with... Behavioral preference tags identify the activity intensity and preferred behaviors of infants and toddlers in the area (e.g., a preference for activity, stillness, or socialization). Spatiotemporal neighborhood matching of continuous behavioral trajectory points of infants and toddlers extracts a spatiotemporal trajectory primitive sequence within the childcare center. Dynamic Time Warping (DTW), a technique for processing time-series data, performs non-linear alignment over time. DTW helps identify changes in infants' and toddlers' behavioral patterns in space across different time periods by matching continuous trajectory points in spatiotemporal neighborhoods. Based on these trajectory points, a spatiotemporal trajectory primitive sequence is generated, recording the migration process of infants' and toddlers' behavioral patterns. Specifically, these primitive sequences contain the changes in infants' and toddlers' behavioral patterns across different time periods from a single point... The transfer of behavioral patterns from one to another is known as the behavioral pattern migration link. The resulting spatiotemporal trajectory primitive sequence reveals the behavioral transfer process of infants and toddlers, such as the transition from a "play" mode to a "rest" mode. A motivational trajectory map of infants and toddlers is constructed by context-aware encoding of the spatiotemporal trajectory primitive sequence using a Graph Attention Network (GAT). A GAT is a graph neural network model that adaptively assigns weights to different nodes in the graph, thereby capturing the relationships between nodes in the spatiotemporal trajectory primitive sequence. GAT can effectively learn the spatiotemporal dependencies in the trajectory, enhancing the contextual information in the sequence by focusing on the important relationships between specific trajectory nodes. After encoding with GAT, a motivational trajectory map is obtained. This motivational trajectory map displays typical trajectory motivational patterns in infants and toddlers (e.g., infants and toddlers prefer to engage in certain activities within specific time periods) and corresponding behavioral intention markers, such as the infants' and toddlers' behavioral intentions (e.g., "play" or "quiet") and motivations (e.g., gaining attention, enjoying independent activities). The motivational trajectory map reveals the intrinsic motivations and intentions behind infants' and toddlers' behavior, helping to analyze the driving forces behind their behavior within the childcare facility. Multimodal alignment of the motivational trajectory map with real-world environmental data from the childcare facility establishes a correlation between infants' and toddlers' behavior and the environment. The real-world environmental data includes toy distribution (which areas have toys, and which toys infants and toddlers show greater interest in) and caregiver locations (the distribution of caregivers within the childcare facility and their interactions with infants and toddlers).By aligning motivational trajectory maps with real-world environmental data, a spatiotemporal trajectory correlation matrix is ​​obtained. This matrix illustrates the relationship between infants' and toddlers' behavioral patterns and their interactions with the environment, reflecting the role of the environment in their behavior. The spatiotemporal trajectory correlation matrix includes environmental interaction intensity, representing the intensity of interaction between infants / toddlers and different elements in the environment (such as toys and caregivers), helping to understand the environmental dependence of infants' and toddlers' behavioral patterns. Based on the spatiotemporal trajectory correlation matrix, features are extracted using a spatiotemporal convolutional network (ST-CNN) to analyze the infants' and toddlers' activity trajectories. The spatiotemporal convolutional network can simultaneously process spatial and temporal information, performing feature extraction within the spatiotemporal correlation matrix. This study aims to capture the complex relationship between infant and toddler behavior patterns and their environment. Through spatiotemporal convolutional layers, ST-CNN can effectively extract key features from spatiotemporal data to identify infant and toddler behavior patterns, activity trajectories, and environmental responses. The resulting infant and toddler activity trajectories contain the following information: spatial movement rate: the speed at which the infant moves within the childcare facility, reflecting the frequency and energy level of their activity; behavior transition probability: the probability of an infant transitioning from one behavior pattern to another, revealing the patterns of behavior pattern transition; and environmental response delay: the reaction time of the infant to changes in the environment, showing the infant's reaction speed and sensitivity during interaction.

[0085] In an optional embodiment, assessing the quality of an infant's activity based on their activity trajectory to obtain the quality of their activity includes:

[0086] The activity trajectories of infants and toddlers are analyzed to obtain activity intensity fingerprints, which include movement speed, trajectory curvature, and spatial residence heat.

[0087] The activity intensity fingerprint and the distribution of teaching aids in the childcare center were spatiotemporally aligned to obtain a behavioral sequence map; the behavioral sequence map is marked with cognitive stage markers; the cognitive stage markers include observation, manipulation and creation;

[0088] The frequency and duration of interactions between infants and toddlers and physical entities were statistically analyzed using behavioral sequence mapping to obtain a cognitive engagement matrix. The cognitive engagement matrix includes social initiative, operational focus, and exploration depth. Physical entities include peers, caregivers, and teaching aids.

[0089] The activity intensity fingerprint, behavioral sequence map, and cognitive engagement matrix are fused to obtain the activity quality of infants and toddlers.

[0090] It's important to note that the rate of movement of infants and toddlers within a childcare facility can reflect their activity level. For example, a faster movement speed may indicate a higher frequency of activity, suggesting exploration or social behavior. Trajectory curvature reflects the degree of curvature in an infant's movement trajectory and is typically used to indicate variability in movement; high curvature may mean the infant is exhibiting more exploratory behavior or showing greater attention to their surroundings. Spatial dwell heat indicates the frequency and duration an infant or toddler spends in a specific area, usually reflecting the area's attractiveness to them. For example, if an infant or toddler spends a long time in a particular toy area, it may indicate a higher level of interest in that area or that the area fulfills certain needs of the infant or toddler (such as emotional needs). (Support, games, etc.); By aligning activity intensity fingerprints with the distribution of teaching aids within the childcare center through spatiotemporal alignment, a behavioral sequence map of infants and toddlers is obtained; the behavioral sequence map reveals the sequence of infants and toddlers' activities and is marked according to their cognitive stages; by aligning the activity intensity fingerprints of infants and toddlers with the distribution data of teaching aids within the childcare center, behavioral patterns of infants and toddlers in different times and spaces are determined; the distribution of teaching aids within the childcare center can influence the types of activities and cognitive levels of infants and toddlers, for example, some teaching aids may promote manipulative behavior, while others may encourage creative activities; cognitive stage marking is based on the infants and toddlers' behavioral trajectories to mark their current cognitive stage; typical cognitive stages include: observation stage, where infants and toddlers mainly learn through observing the behavior of others or In the environment stage, infants learn and understand the world around them; activities at this stage are relatively static, possibly characterized by low movement speed and high spatial dwell time. In the manipulative stage, infants begin to actively explore and manipulate objects in their environment, such as toys, exhibiting higher movement speed and more frequent behavioral changes. In the creative stage, infants begin to apply existing knowledge and experience to new scenarios or creative activities, demonstrating more creative interactions and higher behavioral complexity. By analyzing the interactions between infants and environmental entities, their cognitive engagement is calculated. The cognitive engagement matrix reveals infants' social, manipulative, and exploratory performance in activities. The frequency of interactions between infants and various entities within the childcare facility (such as peers, caregivers, and teaching aids) is statistically analyzed. Interaction duration; for example, frequent interactions between infants and toddlers with peers indicate active social behavior; longer interactions with caregivers may indicate stronger dependence on caregivers; Cognitive engagement matrix: This matrix includes the following three dimensions: Social initiative: the frequency with which infants and toddlers actively interact with peers or caregivers, reflecting the initiative of their social behavior; Operational focus: the degree of focus of infants and toddlers when manipulating teaching aids or other objects, reflecting the depth of their participation in the exploration process; Exploration depth: the degree to which infants and toddlers explore different areas or objects in the environment, reflecting their interest in and exploration intensity of new environments and new things; By comprehensively analyzing the intensity of infants and toddlers' activities, behavioral patterns, and cognitive engagement, the quality of their activities can be fully assessed;This fusion process helps us better understand the overall developmental status of infants and toddlers; by integrating activity intensity fingerprints, behavioral sequence maps, and cognitive engagement matrices, a comprehensive activity quality assessment model is formed; through this multi-dimensional fusion analysis, we can reveal the comprehensive performance of infants and toddlers in spatial activities, cognitive development, and social interaction.

[0091] In an optional embodiment, childcare centers are categorized according to the quality of infant and toddler activity to obtain a childcare demand cycle benchmark, including:

[0092] The demand patterns of infants and toddlers are classified based on the quality of their activities using an unsupervised clustering algorithm to obtain a baseline of periodic behavior patterns. The baseline of periodic behavior patterns is labeled with demand type. Demand patterns include high-intensity physical activity, deep cognitive exploration, and socio-emotional connection.

[0093] The baseline of periodic behavior patterns and the schedule data of childcare centers are spatiotemporally aligned to obtain the matching deviation between standard procedures and the baseline of periodic behavior patterns. Based on the matching deviation, a heat map of the coupling degree between demand and supply is generated.

[0094] Time-series predictions are made based on the heatmap of demand-supply coupling using long short-term memory networks to obtain a demand response baseline curve. The demand response baseline curve includes a demand surge warning threshold, a supply redundancy critical point, and an intervention response window.

[0095] The demand response baseline curve and the standardized demand cycle template are integrated to obtain the childcare demand cycle benchmark; the childcare demand cycle benchmark includes real-time demand intensity and cyclical fluctuation patterns.

[0096] It should be noted that unsupervised clustering algorithms (such as K-means, DBSCAN, etc.) are used to classify the activity data of infants and toddlers; the goal is to identify different demand patterns. Common demand patterns include: High-intensity movement: Infants and toddlers are active and move quickly, usually involving physical activity, which may be exploration or interaction with peers; Deep cognitive exploration: Infants and toddlers concentrate on thinking, manipulating, and exploring in a relatively quiet environment, usually manifested as prolonged stay and in-depth manipulation of teaching aids; Social-emotional connection: Infants and toddlers engage in emotional interaction with caregivers or peers, manifested as close contact, communication, and interaction; Periodic behavior pattern baseline: After clustering, different demand types are obtained. The baseline of cyclical behavioral patterns is defined as the typical needs patterns exhibited by infants and toddlers at different times within the childcare center. This baseline is then spatiotemporally aligned with the childcare center's schedule data to identify the match between infants' and toddlers' needs and the center's supply. The childcare center's schedule includes activities at different times, such as morning opening activities, lunchtime, nap time, and playtime. The matching deviation between the baseline of cyclical behavioral patterns and the childcare center's schedule is calculated, i.e., whether the infants' and toddlers' needs are consistent with the childcare center's actual supply (activity schedule). For example, if an infant or toddler needs in-depth cognitive exploration at a certain time, but the childcare center schedules high-intensity physical activities, this may create a deviation. This study analyzes matching deviations to generate a heatmap of supply-demand coupling, showcasing the matching between supply and demand under different time periods and activity arrangements. This helps identify periods of insufficient or excessive supply. A Long Short-Term Memory (LSTM) network is used to perform time-series forecasting on the heatmap, aiding in predicting future demand changes and providing decision support for resource allocation in childcare centers. LSTM is a deep learning-based time-series data analysis model capable of learning and predicting trends in time-series data. Here, an LSTM network is used to perform time-series forecasting on the supply-demand coupling heatmap to predict future changes in infant and toddler demand. Based on the prediction results of the LSTM model, a demand response basis is obtained. A quasi-curve, which illustrates the time series of changes in infants' and toddlers' needs, may include important information such as: a surge in demand warning threshold: predicting when infants' and toddlers' needs might suddenly increase, indicating moments requiring special attention; for example, a surge in demand for high-intensity physical activity or deep cognitive exploration during a certain period; a supply redundancy threshold: indicating that childcare resources (such as teaching aids and activity arrangements) may become excessive after exceeding a certain threshold, potentially leading to resource waste or excessive passivity for infants and toddlers; and an intervention response window: the time window during which childcare facilities need to make timely adjustments when changes in demand occur; for example, when infants' and toddlers' needs suddenly increase, childcare facilities should adjust their activity arrangements within a certain timeframe to meet the demand.By integrating the demand response baseline curve with a standardized demand cycle template, a childcare demand cycle benchmark is generated, providing childcare centers with a systematic and real-time reference for demand forecasting and response. The standardized demand cycle template is a standard demand cycle based on historical data or theoretical analysis, typically including common demand patterns and fluctuations of infants and toddlers within a day or week at the childcare center. Combining the demand response baseline curve (the result of LSTM prediction) with the standardized demand cycle template forms the final childcare demand cycle benchmark. This benchmark provides a comprehensive analysis of childcare centers' needs for infants and toddlers, including: real-time demand intensity: the real-time intensity of infants' and toddlers' demand, indicating how childcare center resources need to be adjusted at the current time; and cyclical fluctuation patterns: the cyclical fluctuation patterns of infants' and toddlers' demand, helping childcare centers predict peak and trough periods of demand, thereby enabling advance resource allocation.

[0097] In an optional embodiment, environmental adaptation identification is performed on the quality of infants' activities based on real-time environmental data to obtain an environmental adaptation influence coefficient, including:

[0098] Real-time environmental data and the quality of infants' activities are synchronized to obtain a spatiotemporally aligned data chain; the spatiotemporally aligned data chain is labeled with environmental tags.

[0099] Semantic association is performed on the behavioral intensity fingerprint, cognitive engagement matrix, and real-time environmental data in the spatiotemporally aligned data chain to obtain a feature fusion tensor; wherein, the feature fusion tensor includes environmental and behavioral interaction patterns;

[0100] The adaptation curve between real-time environmental data and the quality of infant activity was determined based on the infant physiological tolerance database.

[0101] It's important to note that synchronizing real-time environmental data with infant activity quality establishes a complete data chain, enabling the analysis of the relationship between environmental changes and infant behavior. Real-time environmental data includes information such as temperature, humidity, air quality, noise, and light intensity, reflecting current environmental changes. Environmental data and infant activity quality data are aligned temporally and spatially, ensuring they match within the same timeframe. This step generates a "spatiotemporally aligned data chain" and labels these data with environmental tags to differentiate the impact of different environmental conditions on activities during subsequent analysis. Semantic association between behavioral and environmental data yields a feature fusion tensor containing environmental and behavioral interaction patterns, allowing for a deeper understanding of their relationship. Intensity fingerprints represent the intensity of infant activity within a specific timeframe, such as movement frequency and amplitude. A cognitive engagement matrix reflects the degree of cognitive activity participation within a given timeframe, typically correlated with attention span and exploration. The system includes: behavioral data; environmental data (including temperature, humidity, light intensity, etc.) that affect the quality of infants' activities; semantic association (via feature fusion, linking environmental data with infants' behavioral characteristics such as intensity and cognitive engagement); this process generates a feature fusion tensor containing the interaction patterns between environment and behavior, i.e., how environmental factors influence infants' behavioral patterns; adaptation curves between real-time and fixed-time environmental data and the quality of infants' activities, helping to understand how the environment affects the quality of infants' activities and their physiological tolerance; an infant physiological tolerance database containing data on infants' physiological responses under different environmental conditions, such as the impact of temperature and humidity on activity intensity and cognitive engagement; and adaptation curves based on the physiological tolerance database, which represent the trend of changes in the quality of infants' activities when environmental conditions change. This curve reflects how infants' activity performance adapts to different environments, helping to determine whether infants experience maladaptation under certain environmental conditions.

[0102] In an optional embodiment, the method further includes identifying the environmental adaptation of infants' activity quality based on real-time environmental data to obtain an environmental adaptation influence coefficient, and also includes:

[0103] Structural learning is performed on the feature fusion tensor to obtain causal links;

[0104] Among them, the causal link includes identifying sudden temperature changes, surges in exercise volume, and exceeding the fatigue index limit, and generating an environmental impact topology map with the direction of action based on the causal link;

[0105] The environmental impact topology map and the adaptation curve are weighted and fused to obtain the environmental fitness impact coefficient; the environmental fitness impact coefficient includes stimulus intensity, tolerance threshold and compensation demand.

[0106] It should be noted that, through structural learning, the causal relationship between environmental factors (such as temperature, exercise level, fatigue, etc.) and the quality of infants' activities is identified, further optimizing the activity arrangements for infants. Machine learning or statistical models (such as Bayesian networks, graph neural networks, etc.) are used to analyze the feature fusion tensor to learn the causal relationship between environment and behavior. This process will reveal the direct or indirect impact of various environmental factors (such as sudden temperature changes, surges in exercise, and exceeding fatigue limits) on the quality of infants' activities. Through structural learning, causal links are identified, i.e., which environmental changes (such as sudden temperature changes) directly or indirectly lead to changes in infants' activity behavior. Based on the analysis results of causal links, an environmental impact topology map with direction of action is generated. This map shows the impact path and direction of different environmental factors on the quality of infants' activities; for example, sudden temperature changes may lead to an increase in infants' fatigue index, further affecting cognitive engagement. The environmental impact topology map is then weighted and fused with the adaptation curve to obtain... A comprehensive environmental adaptability impact coefficient helps quantify the specific impact of environmental factors on the quality of infants' and toddlers' activities. This is achieved by weighted fusion of environmental impact topology maps and adaptation curves, linking environmental factors with the physiological responses and behavioral performance of infants and toddlers. The purpose of weighted fusion is to consider the comprehensive impact of different factors on the quality of infants' and toddlers' activities. The environmental adaptability impact coefficient, generated after fusion, includes: Stimulus intensity: the intensity of the impact of changes in environmental factors (such as temperature, humidity, and light) on the quality of infants' and toddlers' activities; for example, a sudden temperature change may have a significant impact on the intensity of infants' and toddlers' activities and cognitive engagement; Tolerance threshold: the limit of environmental conditions that infants and toddlers can adapt to; for example, the quality of an infant's and toddler's activities may begin to decline under certain temperature or humidity conditions, forming a tolerance threshold; Compensation needs: when environmental changes are significant, infants and toddlers may need additional support or compensation, such as adjusting activity intensity or increasing rest time to adapt to environmental changes.

[0107] In an optional embodiment, the childcare demand cycle benchmark is corrected for childcare quality environment based on the environmental adaptability impact coefficient to obtain the childcare environment quality, including:

[0108] The environmental adaptability impact coefficient and the childcare demand cycle benchmark are synchronized to obtain the demand cycle spectrum; the demand cycle spectrum is labeled with environmental intervention.

[0109] Environmental sensitivity analysis of the demand period spectrum is performed using an adaptive gated cyclic network to obtain an environmentally coupled demand correction curve, which includes an environmental demand attenuation mode.

[0110] Counterfactual inference is performed on the environment-coupled demand correction curve to obtain the environment correction weight matrix; the childcare quality environment is corrected based on the physiological tolerance characteristics of infants and young children and the environment correction weight matrix to obtain the childcare environment quality.

[0111] It should be noted that synchronizing the environmental adaptability impact coefficient with the childcare demand cycle of infants and toddlers generates a demand cycle spectrum to analyze the relationship between infants' and toddlers' needs and environmental changes. The environmental adaptability impact coefficient is a coefficient that quantifies the impact of environmental conditions on the quality of infants' and toddlers' activities, typically including parameters such as stimulus intensity, tolerance threshold, and compensatory needs, reflecting the infants' and toddlers' adaptation status in different environments. The childcare demand cycle baseline is usually the periodic pattern of the physiological and psychological needs of infants and toddlers during the childcare process, including their needs for sleep, eating, and activity, and is usually planned according to the infants' and toddlers' physiological development and activity patterns. The demand cycle spectrum synchronizes the environmental adaptability impact coefficient with the childcare demand cycle to obtain a... A new data sequence, called the "Demand Cycle Spectrum," illustrates the periodic changes in infants' and toddlers' needs under environmental variations. It also includes environmental intervention labels, indicating the specific impact of environmental factors (such as temperature and humidity) on infants' and toddlers' needs at different stages of demand. An Adaptive Gated Recurrent Network (AGCRN) is used to perform environmental sensitivity analysis on the demand cycle spectrum, exploring the sensitivity of infants' and toddlers' needs to environmental changes and obtaining an environmentally coupled demand correction curve. The AGCRN is an improved recurrent neural network that adjusts network weights through a "gating" mechanism, automatically learning and adapting to changes in input data. Here, the AGCRN is used to analyze the demand cycle spectrum. Analysis of the interaction between infants' and toddlers' needs and environmental factors; Environmental sensitivity analysis, using AGCRN, can identify which environmental factors have a significant impact on changes in the spectrum of infants' and toddlers' needs, thus analyzing their sensitivity to environmental changes; for example, certain environmental factors (such as temperature and noise) may have a significant impact on infants' and toddlers' sleep or activity needs; Environmentally coupled demand correction curves describe how infants' and toddlers' needs are corrected or diminished with environmental changes; for example, when the temperature is too high, infants and toddlers may need more rest time, leading to a decrease in their activity needs; this curve reveals the interaction between infants' and toddlers' needs and environmental conditions; understanding how infants' and toddlers' needs gradually decline over time under different environmental conditions. The environmental demand decay pattern is part of the environmentally coupled demand correction curve, representing the decay process of infants' and toddlers' needs under specific environmental conditions. For example, when the ambient temperature is consistently too high, infants' and toddlers' activity levels and cognitive engagement may decrease, and their needs will decay accordingly. The decay pattern helps predict and quantify the long-term impact of the environment on infants' and toddlers' activities. Counterfactual inference analyzes how infants' and toddlers' needs would change without certain environmental factors, further determining the weight of environmental corrections. Counterfactual inference is a reasoning method used to analyze how infants' and toddlers' needs would change under "hypothetical" conditions. For example, would the intensity of infants' and toddlers' activities be different without sudden temperature changes or excessive humidity?The environmental correction weight matrix, through counterfactual inference, determines the degree to which environmental factors correct for needs, thus deriving an environmental correction weight matrix. This matrix reflects the impact of each environmental factor (such as temperature, humidity, and light) on the needs of infants and toddlers at different stages of needs, and provides a basis for subsequent adjustments to childcare quality. Based on the physiological tolerance characteristics of infants and toddlers and the environmental correction weight matrix, environmental correction of childcare quality is performed, ultimately resulting in an optimized childcare environment quality scheme. Infant and toddler physiological tolerance characteristics include their physiological adaptability to different environmental conditions, such as their tolerance to different temperatures and humidity levels. By establishing a physiological tolerance model, the adaptability of infants and toddlers in specific environments is determined. Childcare quality environmental correction: Based on the environmental correction weight matrix and combined with the physiological tolerance characteristics of infants and toddlers, the settings of the childcare environment are adjusted. This includes optimizing environmental parameters such as temperature, humidity, and light to ensure that the needs of infants and toddlers are met while maintaining a comfortable and healthy childcare environment. Childcare environment quality: Ultimately, the childcare environment quality after environmental correction reflects the overall activity performance and health status of infants and toddlers under specific environmental conditions. The optimized childcare environment quality ensures that infants and toddlers receive appropriate care and development under optimal environmental conditions.

[0112] Example 2, please refer to Figure 2 This invention provides a technical solution: a data analysis-based method for assessing the quality of infant and toddler care, applicable to the aforementioned data analysis-based infant and toddler care quality assessment system, comprising:

[0113] S1. Obtain videos of the daily activities of infants and toddlers in the childcare center; perform behavioral identification on the videos to obtain behavioral characteristics of the infants and toddlers; perform behavioral statistics on the childcare center based on the behavioral characteristics of the infants and toddlers to obtain the behavioral distribution of the infants and toddlers in the childcare center.

[0114] S2. Extract activity trajectories from the childcare center based on the distribution of infant and toddler behavior to obtain the activity trajectories of infants and toddlers within the childcare center; assess the activity quality of infants and toddlers based on the activity trajectories to obtain the activity quality of infants and toddlers;

[0115] S3. Classify childcare centers according to the quality of infant and toddler activities to obtain a childcare demand cycle benchmark; obtain real-time environmental data of childcare centers; identify the environmental adaptation of infant and toddler activities based on the real-time environmental data to obtain the environmental adaptation influence coefficient.

[0116] S4. Adjust the childcare quality environment based on the childcare demand cycle benchmark according to the environmental adaptability influence coefficient to obtain the childcare environment quality; map the childcare service quality according to the childcare environment quality and the quality of infant and toddler activities to obtain the comprehensive childcare quality.

[0117] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A data analysis-based infant and toddler childcare quality assessment system, characterized in that... ,include: A behavior statistics unit (1) is used to acquire videos of the daily activities of infants and toddlers in the childcare center; to identify the behavior of the infants and toddlers based on the videos of the daily activities, so as to obtain the behavior characteristics of the infants and toddlers; wherein, the behavior characteristics of the infants and toddlers include the intensity of the action, the frequency of the behavior, and the interactive objects; to perform behavior statistics on the childcare center based on the behavior characteristics of the infants and toddlers, so as to obtain the behavior distribution of the infants and toddlers in the childcare center; Activity assessment unit (2), the activity assessment unit (2) is used to extract the activity trajectory of the childcare center according to the distribution of infant and toddler behavior, so as to obtain the activity trajectory of infant and toddler in the childcare center; wherein, the activity trajectory of infant and toddler includes spatial movement rate, behavior conversion probability and environmental response delay; and toddler activity quality is assessed according to the activity trajectory of infant and toddler, so as to obtain the activity quality of infant and toddler; An environmental identification unit (3) is used to classify the childcare center according to the quality of the infants' activities to obtain a childcare demand cycle benchmark; to obtain real-time environmental data of the childcare center; and to identify the environmental adaptation of the infants' activities according to the real-time environmental data to obtain an environmental adaptation influence coefficient; wherein the environmental adaptation influence coefficient includes stimulus intensity, tolerance threshold and compensation demand. The comprehensive evaluation unit (4) is used to correct the childcare quality environment based on the childcare demand cycle benchmark according to the environmental adaptability influence coefficient, so as to obtain the childcare environment quality; and to map the childcare service quality based on the childcare environment quality and the infant activity quality, so as to obtain the comprehensive childcare quality. Specifically, behavioral statistics are performed on the childcare center based on the infants' behavioral characteristics to obtain the distribution of infant behaviors within the childcare center, including: Threshold cleaning is performed on the infant and toddler behavioral characteristics to obtain a set of effective behavioral fragments; A behavior pattern topology graph is constructed based on the set of effective behavior fragments; the action chain is clustered using the behavior pattern topology graph through a graph convolutional network to obtain a topology feature matrix; wherein, the topology feature matrix includes action connection relationships and behavior transformation probabilities; The continuous monitoring period is divided into time windows of preset duration using time series binning technology. Within each time window, the node activation frequency of the behavior pattern topology map is counted to obtain a behavior pattern heatmap. The behavior pattern heatmap has a periodicity marked by timestamps. The childcare center is divided into interactive units of a preset size, and the dwell time ratio of behavioral patterns in each interactive unit is counted to obtain a spatial behavior density cloud map. The behavioral pattern heatmap and the spatial behavioral density cloud map are fused to obtain the distribution of infant and toddler behaviors.

2. The data analysis-based infant and toddler childcare quality assessment system according to claim 1, characterized in that... Based on the videos of the infants' daily activities, behavioral recognition is performed on the infants to obtain their behavioral characteristics, including: The daily activity video is divided into video frames to obtain a daily activity image sequence; the action segmentation network is used to identify the boundaries of infant limb movements through the daily activity image sequence to obtain action video clips; wherein, the action video clips include the action start frame, duration and spatial occupancy information; The motion video clips are localized using a 3D human pose estimation model to obtain a pose sequence; wherein, the pose sequence includes the 3D coordinates and motion trajectories of multiple key human joints. The preset basic behavior pattern library for infants and toddlers is matched with the posture sequence to obtain the behavior type label of the infants and toddlers; the action video clip, the posture sequence and the behavior type label are fused to obtain the behavior characteristics of the infants and toddlers.

3. The data analysis-based infant and toddler childcare quality assessment system according to claim 2, characterized in that, Based on the infant and toddler behavior distribution, activity trajectories of the childcare center are extracted to obtain the activity trajectories of infants and toddlers within the childcare center, including: The childcare center is divided into functional areas, and a gridded spatial topology map is generated by combining the spatial behavior density cloud map of the infant behavior distribution; wherein, the gridded spatial topology map is labeled with behavior preference tags; According to the dynamic time warping algorithm, the spatiotemporal neighborhood matching of continuous trajectory points of infants and young children is performed through the gridded spatial topology map to obtain a spatiotemporal trajectory primitive sequence; wherein, the spatiotemporal trajectory primitive sequence includes behavioral pattern transfer links; The spatiotemporal trajectory primitive sequence is context-aware encoded using a graph attention network to obtain a motivation trajectory map; wherein, the motivation trajectory map includes typical trajectory motivation patterns and behavioral intention markers; The motivation trajectory map and the environmental data of the childcare center are multimodally aligned to obtain a spatiotemporal trajectory correlation matrix; wherein, the spatiotemporal trajectory correlation matrix includes the intensity of environmental interaction; wherein, the environmental data includes toy distribution and caregiver location; The spatiotemporal trajectory correlation matrix is ​​used to extract features based on a spatiotemporal convolutional network to obtain the activity trajectory of infants and young children.

4. The data analysis-based infant and toddler childcare quality assessment system according to claim 3, characterized in that, The activity quality of the infants is assessed based on their activity trajectories to obtain the quality of their activities, including: The activity trajectory of the infant is analyzed for features to obtain an activity intensity fingerprint; wherein, the activity intensity fingerprint includes movement speed, trajectory curvature and spatial residence heat. The activity intensity fingerprint and the distribution of teaching aids in the childcare center are spatiotemporally aligned to obtain a behavioral sequence map; wherein the behavioral sequence map is marked with cognitive stage markers; wherein the cognitive stage markers include observation, manipulation and creation; The frequency and duration of interactions between infants and toddlers and entities are statistically analyzed based on the behavioral sequence map to obtain a cognitive engagement matrix; wherein, the cognitive engagement matrix includes social initiative, operational focus, and exploration depth; the entities include peers, caregivers, and teaching aids; The activity intensity fingerprint, the behavior sequence map, and the cognitive participation matrix are fused to obtain the activity quality of infants and toddlers.

5. The data analysis-based infant and toddler care quality assessment system according to claim 4, characterized in that: The childcare facilities are categorized into childcare demand cycles based on the quality of infant and toddler activities to obtain a childcare demand cycle benchmark, including: The demand patterns of infants and toddlers are classified according to the quality of their activities using an unsupervised clustering algorithm to obtain a baseline of periodic behavior patterns; wherein the baseline of periodic behavior patterns is labeled with demand type; the demand patterns include high-intensity physical activity, deep cognitive exploration, and socio-emotional connection. The periodic behavior pattern baseline and the childcare center's schedule data are spatiotemporally aligned to obtain the matching deviation between the standard process and the periodic behavior pattern baseline, and a demand-supply coupling degree heat map is generated based on the matching deviation. The demand-supply coupling heatmap is used to perform time-series prediction based on the Long Short-Term Memory Network to obtain the demand response baseline curve; wherein, the demand response baseline curve includes the demand surge warning threshold, the supply redundancy critical point, and the intervention response window period; The demand response baseline curve and the standardized demand cycle template are fused together to obtain the childcare demand cycle baseline; wherein, the childcare demand cycle baseline includes real-time demand intensity and cyclical fluctuation pattern.

6. The data analysis-based infant and toddler childcare quality assessment system according to claim 5, characterized in that, Based on the real-time environmental data, environmental adaptation is identified in the quality of the infants' activities to obtain an environmental adaptation influence coefficient, including: The real-time environmental data and the quality of infant activity are synchronized to obtain a spatiotemporally aligned data chain; wherein the spatiotemporally aligned data chain is labeled with an environment tag. Semantic association is performed on the behavior intensity fingerprint, the cognitive participation matrix, and the real-time environment data in the spatiotemporal aligned data chain to obtain a feature fusion tensor; wherein, the feature fusion tensor includes environment and behavior interaction patterns; The adaptation curve between the real-time environmental data and the quality of the infant's activities was determined based on the infant physiological tolerance database.

7. The data analysis-based infant and toddler childcare quality assessment system according to claim 6, characterized in that, Based on the real-time environmental data, environmental adaptation identification is performed on the quality of the infants' activities to obtain the environmental adaptation influence coefficient, which also includes: Structural learning is performed on the feature fusion tensor to obtain causal links; The causal link includes identifying sudden temperature changes, surges in exercise volume, and exceeding fatigue limits, and generating an environmental impact topology map with the direction of action based on the causal link; The environmental impact topology map and the adaptation curve are weighted and fused to obtain the environmental fitness impact coefficient.

8. The data analysis-based infant and toddler childcare quality assessment system according to claim 7, characterized in that, The childcare environment quality is adjusted based on the environmental adaptability influence coefficient to obtain the childcare demand cycle benchmark, including: The environmental adaptability impact coefficient and the childcare demand cycle benchmark are synchronized to obtain the demand cycle spectrum; wherein the demand cycle spectrum is labeled with environmental intervention. An environmental sensitivity analysis is performed on the demand period spectrum based on an adaptive gated cyclic network to obtain an environmentally coupled demand correction curve; wherein, the environmentally coupled demand correction curve includes an environmental demand attenuation mode. Counterfactual inference is performed on the environmentally coupled demand correction curve to obtain the environmental correction weight matrix; the childcare quality environment is corrected based on the physiological tolerance characteristics of infants and young children and the environmental correction weight matrix to obtain the childcare environment quality.

9. A data-based method for assessing the quality of infant and toddler care, applicable to the data-based infant and toddler care quality assessment system described in any one of claims 1-8, characterized in that, include: Obtain videos of the daily activities of infants and toddlers in childcare centers; The infants' behavior is identified based on the videos of their daily activities to obtain their behavioral characteristics. Behavioral statistics of the childcare center were performed based on the infant and toddler behavioral characteristics to obtain the distribution of infant and toddler behavior within the childcare center; Based on the distribution of infant and toddler behavior, the activity trajectory of the childcare center is extracted to obtain the activity trajectory of the infants and toddlers within the childcare center; The activity quality of the infants is assessed based on their activity trajectories to obtain the quality of their activities. The childcare center is classified into childcare demand cycles based on the quality of infant and toddler activities to obtain a childcare demand cycle benchmark; real-time environmental data of the childcare center is obtained. Based on the real-time environmental data, environmental adaptation is identified in the quality of the infants' activities to obtain the environmental adaptation influence coefficient. The childcare environment quality is corrected based on the environmental adaptability influence coefficient to obtain the childcare demand cycle benchmark; the childcare service quality is mapped based on the childcare environment quality and the infant activity quality to obtain the comprehensive childcare quality.

Images