Rehabilitation space adaptation method and system based on special child behavior dynamic recognition

Abstract

The invention discloses a rehabilitation space adaptation method and system based on special child behavior dynamic recognition, and the method comprises the steps: obtaining video resources related to special child rehabilitation training, carrying out the classification, and constructing a standardized video sample set; identifying skeleton nodes of a special child in the video through an AI motion capture tool, mapping the skeleton nodes to a standard character model, and exporting skeleton data; converting the skeleton data format into space coordinate point data, importing the space coordinate point data into a parameterized design tool, and extracting an activity volume and an activity position; the adaptation rules of different behavior states, behavior types and space sizes are analyzed, and data association of behavior characteristics, space layout and element sizes is established; and formulating a targeted rehabilitation space optimization scheme according to an adaptation rule. According to the method, AI video motion capture and Grasshopper parametric modeling technologies are integrated, so that accurate mapping of behavior data and spatial layout is realized, and the suitability and accuracy of spatial design are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of architectural design and rehabilitation engineering, and particularly relates to a rehabilitation space adaptation method and system based on dynamic identification of special children's behavior. BACKGROUND

[0002] At present, in the field of special children's rehabilitation medical buildings, the space design and practice still mainly rely on traditional architectural experience and perceptual judgment, and lack systematic "patient-centered" evidence-based design guidance. The existing research results mainly focus on the qualitative description and flexible layout of environmental elements such as color, material and furniture, and lack of in-depth analysis and design guidance based on objective behavior and environmental quantitative data, which leads to the need for improvement in the scientificity, adaptability and accuracy of the rehabilitation environment.

[0003] With the progress of technology, various new technologies and methods have emerged in the field of behavior observation and space analysis. For example, in the field of elderly health monitoring, some studies have deployed multi-sensor systems in living environments to unobtrusively monitor physiological, motion and environmental parameters, providing a new perspective for understanding home behaviors and neurological diseases. In the study of campus environment, by comparing the video observation data before and after greening, quantitative evidence is provided for the influence of environmental intervention on children's game behavior. In terms of research methods, computer programming simulation of human motion trajectory, interval image acquisition matrix method, video flow statistics based on MATLAB, etc. have been used to analyze behavior patterns, space use efficiency and passenger flow, promoting the development of behavioral architecture towards three-dimensionality and quantification. In addition, in the design for the elderly and passenger flow statistics in commercial space, the application of three-dimensional data extraction, behavior recognition and head target detection technologies further demonstrates the potential of data-driven design in meeting the needs of specific groups.

[0004] In particular, modern video motion capture and parametric modeling technologies provide powerful tools for in-depth understanding of the relationship between behavior and space. In the study of special children's behavior, motion capture systems have been used to quantitatively analyze the social symptoms of autistic children; platforms based on artificial intelligence (such as Plask AI) can even generate 3D motion data from ordinary 2D videos, greatly reducing the data collection threshold. In terms of data analysis and space modeling, parametric design platforms such as Grasshopper can process skeletal coordinate data, perform motion parameter calculation and posture simulation, and provide possibilities for the optimal design of built environments.

[0005] However, although the above technologies have made some progress, their integrated application in the field of special children's rehabilitation medical buildings still has significant limitations and defects: Technical coordination and data conversion barrier: Artificial intelligence video motion capture technology and Grasshopper parametric modeling technology have not achieved effective coordination and smooth data conversion in this field, making it difficult to establish a precise mapping relationship between the behavior characteristics of special children and the elements of the space environment.

[0006] Insufficient model universality and adaptability: AI-based motion capture technology lacks a standard character skeleton model suitable for most special children's body types. When facing different objects or needing cross-object uniform analysis and comparison, the real-time recognition ability and adaptability of the model decrease significantly, affecting the reliability and comparability of the data.

[0007] Lack of quantitative association between behavior data and space elements: Existing technologies can convert skeleton data into spatial coordinates and generate geometric shapes such as curves and surfaces, but the behavior space data (such as activity volume and trajectory) has not yet been associated with key design elements such as the specific layout, enclosure range, and furniture size of the rehabilitation space. This makes it difficult for behavior analysis results to directly guide precise space adaptation design.

[0008] The above problems make it difficult for current rehabilitation space design to dynamically respond to the real needs of special children in different emotional and behavioral states (such as normal state and attack state), and cannot achieve flexible adaptation of space, which may exacerbate the anxiety of children in the environment and affect the rehabilitation effect. Therefore, it is urgent to develop a systematic method that integrates cutting-edge behavior recognition and space modeling technology, targets the special behavior of special children, and establishes a quantitative conversion relationship from behavior data to space parameters, to improve the scientificity, adaptability, and humanistic care level of rehabilitation medical space design. SUMMARY

[0009] The purpose of the present application is to provide a rehabilitation space adaptation method and system based on dynamic identification of special children's behavior, to solve the problems of insufficient coordination, difficult data conversion, and lack of quantitative relationship between behavior and space elements in the field of special children's rehabilitation using AI video motion capture and Grasshopper parametric modeling technology. By integrating the two technologies, a rehabilitation space adaptation method based on dynamic behavior identification is established, achieving precise mapping of behavior data and space layout, and improving the adaptability and precision of space design.

[0010] To achieve the above purpose, the present application provides the following solutions: A rehabilitation space adaptation method based on dynamic identification of special children's behavior, comprising the following steps: S1, special children behavior video sample construction: obtain special children rehabilitation training related video resources, classify them by space type, behavior type, and behavior state, and construct a standardized video sample set; S2, behavior data extraction and space data conversion: based on the standardized video sample set, the AI motion capture tool is used to identify the skeleton nodes of the special children in the video, which are mapped to the standard character model and the skeleton data is exported; the Grasshopper parameterized modeling technology is used to convert the skeleton data format into spatial coordinate point data, which is imported into the parameterized design tool to extract the two core space indicators of activity volume and activity position; S3, behavior and rehabilitation space adaptation relationship establishment: based on the activity volume and activity position data, the adaptation rules of different behavior states, behavior types and space sizes are analyzed, and the data correlation between behavior characteristics and space layout, element size is established; S4, rehabilitation space adaptation design output: according to the adaptation rules, specific rehabilitation space optimization scheme is developed, including space enclosing range, furniture layout and size design content.

[0011] Further, the S1, special children behavior video sample construction, specifically includes: S101, video resource acquisition: special children related videos are obtained from professional rehabilitation service platform or through cooperation with rehabilitation institutions, the video captures the full body of the video object, the feet of the video object should be at the bottom of the picture, the video object is directly shot from the front angle, the video resolution is not less than 720p, and the time length is controlled within 3 seconds; S102, video object selection: multiple 4-6 year-old special children are selected, all objects come from the same rehabilitation institution, and a unified training system is adopted; S103, sample classification and labeling: the space types include large space 7.00m x 8.00m x 3.20m, medium space 4.00m x 6.00m x 3.00m, small space 3.00m x 4.00m x 2.80m and public space, the public space includes activity area 4.80m x 3.80m and corridor 1.80m wide; the behavior types include learning, playing toys and searching for objects; the behavior states include normal state and attack state; S104, sample set construction: the standardized sample set containing multiple behavior videos is formed according to the three-dimensional classification of space type, behavior type and behavior state.

[0012] Further, the S2, behavior data extraction and space data conversion, specifically includes: S201, standard character model construction: according to GBT 26158-2010 standard, the standard character model is constructed by using 4-6 year-old children nine five percentile P95 bone size parameters; S202, skeleton data extraction: the video is imported through Plask artificial intelligence tool, multiple skeleton nodes are identified and mapped to the standard character model, and BVH format animation data is exported; S203, data format conversion: use Python to call open source bvh-converter code, extract joint global position coordinates with ZYX Euler angle sequence, convert BVH format data to CSV format spatial coordinate point data; S204, spatial index extraction: Activity volume extraction: import CSV data through Grasshopper's LunchBox plug-in, read multiple data at 0.2 second intervals, select several key skeleton point connecting lines to form curves, and convert to three-dimensional spatial volume data using Dendro plug-in; Activity position extraction: through the plane view and the section view, the position change of special children in the horizontal and vertical directions is observed, and the activity position of special children is extracted.

[0013] Further, the S3, behavior and rehabilitation space adaptation relationship establishment, specifically includes: S301, spatial data difference analysis: in the attack state, the activity volume of special children is larger than that in the normal state, and the difference is most significant in the large space, and the learning behavior volume increases, the searching behavior volume decreases, the activity position is more dispersed and discontinuous, the searching and playing toy positions change significantly in the large and medium spaces, the learning position changes significantly in the small space, and the activity position is distributed along the wall and has higher dispersion in the public space; In the normal state, the activity volume satisfies searching>playing toy>learning; S302, behavior-space adaptation analysis: large space adapts to playing toys and searching, and does not adapt to learning; Medium space adapts to learning and training, and has the best stability; Small space adapts to playing toys and searching, and has strong activity concentration; Activity area adapts to game socialization, and corridor adapts to searching and searching training; S303, space design influence analysis: large space reserves central activity area and learning area away from door, medium space furniture is placed away from door and close to center, small space furniture is arranged close to door wall, and public space is designed with wall side safety area and corridor horizontal guide design.

[0014] Further, the S301, spatial data difference analysis, specifically includes: Activity volume difference: in the attack state, the activity volume of special children in various spaces is larger than that in the normal state, and the difference is most significant in the large space; In the normal state, the activity volume satisfies searching>playing toy>learning, in the attack state, the learning behavior volume increases, the searching behavior volume decreases, and the playing toy behavior volume does not change; In the public space, the activity area and the corridor in the attack state are higher than those in the normal state; Activity position difference: in the attack state, the activity position is more scattered, the distribution range is wider and discontinuous; in the large space and the medium space, the position change of searching for toys and playing toys is obvious, and in the small space, the position change of learning behavior is more prominent; in the public space, the activity position in the attack state is more inclined to be distributed along the wall, showing higher dispersity and escape behavior tendency, and the position change in the horizontal direction of the corridor is particularly significant.

[0015] Further, the S4, the rehabilitation space adaptive design output, specifically comprises: S401, training space design: large space adaptive play toys and searching behavior, medium space adaptive learning behavior, small space adaptive play toys and searching behavior; S402, public space design: the activity area is arranged near the wall to set a safe space to adapt to the behavior of children along the wall in the attack state; the corridor is strengthened to be horizontally guided to guide the children to move along the length direction; S403, furniture layout design: the furniture is arranged in the large space according to the learning area and the activity area, and the central area and the wall space away from the door are left in the design; the furniture is arranged in the medium space away from the door, and the furniture size is designed according to the learning behavior scale; the furniture is arranged in the small space close to the door, and the furniture size is designed according to the play toy and searching behavior scale.

[0016] The application also discloses a rehabilitation space adaptive system based on special child behavior dynamic identification, which is applied to the rehabilitation space adaptive method based on special child behavior dynamic identification. A special child behavior video sample construction module is used to acquire special child rehabilitation training related video resources, classify according to space types, behavior types and behavior states, and construct a standardized video sample set; A behavior data extraction and space data conversion module is used to identify the bone nodes of special children in the video based on the standardized video sample set through an AI motion capture tool, map to a standard character model and export bone data; the bone data format is converted into space coordinate point data by using a Grasshopper parameterized modeling technology, imported into a parameterized design tool, and two core space indexes of activity volume and activity position are extracted; A behavior and rehabilitation space adaptive relationship establishment module is used to analyze the adaptive rules of different behavior states, behavior types and space sizes based on the activity volume and activity position data, and establish the data correlation of behavior characteristics and space layout and element size; A rehabilitation space adaptive design output module is used to formulate a targeted rehabilitation space optimization scheme according to the adaptive rules, including space enclosing range, furniture layout and size design content.

[0017] Further, the special child behavior video sample construction module supports batch import of video resources, has an automatic classification and labeling function, and can generate a sample classification report according to three dimensions of space type, behavior type and behavior state.

[0018] Further, the behavior data extraction and space data conversion module has a built-in standard skeleton model library and supports automatic conversion of BVH and CSV formats.

[0019] Further, the behavior and rehabilitation space adaptation relationship establishment module integrates LunchBox and Dendro plug-in functions, and can output activity volume and position distribution data in real time, with synchronous data update frequency and sampling interval.

[0020] According to the technical solutions described above, compared with the prior art, the rehabilitation space adaptation method and system based on special child behavior dynamic identification provided by the application have the following beneficial effects: The application constructs a video behavior sample library of autistic children, acquires skeleton data through AI motion capture technology, converts the skeleton data into activity volume and activity position through Grasshopper, further analyzes the differences between normal state and aggressive state of autistic children in different spaces, and summarizes the behavior characteristics and space requirements of autistic children through in-depth analysis of these behavior actions.

[0021] The application establishes a data correlation between behavior motion skeleton data and rehabilitation space layout and element size, not only helps researchers understand the behavior patterns of autistic children, but also provides a scientific basis for the design of rehabilitation spaces; through data processing and analysis, it is found that in the aggressive state, the activity volume increases in all space sizes, especially in large spaces, the activity volume during learning significantly increases, but the activity volume for searching objects is significantly smaller than that in the normal state, in addition, compared with the normal state, the motion position of autistic children in the aggressive state is more dispersed; the motion position of the activity area and the corridor is closer to the wall, the motion for playing toys and searching objects overlaps more with furniture, the application provides a method for spatial adaptability intervention service for autistic children, and provides a reference for comparison with the current design standard.

[0022] The application provides a new idea and method for spatial design research of special groups. The application is beneficial to solve the spatial adaptation problems of special children such as autism, attention deficit disorder and language development retardation due to limited cognitive ability, abnormal perception and communication difficulties. By integrating video motion capture and Grasshopper interface space fitting analysis technology, the mapping relationship between behavior characteristics and space requirements of special children is quantified, which provides evidence-based design basis for optimization of spatial environment of special child medical rehabilitation centers, special education schools and related rehabilitation institutions.

[0023] The application aims to improve the adaptability of autism children rehabilitation space design, provide guidance for the space arrangement of autism children behavior training, and optimize the existing space environment to adapt to the behavior needs of autism children in the whole life cycle, and promote the construction and development of China's autism children whole-process service. BRIEF DESCRIPTION OF DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0025] Figure 1 It is a space category schematic diagram of the embodiment of the present application, wherein (a) is a large space, (b) is a medium space, (c) is a small space, (d) is an activity area, and (e) is a corridor. Figure 2 It is a behavior category schematic diagram of the embodiment of the present application. Figure 3 It is a behavior state schematic diagram of the embodiment of the present application. Figure 4 It is a video sample classification schematic diagram of the embodiment of the present application. Figure 5 It is a technical roadmap of video extraction of human behavior motion skeleton data of the embodiment of the present application. Figure 6 It is a technical roadmap of calculating and processing human behavior motion skeleton data of the embodiment of the present application. Figure 7 It is an extraction schematic diagram of activity volume of the embodiment of the present application. Figure 8 It is an extraction schematic diagram of activity position of the embodiment of the present application. Figure 9 The present application provides a rehabilitation space adaptation method based on special children behavior dynamic identification. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0027] The application is based on the behavior characteristics of special children in different states (normal, attack), interprets the space demand, establishes the data correlation between the behavior motion skeleton data and the rehabilitation space layout and the element size, and proposes a rehabilitation space adaptation method based on dynamic identification of special child behavior.

[0028] In order to make the above-mentioned purposes, characteristics and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below in combination with the drawings and specific embodiments.

[0029] As Figure 9 shown, the present application provides a rehabilitation space adaptation method based on dynamic identification of special child behavior, which comprises the following steps: S1, special child behavior video sample construction: obtaining special child rehabilitation training related video resources, classifying according to space type, behavior type and behavior state, and constructing a standardized video sample set; S2, behavior data extraction and space data conversion: based on the standardized video sample set, identifying the skeleton nodes of special children in the video through an AI motion capture tool, mapping to a standard character model and exporting skeleton data; using Grasshopper parametric modeling technology to convert the skeleton data format into space coordinate point data, importing into a parametric design tool, and extracting two core space indicators of active volume and active position; S3, establishment of the adaptation relationship between behavior and rehabilitation space: based on the active volume and active position data, analyzing the adaptation rules of different behavior states, behavior types and space sizes, and establishing the data correlation between behavior characteristics and space layout and element size; S4, rehabilitation space adaptation design output: according to the adaptation rules, formulating a targeted rehabilitation space optimization scheme, including space enclosing range, furniture layout and size design content.

[0030] In combination with specific embodiments, the specific description of each step of the above rehabilitation space adaptation method based on dynamic identification of special child behavior is as follows: 1. Construction of special child behavior video sample: In this scheme, in addition to obtaining videos through a website, special child behavior video samples can also be constructed by cooperating with rehabilitation medical institutions and children's parents, and tracking and recording videos of special child rehabilitation training processes. In addition, special child behavior data extraction can also be achieved by using motion capture devices and systems to directly collect relevant skeleton data according to research content.

[0031] Embodiments of the present application obtain video resources related to special children on professional websites such as APSPARK, autism rehabilitation service platform and autism home network. The video resolution is not less than 720p, the time length is controlled at about 3 seconds, the shooting angle is front, and the whole body movement of the object can be clearly captured. A total of 15 special children aged 4 to 6 are selected as video samples, which are classified according to space type, behavior type and behavior state. The space type includes training space (large, medium and small) and public space (activity area, corridor); the behavior type includes learning, playing toys and searching for objects, and the behavior state includes normal state and attack state. Finally, a sample set containing 330 behavior videos is constructed. The specific process is as follows: 1.1 Autism children behavior sample video acquisition The video sample set is obtained from professional websites, and the current professional websites such as APSPARK, autism rehabilitation service platform and autism home network provide highly relevant video resources for autism rehabilitation due to their professionalism and authority. In particular, APSPARK, established by the Hong Kong Autism Partnership, provides high-quality videos and applied behavior analysis methods, which are highly consistent with the needs of the present application. This study has obtained the video use and analysis rights of the APSPARK website.

[0032] In order to ensure the effective application of AI motion capture technology, the quality of video samples is crucial. The video samples should meet the following basic quality requirements: the video picture should be able to capture the whole body of the video object, ensuring the integrity and continuity of the motion; the feet of the video object should be located at the bottom of the picture, facilitating the determination of the starting point and ending point of the motion; the video object should be shot from a direct angle to capture the motion details more accurately; the video resolution should be at least 720p to ensure the clarity of the image; the video time length should be controlled at about 3 seconds to capture the complete motion period and ensure the efficiency of data processing.

[0033] After constructing a high-quality video sample set, it can be applied to motion analysis, rehabilitation program design and training effect evaluation, etc. Through AI motion capture technology, the motion of autism children is analyzed in detail, the behavior pattern is recognized, and individualized rehabilitation training programs are designed to improve the pertinence and effectiveness of rehabilitation training. At the same time, the effect of rehabilitation training is evaluated by using the video sample set, and the training program is adjusted in time to achieve the best rehabilitation effect.

[0034] The construction of a video sample set is a systematic project that requires careful consideration of the diversity of acquisition channels, the representativeness of samples, and the control of quality. A well-constructed video sample set will provide rich data support for autism children's rehabilitation research, promote the application of AI technology in the field of rehabilitation, and open up new avenues for the rehabilitation training of autistic children. With the continuous advancement of technology and the deepening of research, video sample sets will play an increasingly important role in the field of autism children's rehabilitation.

[0035] 1.2 Video classification of behavior categories of autistic children In the rehabilitation research of autistic children, the construction of a video sample set is a complex and meticulous process that involves in-depth analysis and classification discussion of various factors affecting children's behavior. First, the selection of subjects is considered, i.e., 4 to 6-year-old autistic boys, providing a specific group to observe changes in behavior. Second, the spatial category is considered, i.e., training space and public space, which has an undeniable influence on the training behavior patterns of autistic children. The size and layout of different spaces may stimulate or limit the behavior of children. In addition, behavior categories such as learning, object recognition (e.g., playing with toys), and object search are considered, reflecting the participation and reactivity of children in rehabilitation training. Furthermore, the distinction of behavior states, including normal state and attack state, reveals the behavior differences of autistic children in different psychological states. These classifications not only help to better understand the behavior characteristics of autistic children, but also provide scientific basis for the individualized design and effectiveness evaluation of rehabilitation training.

[0036] 1.3 Selection of video subjects In the common period of rehabilitation treatment for autistic children, i.e., 4 to 6 years old, selecting appropriate video subjects is the first step of the research. Autistic children in this age group usually have difficulty expressing themselves in language, so the collection of behavior data is particularly important. In addition, relevant research shows that autistic boys have more obvious changes in behavior than girls, which provides a more explicit direction for observation and analysis. Referring to the effect size in previous research, the total sample size of 15 cases is calculated using G*Power 3.1 software. This calculation is based on Cohen's effect size standard, with the following parameter settings: effect size f=0.25, significance level α=0.05, and statistical power 1-β=0.80. Based on these considerations, the present embodiment selects 15 autistic boys aged 4 to 6 years as video subjects. These children come from the same rehabilitation training institution, and they complete the course in a standardized training room according to a unified training system, which provides consistency and comparability for the research.

[0037] 1.4 Basis for video classification (1) The influence of space categories. In the autism rehabilitation center, the core space for carrying out behavioral activities is the training space. In addition, the activity area and the corridor are also indispensable parts of the space in the center. In the activity area, autistic children mainly carry out games and social activities, while the corridor is used to perform training tasks such as route finding and object finding. The design of these spaces aims to provide a comprehensive behavioral intervention and skill training environment for autistic children. The rehabilitation center has various functional rooms, such as training rooms, classrooms, and multi-purpose rooms. The size, layout, and environmental factors of these rooms can influence the behavior of autistic children. Among them, the training space is divided into three categories according to the size of the room: large, medium, and small. Through video observation and measurement, the specific dimensions of the three spaces are determined: large space is 7.00 meters long, 8.00 meters wide, and 3.20 meters high; medium space is 4.00 meters long, 6.00 meters wide, and 3.00 meters high; small space is 3.00 meters long, 4.00 meters wide, and 2.80 meters high. In addition, the positions of doors and furniture are also recorded in detail according to the video information. The public space is divided into: activity area 4.80 meters long, 3.80 meters wide; corridor 1.80 wide. The specific space category diagram is shown in Figure 1 , Figure 1 The specific space category diagram is shown in

[0038] (2) Classification of behavior categories. The behaviors of autistic children can be divided into various types, which are crucial for understanding their rehabilitation training effects. According to the 0-6-year-old children's autism screening intervention service specification, the training behaviors involved in the video are classified. These behavior categories include learning, object recognition (playing with toys), and object finding. By classifying and counting the training behaviors of autistic children in different size spaces, the three most frequent training behaviors are finally selected as the basis for classification. The behavior category diagram is shown in Figure 2 .

[0039] (3) Differentiation of behavior states. The behaviors of autistic children differ significantly in different states, and aggressive behavior is a frequently occurring behavior characteristic. This type of aggressive behavior may manifest itself in the form of behavioral outbursts, which include aggressive physical actions and self-harming behaviors. These behaviors have a larger movement range and can more significantly affect the design of the space, making them suitable as classification elements for this study. Behaviors in the defensive state are restricted, and in this state, the behavior exhibits inhibition and stability. This stable behavior performance has the same space requirements as normal behavior. Therefore, in the classification of video samples, the behavior in the defensive state is combined with the normal behavior as the normal state, but special attention is paid to distinguishing the behavior in the aggressive state. The behavior state diagram is shown inFigure 3 .

[0040] 1.5 Classification of the sample set The classification of video samples is an important foundation for the study of autistic children's behavior. According to the behavior state of autistic children, video samples are divided into two categories: normal state and attack state. The purpose of this dichotomy is to capture the behavioral differences of autistic children in different psychological states, thereby providing more abundant data for research. For example, in the normal state, autistic children may exhibit more social interaction and learning participation, while in the attack state, they may exhibit more anxiety or escape behavior.

[0041] Further, it is divided into rehabilitation training room, activity area and corridor. According to the size of the room where the video is taken, the sample is subdivided into three subcategories: large space, medium space and small space. This subdivision not only considers the potential impact of the physical environment on the behavior of autistic children, but also provides a framework for evaluating how space size affects the behavior of autistic children.

[0042] After determining the behavior state and space size, further video samples of autistic children under these different conditions are collected, which complete the learning, object recognition and object search behaviors. The selection of these three behaviors is based on their importance and universality in the rehabilitation training of autistic children. Learning behavior reflects the cognitive ability of autistic children, object recognition behavior tests their memory and recognition ability, and object search behavior examines their problem-solving ability and spatial perception.

[0043] Through this classification method, not only can we more comprehensively understand the behavioral patterns of autistic children in different environments, but also can we discover how different behavior states and space sizes interact and affect the behavior of autistic children. Ultimately, this research method forms a sample set containing 330 behavioral videos, as shown in Figure 4 .

[0044] 2. Extraction of special children behavior data and conversion of spatial data: The Plask artificial intelligence tool is used to accurately identify the bone nodes of special children in the video samples and map them to the standard character model. The BVH format animation model is derived to record the detailed information of the video object. Then, the BVH format bone data is converted into CSV format spatial coordinate point data using Python. Subsequently, with the help of the parametric design tool Grasshopper and its plug-in LunchBox, the motion data is converted into a spatial dimension data set, and the data is carefully processed, including selecting specific capture time points and bone points: the bone data is read at intervals of 0.2 seconds, a total of 15 times, and 13 key points are carefully selected from the 24 bone points, including leg, arm, neck and other bone points, forming 15 bone point curves. Next, these curves are used to create lofted surfaces, and the motion of special children is converted into visual three-dimensional spatial data. Finally, the Dendro plug-in of Grasshopper is used to convert the 15 bone point curve data into three-dimensional spatial volume data, and the activity volume of special children is extracted; through the perspective of plane and section, the position change of special children in the horizontal and vertical directions is observed, and the activity position of special children is extracted.

[0045] After the construction of the video sample set is completed, the embodiment of the application enters the key stage of extracting the behavior and motion bone data of autistic children. In order to achieve this goal, the Plask advanced artificial intelligence tool is adopted, which is specially used for video motion capture and can import motion data into 3D resources to create animation. Through the Plask platform, the bone nodes of autistic children in the video can be accurately identified, and the captured behavior trajectory is transmitted to the standard model of the platform. This standard model not only simulates the behavior and motion trajectory of children, but also embodies the strict protection of the privacy of video objects by simulating rather than directly using the original data of children. Finally, the behavior and motion data of each bone node in the standard model are output, which provides important quantitative information for further analyzing the behavior patterns of autistic children and the effects of rehabilitation training, and also ensures the ethics of the research and the privacy rights and interests of children.

[0046] 2.1 Software sets the size data of the standard model of the human body In constructing the standard human body model, this study strictly adhered to ergonomic principles to ensure the scientific validity and applicability of the model parameters. Based on the standard "Anthropometric Dimensions of Chinese Minors GBT 26158-2010," particular attention was paid to the skeletal dimensions of children aged 4 to 6. To meet the scale requirements of most children while considering individual differences, the 95th percentile (P95) dimensional parameters were used for modeling. This method not only covers the anthropometric dimensions of the vast majority of children but also provides appropriate fit for children with special body types. This approach ensures a balance between the universal applicability of the standard model and individualized needs, providing a more accurate and comprehensive reference framework for the behavioral and motor analysis of children with autism. The platform's standard model dimensional parameters are shown in Table 1.

[0047] Table 1. Platform Standard Model Dimensions

[0048] 2.2 Extracting skeletal data of human behavior and movement from video. In the data processing phase of the autism children's behavior research, the first step was to import the videos from the video sample set into the Plask platform. This step marked the substantial beginning of the research. The Plask platform, with its advanced motion capture technology, successfully extracted the children's behavioral movements from the videos and mapped them onto a standard persona model. This not only completed the uploading of videos and the recognition of behavioral movements but also enabled the effective application of the standard persona model. Subsequently, the behavioral motion data of the standard persona model was exported. This data was presented in the form of motion skeletons, providing detailed motion information of the video subjects.

[0049] There was a format incompatibility issue between the motion skeleton data exported from the Plask platform and the planned subsequent analysis software. To overcome this obstacle, it was decided to leverage the powerful programming capabilities of Python to perform the necessary format processing on the motion skeleton data. A custom script was written to convert the skeleton data into spatial coordinate point data, a format that perfectly matches the input requirements of the subsequent software. This conversion involved not only adjusting the data structure but also optimizing data accuracy and format specifications, ensuring the accuracy and integrity of the data during the conversion process. The technology roadmap is as follows: Figure 5 As shown, converting skeletal data into spatial coordinate point data is a key step in this technical process. Specifically, in Python, the open-source bvh-converter code provided on GitHub is used to process the skeletal motion data exported from the Plask platform. This code extracts the global position coordinates of the joints (using the "ZYX" Euler angle sequence), thereby converting the BVH format motion capture data into CSV format, bridging the gap for subsequent applications in Grasshopper software.

[0050] 2.3 Calculating and processing skeletal data for human behavior and movement After successfully converting the spatial coordinate point data into a format suitable for subsequent analysis, the next stage is data processing and analysis. The core task of this stage is to import this coordinate point data into Grasshopper, a powerful parametric design tool that can transform action-based data into datasets with spatial dimensions.

[0051] To achieve this goal, the LunchBox plugin was used, which efficiently imports spatial coordinate point data files into the Grasshopper environment. This data is organized in a table format, where each row represents a captured motion time point, and each column corresponds to the coordinate data of different skeletal points. The dataset contains coordinate information for 24 skeletal points (as shown in Table 1), covering key parts of the child's body.

[0052] In Grasshopper, the data structure was meticulously processed first. This included selecting specific capture time points and skeleton points to ensure the accuracy and representativeness of the data. Data was read at 0.2-second intervals, for a total of 15 reads, ensuring the continuity and integrity of motion capture.

[0053] Next, 13 key points were carefully selected from the 24 skeletal points (such as...). Figure 6 As shown, these points include skeletal points in areas such as the legs, arms, and neck. The selection of these points was based on their importance in motion analysis and their contribution to describing movement patterns. Connecting these skeletal points in the order of the body's contours created 15 skeletal point curves.

[0054] Ultimately, these curves are used to create lofted surfaces, a technique that connects a series of curves in a specific order and according to certain rules to form a continuous surface. For example... Figure 6 As shown, this method can transform the movements of children with autism into visualized three-dimensional spatial data, providing strong support for further movement analysis and pattern recognition.

[0055] 2.4 Transformation of behavioral spatial data of children with autism 2.4.1 Extraction of active volume The next phase of research into the behavior of children with autism explored the impact of room size and behavior type on children's behavioral performance, as well as the spatial volume differences in behavior between normal and aggressive states. To achieve this, an innovative approach was employed, collecting and analyzing corresponding motion data by varying room dimensions and behavior types within a simulated environment. This process involved not only in-depth analysis of motion capture data but also the generation and comparison of spatial volume data.

[0056] The method route is shown as Figure 7 The Dendro plugin is used to convert the 15 curves of each group into mash, which is an approximate volume geometry. The Dendro plugin has a unique algorithm that can convert curve data into a three-dimensional volume, providing an intuitive way to observe and quantify the behavior space of autistic children under different conditions. Through reading and calculating these mash volumes, the conversion from spatial coordinate point data to action-based spatial volume data is achieved, which provides a new perspective for studying the spatial characteristics of children's behavior.

[0057] Further, the spatial volume data under the attack state is compared and analyzed with the data under the normal state. This comparison reveals the significant differences in the behavior of autistic children under two different psychological states, providing important clues for understanding the behavior adaptability of autistic children in different environments. Through this method, not only can the key environmental factors affecting the behavior of autistic children be identified, but also the effectiveness of different rehabilitation training environments can be evaluated.

[0058] 2.4.2 Extraction of activity position In the in-depth analysis of autistic children's behavior, action-based lofted surfaces are used to observe and reflect the behavior trends of children in the room. This surface not only captures the complexity of children's actions, but also intuitively presents the behavior patterns and trends of autistic children through their spatial positions in the room. Through this method, it can be explored how room size and behavior type jointly affect the behavior of autistic children.

[0059] In order to further reveal the spatial differences in children's behavior between normal and attack states, the spatial position data under the two states is superimposed and compared. This comparative analysis helps to identify the spatial variation trends of autistic children's behavior under different psychological states, providing a new perspective for understanding the adaptability and reactivity of children to environmental changes.

[0060] During the observation process, two different perspectives are adopted: planar perspective and cross-sectional perspective. The planar perspective mainly reflects the position relationship of children in the horizontal direction, and can observe the horizontal activity range and movement path of children in the room. The cross-sectional perspective reveals the position changes of children in the vertical direction, including standing, sitting or other vertical actions. The combination of these two perspectives provides a comprehensive spatial analysis framework, which can understand the behavior characteristics of autistic children from different dimensions.

[0061] Through this multi-angle, multi-dimensional spatial position analysis, as Figure 8The results show that the proposed method can not only accurately evaluate the behavior patterns of autistic children, but also provide scientific basis for the design and adjustment of rehabilitation training environment. The application of this analysis method not only enriches the understanding of the spatial characteristics of autistic children's behavior, but also provides new strategies and directions for personalized rehabilitation training of autistic children.

[0062] 3. Relationship between rehabilitation space and behavior: Based on the special children's spatial data information (activity volume and activity location) extracted from video samples, the behavior types and behavior states of special children and the adaptability of rehabilitation space are analyzed. The analysis is mainly from the following three aspects: first, the difference analysis of spatial data under different behavior states. The activity volume of autistic children in the attack state is generally larger than that in the normal state, especially in large spaces. In terms of behavior type, the activity volume in the normal state is in the order of searching > playing toys > learning, while in the attack state, the learning behavior volume increases significantly, the searching behavior decreases, and the playing toys do not change significantly. In terms of activity location, the activity location form in the attack state is more dispersed, the distribution range is wider and not continuous. The location of searching and playing toys changes significantly in large and medium spaces, while the location of learning behavior changes more significantly in small spaces. In public spaces, the activity area and the movement in the corridor increase in the attack state, and the activity location tends to be distributed along the wall, showing higher dispersity and escape behavior tendency. Second, the behavior and space adaptation under activity volume difference analysis. Large spaces are not suitable for learning training in the attack state, but are suitable for playing toys and searching; medium spaces are best in maintaining learning stability and are suitable for learning training; small spaces have higher adaptation degree in playing toys and searching behavior, which helps to improve the behavior concentration and stability. Overall, large spaces are not recommended as the main training space, medium spaces are suitable for learning, and small spaces are more suitable for playing toys and searching. In public spaces, the activity area is suitable for games and socialization, and the corridor is suitable for searching and searching training due to its linear structure. Third, the influence of behavior on space under activity location difference analysis. In large spaces, the central area should be reserved for playing toys and searching, and the wall area far from the door is suitable for arranging learning space; in medium space design, furniture should be placed far from the door and close to the center to support stable learning behavior; in small spaces, furniture should be placed close to the door and wall to promote the concentrated performance of playing toys and searching activities. In addition, safety spaces should be set up at the wall of the activity area, and the corridor should be strengthened in the horizontal direction to meet the behavior needs and emotional regulation of children in the attack state. The specific analysis process is as follows: 3.1 Difference analysis of spatial data under different behavior states 3.1.1 Differences in activity volume In this study, we investigated the changes in the activity volume of experimental subjects in different-sized rooms under attack and normal conditions. The results showed that under attack conditions, the activity volume of experimental subjects in three different-sized rooms generally increased, especially in large spaces. This may be related to the tendency of experimental subjects to increase their activity range to cope with environmental changes or find safe areas when facing non-normal environmental stress.

[0063] However, despite the overall increase in activity volume in large spaces under attack conditions, it was observed that the activity volume of certain behavior types actually decreased. Under normal conditions, the size order of activity volume was searching greater than playing with toys greater than learning. This may reflect the differences in space requirements for different behavior types, with searching behavior requiring larger space for exploration, while learning may be conducted in smaller spaces.

[0064] Under attack conditions, the activity volume of learning behavior increased significantly, which may indicate that experimental subjects become more active in learning behavior when facing stress or environmental changes, possibly to better adapt to the environment or find problem-solving methods. Conversely, the activity volume of searching behavior decreased under attack conditions, which may mean that the exploration behavior of experimental subjects is inhibited in a stressful environment, or they tend to seek smaller spaces for safety.

[0065] The activity volume of playing with toys did not change significantly under attack conditions, which may indicate that playing with toys is less sensitive to environmental changes, or that the interest of experimental subjects in toys does not change significantly under this condition.

[0066] It is worth noting that in large spaces, the difference in activity volume of different behavior types under normal conditions is significant, which may be related to the space requirements of experimental subjects for different behaviors in normal environments. However, under attack conditions, this difference is no longer significant, which may indicate that in a stressful environment, the behavior patterns of experimental subjects tend to be consistent, or their space requirements become more flexible and variable.

[0067] In the activity area and the corridor, the activity amount of autistic children in both spaces under attack conditions was higher than that under normal conditions, showing the regulatory effect of space environment on behavior. The corridor, with its narrow linear characteristics, may encourage autistic children to perform higher activity amounts when performing specific training tasks. While the activity area, as the main place for games and social interaction, the increase in movement may be related to the diverse activities performed by autistic children in this area, including not only physical movement but also social interaction, etc.

[0068] 3.1.2 Differences in activity location Overall, the activity locations in the aggressive state showed more dispersed, widely distributed, and discontinuous patterns. This phenomenon may be related to changes in the individual's perception of the environment and behavioral responses in the aggressive state. In the aggressive state, individuals may be more inclined to explore unknown areas, or due to the influence of emotions such as anxiety and unease, leading to an expansion of the activity range and dispersion of activity patterns.

[0069] (1) Comparison between large and medium spaces: In large and medium spaces, the activity locations of searching and playing with toys in the aggressive state changed significantly, which may be related to the space and flexibility required by these behaviors. In these larger spaces, individuals have more freedom to explore and move, so in the aggressive state, this tendency to explore and move may be more pronounced. In contrast, learning activities may be less sensitive to space size due to their need for higher concentration and stability, so the changes in the aggressive state may not be as obvious as searching and playing with toys.

[0070] (2) Special case analysis of small spaces: In small spaces, the location of learning activities in the aggressive state changed more significantly, which is in contrast to the situation in large and medium spaces. The limitations of small spaces may force individuals to adjust their activity locations when performing learning activities to adapt to the limitations of the space. In addition, due to the limited range of activities in small spaces, any changes in behavior patterns may be more noticeable. At the same time, playing with toys and searching activities may not show similar significant changes in small spaces due to space limitations compared to large and medium spaces.

[0071] (3) Comparison between activity areas and corridors: Activity locations in the aggressive state show higher dispersion, especially in activity areas, where aggressive autistic children tend to be distributed along the walls, while in the normal state, they are relatively concentrated. In contrast, the spatial utilization patterns of aggressive children in the corridor are similar to those in the activity area, but the horizontal spatial location changes are more significant, which may be related to escape behavior. These differences reflect the specific needs of autistic children for space utilization in different emotional states.

[0072] From these results, we can see that the size of the space and the changes in the individual's state have a complex interaction on the activity location. Future research can further explore the specific performance of different behavior types in different space conditions and the psychological and physiological mechanisms behind these performances. In addition, research can also consider individual differences, environmental factors, and other factors that may affect activity location to gain a more comprehensive understanding.

[0073] 3.2 Activity volume difference analysis of behavior and space adaptation Considering the results of activity volume and activity location, we can discuss the adaptation of three training behaviors to three space sizes. This provides important spatial adaptation guidance for behavior training of autistic children.

[0074] (1) Adaptability analysis of large space: In a large space environment, it is found that the learning activity volume of autistic children significantly increases in the attack state, indicating that a large space may not be the ideal choice for learning training. The openness of large space may make children more easily distracted in the attack state, leading to decreased learning efficiency. However, for the object search training, large space seems to provide more flexibility and exploration, which helps to reduce the activity volume in the attack state. In addition, large space performs well in the adaptability of playing with toys behavior, with no significant difference in activity volume between normal and attack states, showing higher adaptability. However, in terms of activity location, the overlap of object search and playing with toys in the attack state is lower than that in the normal state, indicating that large space may not be conducive to the stability of activity location. Considering the results of activity volume and activity location, we can conclude that large space is not suitable for the three behavior training of autistic children.

[0075] (2) Adaptability analysis of medium space: In a medium-sized space, the activity volume of the three behaviors in the attack state has no significant difference compared with the normal state, indicating that medium space has good adaptability in activity volume for behavior training of autistic children. Especially in learning behavior, the overlap of activity location in the attack state and the normal state is high, indicating that medium space is conducive to maintaining the stability and continuity of learning activity. However, for playing with toys and object search behavior, the overlap of activity location in the attack state and the normal state is low, which may mean that medium space has certain limitations in the stability of these behaviors. However, the adaptability of medium space in learning training of autistic children is still high.

[0076] (3) Adaptability analysis of small space: In a small space environment, the activity volume of the three behaviors in the attack state also has no significant difference compared with the normal state, showing the adaptability of small space in activity volume. Especially in playing with toys and object search behavior, the overlap of activity location in small space is high, the form is more concentrated, and the distribution range is smaller, indicating that small space is conducive to improving the stability and concentration of these behaviors. The limitations of small space may prompt children to pay more attention when performing these activities, reducing unnecessary interference. Therefore, small space has high adaptability for playing with toys and object search.

[0077] In summary, while large spaces may exhibit some adaptability in certain behaviors, they are not overall suitable for the behavior training of autistic children. Medium spaces have advantages in maintaining the stability of learning activities and are suitable for learning training. Small spaces show higher adaptability in playing with toys and searching for objects. Future research can further explore the specific effects of different spatial characteristics on the behavior of autistic children and how to optimize the design of space according to individual differences and behavior needs. At the same time, it is recommended to choose appropriate space environment according to the specific needs and behavior characteristics of autistic children in actual training to improve the training effect and the adaptability of children.

[0078] Activity area and corridor adaptability analysis: The design of activity areas and corridors should promote the amount of children's movement. Activity areas are suitable for games and social interaction, while the linear characteristics of corridors are suitable for route-finding and object-finding training. These spatial layouts play an important role in regulating children's behavior and help promote their overall rehabilitation and development.

[0079] 3.3 Activity position difference analysis of the influence of behavior on space This study explores the behavior characteristics of autistic children in different spatial environments and the influence of these behaviors on spatial design elements. This provides important guidance for the design of education and training environments for autistic children.

[0080] (1) Influence of spatial element design in large spaces. In large space environments, autistic children tend to choose positions close to walls and furniture when learning, which may be because these areas provide them with more safety and stability. However, when playing with toys and searching for objects, they tend to choose positions in the center of the room, away from the walls, which may be because the central area provides a more open view and activity space. When playing with toys, they tend to stay away from furniture, possibly to reduce interference, while when searching for objects, they interact with furniture, indicating that furniture plays an auxiliary role in object-finding activities. Therefore, the design of large spaces should consider leaving a central area for autistic children to play with toys and search for objects. At the same time, space should be left on the walls away from the door for them to conduct learning activities. In addition, the size of furniture should be designed according to the behavior scale of autistic children when learning and searching for objects to provide appropriate support and interaction.

[0081] (2) The influence of space element design in the middle space. The middle space environment is more suitable for the learning and training of autistic children. In the middle space, they tend to choose a position away from the door and close to the furniture for learning, and interact more with the furniture. This shows that furniture plays an important role in their learning process. Therefore, the design of the middle space should place furniture away from the door to facilitate the learning activities of autistic children. Furniture can be placed in the center of the room to provide more opportunities for interaction. The size of the furniture should be designed according to the behavior scale of autistic children learning in the middle space to ensure that they can use the furniture comfortably for learning.

[0082] (3) The influence of space element design in the small space. The small space environment is more suitable for the toy playing and object searching training of autistic children. In the small space, they tend to choose a position close to the door and furniture for these activities, and interact with the furniture. This shows that the position and size of furniture are crucial for their activities in the small space. Therefore, the design of the small space should place furniture close to the door to facilitate their toy playing and object searching activities. Furniture should be placed according to the wall position to maximize the use of space. The size of the furniture should be designed according to the behavior scale of autistic children playing toys and searching for objects in the small space to ensure that they can comfortably carry out activities in limited space.

[0083] Considering the behavior characteristics of autistic children in different space sizes, the following conclusions can be drawn: large space is more suitable for toy playing and object searching activities, but needs to consider leaving a central area and wall space away from the door in the design; middle space is more suitable for learning and training, and furniture should be placed away from the door in the design, and the size of the furniture should be designed according to the learning behavior scale; small space is more suitable for toy playing and object searching training, and furniture should be placed close to the door in the design, and the size of the furniture should be designed according to the toy playing and object searching behavior scale.

[0084] The influence of space element design in the activity area and corridor: the activity area should consider setting safe space near the wall to adapt to the behavior of children distributing along the wall in the attack state. The corridor should be designed to guide children to move along the length to meet their escape behavior needs in the aggressive emotional state. The layout of these space elements helps to meet the space utilization needs of autistic children in different emotional states.

[0085] 4. Rehabilitation space adaptive design output: Training space design: large space for toy playing and object searching behavior, middle space for learning behavior, small space for toy playing and object searching behavior; Public space design: set safe space near the wall in the activity area to adapt to the behavior of children distributing along the wall in the attack state; strengthen the horizontal guidance design of the corridor to guide children to move along the length; Furniture layout design: furniture is arranged in learning area and activity area in large space, and central area and wall space away from the door are left in design; furniture is placed in position away from the door in medium space, and furniture size is designed according to learning behavior scale; furniture is placed in position close to the door in small space, and furniture size is designed according to toy playing and object searching behavior scale.

[0086] In summary, the method uses multi-source technical platforms such as Plask artificial intelligence tools, phython and Grasshopper, breaks through the difficulty of effective conversion of skeleton data and spatial data, establishes a complete technical process, realizes dynamic identification of special children's behavior, converts skeleton data into spatial data, fits and adapts behavior and rehabilitation space, and outputs the whole process.

[0087] The method breaks through the technical bottleneck of conversion difficulty between artificial intelligence and architectural design software, so that the rehabilitation space design no longer depends on existing standards and traditional experience, but uses scientific quantitative indicators to analyze the influence of space factors on behavior, realizes the precise adaptation of special children's behavior and rehabilitation space, and uses the rehabilitation space adaptation method of special children's behavior dynamic identification, which is beneficial to establish design reference standards suitable for the special behavior of special children, identify and improve the shortcomings of the current design standards, and provide design basis for the layout and furniture size of the rehabilitation space suitable for the behavior of special children.

[0088] Using this method, starting from the behavior results of special children, the space demand behind the behavior is excavated, the relationship between behavior and rehabilitation space is established, the rehabilitation space design method suitable for the behavior of special children is proposed, the use efficiency and psychological comfort of the rehabilitation space are improved, and the anxiety reaction of children is reduced.

[0089] The present application studies the relationship between behavior and space, establishes a video sample library to capture and analyze the behavior of autistic children in the rehabilitation space, extracts activity volume and activity position data, analyzes the adaptation degree of space size and training behavior, and establishes the relationship between behavior data and space composition and interface scale, so as to propose optimization design suggestions for the rehabilitation space.

[0090] The present application can provide beneficial reference and guidance for the space and interface design of the rehabilitation environment of autistic children, and promote the construction and development of the whole service of autistic children in China. At the same time, it is expected that this paper will provide a micro-level supplement to the current research method based on typology and other backgrounds from the behavior research of autistic children. In addition, this research also aims to open up new ideas for the research of similar attribute space or the same type of architectural space in the future.

[0091] The research method combining video motion capture with Grasshopper interface space fitting analysis can be further popularized, and provides a computer simulation method for the related research on "behavior motion capture-space interface" in the field of architecture.

[0092] In other embodiments, the application also provides a rehabilitation space adaptation system based on dynamic identification of special child behaviors, which is applied to the rehabilitation space adaptation method based on dynamic identification of special child behaviors and comprises: A special child behavior video sample construction module is configured to acquire special child rehabilitation training related video resources, classify the video resources according to space types, behavior types and behavior states, and construct a standardized video sample set. The special child behavior video sample construction module supports batch import of video resources, has an automatic classification and labeling function, and can generate a sample classification report according to three dimensions of space types, behavior types and behavior states. A behavior data extraction and space data conversion module is configured to identify the bone nodes of special children in videos based on the standardized video sample set through an AI motion capture tool, map the bone nodes to a standard character model and export bone data, convert the bone data format into space coordinate point data by using a Grasshopper parameterized modeling technology, import the parameterized design tool, and extract two core space indexes of active volume and active position. The behavior data extraction and space data conversion module has a built-in standard bone model library and supports automatic conversion between BVH and CSV formats. A behavior and rehabilitation space adaptation relationship establishment module is configured to analyze the adaptation rules of different behavior states, behavior types and space sizes based on the active volume and active position data, and establish a data correlation between behavior characteristics and space layout and element size. The behavior and rehabilitation space adaptation relationship establishment module integrates LunchBox and Dendro plug-in functions, can output active volume and position distribution data in real time, and synchronizes data update frequency and sampling interval. A rehabilitation space adaptation design output module is configured to develop a targeted rehabilitation space optimization scheme according to the adaptation rules, including space enclosing range, furniture layout and size design content.

[0093] In this paper, specific examples are applied to explain the principles and implementation modes of the application, and the above examples are only used to help understand the method and core idea of the application; meanwhile, for those skilled in the art, according to the idea of the application, the specific implementation mode and application range will be changed. In view of the above, the content of the specification should not be understood as a limitation of the application.

Claims

1. A rehabilitation space adaptation method based on dynamic behavioral recognition of children with special needs, characterized in that, Includes the following steps: S1, Construction of video samples of children with special needs: Obtain video resources related to rehabilitation training for children with special needs, classify them according to spatial type, behavior type, and behavior status, and construct a standardized video sample set; S2, Behavioral Data Extraction and Spatial Data Transformation: Based on a standardized video sample set, AI motion capture tools are used to identify the skeletal nodes of special children in the video, map them to a standard human model, and export the skeletal data. Grasshopper parametric modeling technology was used to convert the skeletal data format into spatial coordinate point data, which was then imported into the parametric design tool to extract two core spatial indicators: activity volume and activity position. S3, Establishing the Adaptation Relationship between Behavior and Rehabilitation Space: Based on activity volume and activity location data, analyze the adaptation patterns between different behavioral states, behavioral types and space size, and establish data correlation between behavioral characteristics and spatial layout and element dimensions; S4, Rehabilitation Space Adaptation Design Output: Based on adaptation principles, develop targeted rehabilitation space optimization plans, including the space enclosure range, furniture layout and size design.

2. The rehabilitation space adaptation method based on dynamic behavioral recognition of children with special needs as described in claim 1, characterized in that, S1, the construction of video samples of special children's behaviors, specifically includes: S101, Video Resource Acquisition: Obtain videos related to children with special needs from professional rehabilitation service platforms or in cooperation with rehabilitation institutions. The video should capture the whole body of the video subject, with the subject's feet at the bottom of the frame. Choose a direct frontal angle to shoot the video subject. The video resolution should be no less than 720p, and the duration should be controlled within 3 seconds. S102, Video Subject Selection: Select multiple children with special needs aged 4-6 years old. All subjects come from the same rehabilitation institution and adopt a unified training system. S103, Sample Classification Labeling: Spatial types include large spaces (7.00m×8.00m×3.20m), medium spaces (4.00m×6.00m×3.00m), small spaces (3.00m×4.00m×2.80m), and public spaces, including activity areas (4.80m×3.80m) and corridors (1.80m wide); behavioral types include learning, playing with toys, and finding objects; behavioral states include normal and aggressive states. S104, Sample set construction: A standardized sample set containing multiple behavioral videos is formed by classifying the data according to three dimensions: spatial type, behavioral type, and behavioral state.

3. The rehabilitation space adaptation method based on dynamic behavioral recognition of children with special needs as described in claim 2, characterized in that, S2, behavioral data extraction and spatial data transformation, specifically includes: S201, Standard Character Model Construction: Based on the GBT 26158-2010 standard, a standard character model is constructed using the 95th percentile P95 skeletal dimension parameters of children aged 4-6 years. S202, Skeletal Data Extraction: Import the video using the Plask AI tool, identify multiple skeletal nodes and map them to a standard human character model, and export BVH format animation data; S203, Data Format Conversion: Using Python to call the open-source bvh-converter code, the global position coordinates of the joints are extracted using the ZYX Euler angle sequence, and the BVH format data is converted into CSV format spatial coordinate point data; S204, Spatial Index Extraction: Activity volume extraction: Import CSV data using Grasshopper's LunchBox plugin, read the data multiple times at 0.2-second intervals, select several key skeletal points and connect them to form a curve, then use the Dendro plugin to convert it into three-dimensional spatial volume data; Activity location extraction: By observing the positional changes of children with special needs in both horizontal and vertical directions from planar and cross-sectional perspectives, the activity locations of children with special needs can be extracted.

4. The rehabilitation space adaptation method based on dynamic behavioral recognition of children with special needs as described in claim 3, characterized in that, The S3 section, establishing the adaptation relationship between behavior and rehabilitation space, specifically includes: S301, Spatial Data Difference Analysis: Under attack conditions, the activity volume of special children is larger than that under normal conditions. This difference is most significant in large spaces. The volume of learning behavior increases, while the volume of retrieval behavior decreases. The activity positions are more dispersed and discontinuous. The positions for retrieval and toy playing are significantly different in large and medium spaces, while the positions for learning are significantly different in small spaces. The activity positions in public spaces are distributed along the walls and are more dispersed. Under normal conditions, the activity volume satisfies the order of retrieval > toy playing > learning. S302, Behavior-Spatial Adaptation Analysis: Large spaces are suitable for playing with toys and finding objects, but not for learning; medium spaces are suitable for learning and training, with the best stability; small spaces are suitable for playing with toys and finding objects, with strong activity concentration; activity areas are suitable for games and social interaction, and corridors are suitable for pathfinding and finding object training. S303, Spatial Design Impact Analysis: Large spaces should reserve a central activity area and a learning area away from the door; medium-sized spaces should have furniture placed away from the door and close to the center; small spaces should have furniture arranged close to the door and wall; and public spaces should have a wall-side safety zone and a horizontal corridor guidance design.

5. The rehabilitation space adaptation method based on dynamic behavioral recognition of children with special needs as described in claim 4, characterized in that, S301, spatial data difference analysis, specifically includes: Differences in activity volume: Children with special needs in aggressive states have a larger activity volume in various spaces than in normal states, with the most significant difference in large spaces; in normal states, activity volume follows the order of finding objects > playing with toys > learning; in aggressive states, the volume of learning behavior increases, the volume of finding objects decreases, and the volume of playing with toys remains unchanged; in public spaces, the amount of movement in activity areas and corridors is higher in aggressive states than in normal states. Differences in activity location: During an attack, activity locations are more dispersed, wider in range, and discontinuous; the location changes for finding objects and playing with toys are significant in large and medium-sized spaces, while the location changes for learning behaviors are more prominent in small spaces; in public spaces, during an attack, activity locations tend to be distributed along the walls, showing a higher degree of dispersion and a tendency to avoidance behaviors, with the horizontal location changes in corridors being particularly significant.

6. The rehabilitation space adaptation method based on dynamic behavioral recognition of children with special needs as described in claim 5, characterized in that, The S4, the rehabilitation space adaptation design output, specifically includes: S401, Training Space Design: Large space is suitable for playing with toys and finding objects, medium space is suitable for learning behavior, and small space is suitable for playing with toys and finding objects. S402, Public Space Design: Activity areas are designed with safe spaces near walls to accommodate children's behavior of spreading along the walls during an attack; corridors are designed with enhanced horizontal guidance to guide children to move along the length of the space. S403, Furniture Layout Design: In large spaces, furniture is arranged in separate learning and activity areas, with a central area and wall space away from the door in the design; in medium-sized spaces, furniture is placed away from the door, and the furniture size is designed according to the scale of learning behavior; in small spaces, furniture is placed near the door, and the furniture size is designed according to the scale of playing with toys and finding objects.

7. A rehabilitation space adaptation system based on dynamic behavioral recognition of children with special needs, characterized in that, The rehabilitation space adaptation method based on dynamic recognition of special children's behaviors as described in any one of claims 1-6 includes: The special needs children's behavior video sample construction module is used to acquire video resources related to rehabilitation training of special needs children, classify them according to spatial type, behavior type and behavior status, and construct a standardized video sample set; The behavioral data extraction and spatial data conversion module is used to identify the skeletal nodes of special children in videos based on standardized video sample sets using AI motion capture tools, map them to standard human models and export skeletal data; Grasshopper parametric modeling technology is used to convert the skeletal data format into spatial coordinate point data, import it into parametric design tools, and extract two core spatial indicators: activity volume and activity position. The module for establishing the adaptation relationship between behavior and rehabilitation space is used to analyze the adaptation patterns between different behavioral states, behavioral types and space sizes based on activity volume and activity location data, and to establish data associations between behavioral characteristics and spatial layout and element dimensions. The rehabilitation space adaptation design output module is used to develop targeted rehabilitation space optimization plans based on adaptation principles, including the space enclosure range, furniture layout and size design.

8. The rehabilitation space adaptation system based on dynamic behavioral recognition of children with special needs as described in claim 7, characterized in that, The special children's behavior video sample construction module supports batch import of video resources and has an automatic classification and labeling function, generating sample classification reports according to three dimensions: spatial type, behavior type, and behavior status.

9. The rehabilitation space adaptation system based on dynamic behavioral recognition of children with special needs according to claim 7, characterized in that, The behavioral data extraction and spatial data conversion module has a built-in standard skeletal model library and supports automatic conversion between BVH and CSV formats.

10. The rehabilitation space adaptation system based on dynamic behavioral recognition of children with special needs according to claim 7, characterized in that, The module for establishing the relationship between behavior and rehabilitation space integrates the functions of LunchBox and Dendro plugins, and can output activity volume and location distribution data in real time, with data update frequency and sampling interval synchronized.

Images