Special child exercise intervention information interaction method and system based on cloud platform

Abstract

The invention relates to the technical field of information interaction processing, and discloses a special child exercise intervention information interaction method and system based on a cloud platform. The method comprises the following steps: acquiring expression, sound, physiology and behavior data of a special child through a multi-source sensing device, and performing feature extraction and time synchronization processing; performing hierarchical analysis and fusion on the data to form an interaction strength data set; accordingly, self-adaptive screening and dynamic configuration are carried out on the training content, and personalized training is achieved; real-time interaction matching is realized through multi-modal analysis; the data security is ensured by adopting distributed encryption storage; finally, multi-dimensional correlation analysis is carried out, and intervention effect evaluation is achieved. According to the method, real-time acquisition, secure storage and personalized analysis of multi-dimensional data are realized, so that accurate adjustment and effect evaluation of a training scheme are supported.

Description

Technical Field

[0001] The present application relates to the field of information interaction processing, and in particular to a special children's sports intervention information interaction method and system based on a cloud platform. Background Art

[0002] With the growing demand for rehabilitation of special children, exercise intervention has been widely used as an important rehabilitation method. Existing exercise intervention methods are mainly carried out through manual observation and evaluation, with the help of wearable devices, smart sensors and other technical means to collect performance data of special children during exercise. At the same time, the development of cloud computing technology provides technical support for remote rehabilitation guidance and data analysis, making data collection, storage and analysis in the process of exercise intervention more convenient and efficient. In particular, in terms of multi-source data fusion, real-time monitoring and personalized training program formulation, relevant technologies have made significant progress.

[0003] However, existing technologies still have obvious deficiencies in the interaction of sports intervention information. First, traditional data collection methods often only focus on single-dimensional data and cannot effectively integrate multi-dimensional information such as sports performance, emotional state, and interactive feedback. Secondly, the data analysis process lacks in-depth consideration of the individual characteristics of special children, making it difficult to accurately adjust the training program. Thirdly, the data security and privacy protection mechanisms are relatively weak and cannot meet the special requirements of rehabilitation data management for special children. In addition, most existing systems use fixed evaluation criteria, which are difficult to adapt to the development characteristics and rehabilitation needs of different special children. Summary of the invention

[0004] The present application provides a method and system for interacting with sports intervention information for special children based on a cloud platform, which realizes real-time collection, secure storage and personalized analysis of multi-dimensional data, thereby supporting precise adjustment of training programs and effect evaluation.

[0005] In the first aspect, the present application provides a method for interactive sports intervention information for special children based on a cloud platform, and the method for interactive sports intervention information for special children based on a cloud platform includes: performing time-series acquisition and processing of expression data streams, sound data, physiological data, and behavioral data of special children's sports processes through multi-source sensing equipment, and performing feature extraction and time synchronization processing on the multi-source data to obtain a multi-dimensional training data stream; performing hierarchical analysis and processing on the interaction parameters according to the multi-dimensional training data stream, and fusing the millisecond-level response data, minute-level adjustment data, and day-level optimization data to obtain an interaction intensity data set; based on the multi-dimensional training data stream and the interaction intensity data set, adaptively screening the training content in the sports project library, and integrating the training difficulty parameters, training rhythm Dynamically configure the parameters and training duration parameters to obtain a personalized training data packet; based on the interaction intensity data set and the personalized training data packet, perform multimodal analysis and processing on the real-time motion data of special children, and match the movement standardization data, emotional change data and interactive feedback data in real time to obtain a collaborative interaction data stream; encrypt and store the multidimensional training data stream, the collaborative interaction data stream and the personalized training data packet through distributed nodes, and configure the data access rights in a hierarchical manner to obtain a security management database; based on the collaborative interaction data stream and the security management database, perform multidimensional correlation analysis on the motor ability data, emotional state data and interaction effect data, and quantitatively evaluate the analysis results to obtain an intervention effect data set.

[0006] In a second aspect, the present application provides a special children's sports intervention information interaction system based on a cloud platform, and the special children's sports intervention information interaction system based on a cloud platform includes: The acquisition module is used to perform time-series acquisition and processing of the expression data stream, sound data, physiological data and behavioral data of the special children's movement process through multi-source sensing equipment, perform feature extraction and time synchronization processing on the multi-source data, and obtain a multi-dimensional training data stream; An analysis module, used to perform hierarchical analysis on the interaction parameters according to the multi-dimensional training data stream, and to fuse the millisecond-level response data, minute-level adjustment data, and day-level optimization data to obtain an interaction intensity data set; A screening module, for adaptively screening the training contents in the sports project library based on the multi-dimensional training data stream and the interactive intensity data set, dynamically configuring the training difficulty parameter, the training rhythm parameter and the training duration parameter, and obtaining a personalized training data packet; A processing module, used to perform multimodal analysis and processing on the real-time motion data of special children according to the interaction intensity data set and the personalized training data packet, and to match the action standardization data, the emotion change data and the interaction feedback data in real time to obtain a collaborative interaction data stream; An encryption module is used to encrypt and store the multidimensional training data stream, the collaborative interaction data stream and the personalized training data packet through distributed nodes, and to configure data access permissions in a hierarchical manner to obtain a security management database; The association module is used to perform multi-dimensional association analysis on the athletic ability data, emotional state data and interaction effect data based on the collaborative interaction data stream and the security management database, and to quantitatively evaluate the analysis results to obtain an intervention effect data set.

[0007] In the technical solution provided by this application, the facial expressions, sounds, physiological and behavioral data of special children during exercise are collected and extracted in an all-round time series through multi-source sensing equipment, which realizes the accurate perception and quantitative representation of the movement state of special children, and provides a rich data basis for subsequent personalized intervention. Based on the multi-dimensional training data stream, hierarchical analysis and processing are carried out, and the millisecond-level response data, minute-level adjustment data and day-level optimization data are integrated to build a hierarchical interaction intensity evaluation system, which can accurately grasp the immediate needs and long-term development trends of special children from different time scales. On this basis, the solution uses an adaptive screening algorithm to intelligently screen the sports project library, dynamically optimizes and configures parameters such as training difficulty, rhythm and duration, realizes accurate matching and personalized adjustment of training content, and effectively improves the pertinence and adaptability of sports intervention. At the same time, the solution also establishes a real-time interaction mechanism based on multimodal analysis, which matches and coordinates the action standardization, emotional changes and interactive feedback data in real time, which can not only timely discover and respond to the abnormal state of special children during training, but also dynamically adjust the training strategy to ensure the continuous optimization of training effect. In terms of data security, the solution uses distributed nodes for encrypted storage and implements fine-grained access control, effectively ensuring the security and privacy of data related to special children. Finally, the solution builds a complete intervention effect evaluation system. Through multi-dimensional correlation analysis and quantitative evaluation, it can objectively reflect the actual effect of exercise intervention and provide a scientific basis for the continuous improvement of training programs. Overall, this solution innovatively applies artificial intelligence algorithms in the field of sports intervention for special children, including adaptive screening algorithms, multimodal analysis algorithms, and collaborative interaction algorithms. The organic combination of these algorithms not only improves the intelligence level and personalization of sports intervention, but also significantly enhances the real-time and accuracy of the intervention process, providing more scientific and efficient technical support for the rehabilitation training of special children. Through the deep integration and optimized application of algorithm features, the solution successfully solves the problems of insufficient personalization, poor real-time performance, and weak security in traditional sports intervention methods, greatly improving the overall effect of sports intervention for special children. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0009] Figure 1 This is a schematic diagram of an embodiment of a method for interacting with special children's sports intervention information based on a cloud platform in an embodiment of the present application; Figure 2 A timing diagram of performing hierarchical analysis and processing on interaction parameters according to a multi-dimensional training data stream in an embodiment of the present application; Figure 3 This is a schematic diagram of an embodiment of a special children's sports intervention information interaction system based on a cloud platform in an embodiment of the present application. DETAILED DESCRIPTION

[0010] The embodiment of the present application provides a special children's sports intervention information interaction method and system based on a cloud platform. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0011] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiment of the present application, an embodiment of the special children's sports intervention information interaction method based on the cloud platform includes: Step S101, using multi-source sensing equipment to perform time-series acquisition and processing of expression data streams, sound data, physiological data, and behavioral data during the movement of special children, extracting features from the multi-source data and performing time synchronization processing to obtain a multi-dimensional training data stream; Step S102: performing hierarchical analysis on the interaction parameters according to the multi-dimensional training data stream, fusing the millisecond-level response data, minute-level adjustment data, and day-level optimization data to obtain an interaction intensity data set; Step S103: Based on the multidimensional training data stream and the interactive intensity data set, adaptively screen the training content in the sports project library, dynamically configure the training difficulty parameter, the training rhythm parameter and the training duration parameter, and obtain a personalized training data packet; Step S104: Based on the interaction intensity data set and the personalized training data set, multimodal analysis and processing are performed on the real-time motion data of the special children, and the action standardization data, the emotion change data and the interaction feedback data are matched in real time to obtain a collaborative interaction data stream; Step S105, encrypt and store the multidimensional training data stream, the collaborative interaction data stream and the personalized training data packet through distributed nodes, and configure the data access rights in a hierarchical manner to obtain a security management database; Step S106: Based on the collaborative interaction data stream and the security management database, a multi-dimensional correlation analysis is performed on the athletic ability data, the emotional state data, and the interaction effect data, and the analysis results are quantitatively evaluated to obtain an intervention effect data set.

[0012] It is understandable that the execution subject of the present application can be a special children's sports intervention information interaction system based on a cloud platform, or a terminal or a server, which is not limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0013] Specifically, multi-dimensional data of special children during exercise are collected through multi-source sensing equipment. Among them, the expression data stream is collected by a high-speed camera at a sampling rate of 60fps, and the feature points of the collected facial image sequence are annotated to extract the changing features of facial expressions. Then, the frequency domain analysis is performed through discrete Fourier transform to obtain the temporal change data of facial micro-expression features. At the same time, the sound data is collected through an 8-channel circular microphone array. After the collected sound signal is processed by noise reduction and signal enhancement, the acoustic feature parameters such as pitch, speaking speed, and volume are extracted. Physiological data collects physiological signals such as heart rate variability and skin galvanic response through smart bracelets, and data segmentation and feature extraction are performed according to fixed time windows. The behavioral data uses a depth camera to collect three-dimensional spatial coordinate information during the movement process, and spatial mapping and trajectory analysis of the joint point position are performed. These multi-source data are synchronized and aligned in time and feature fused to form a multi-dimensional training data stream.

[0014] Based on the acquired multidimensional training data stream, the interaction parameters are analyzed and processed in layers. First, real-time response analysis is performed at the millisecond level. The sliding window is used to detect sudden states such as emotional fluctuations and abnormal movements to generate millisecond response data; data is accumulated at the minute level and the trend of state changes such as training participation, emotional stability, and movement standardization is analyzed to generate minute-level adjustment data; at the day level, long-term performance indicators such as training effects, emotional change rules, and physical condition are summarized and analyzed to generate day-level optimization data. The data at these three levels are integrated into an interaction intensity data set through weighted fusion. Subsequently, the multidimensional training data stream and the interaction intensity data set are used to screen and configure the training content. First, the sports performance data in the training data stream is evaluated for ability, matched with the training items in the sports project library, and suitable alternative training items are screened out. Then, a secondary screening is performed based on the emotional parameters in the interaction intensity data set to ensure that the training items match the emotional state. The difficulty coefficient of the screened training items is calculated and the difficulty level is quantified to generate the training difficulty parameter; the action rhythm is adjusted according to the reaction speed indicator to generate the training rhythm parameter; the training duration is divided based on the continuity indicator to generate the training duration parameter. These parameters are integrated to form a personalized training data package.

[0015] During the training process, real-time analysis is performed based on the interaction intensity data set and the personalized training data packet. The spatial deviation of the real-time collected action data and the standard action is calculated, and the action standardization is evaluated in combination with the training difficulty parameter; the real-time emotional index is compared with the training rhythm parameter to analyze the trend of emotional changes; the action standardization is scored in real time and paired with the emotional change data to generate interaction status data. These data are time-aligned and correlated to form a collaborative interaction data stream. To ensure data security, a distributed storage architecture is used to manage the data. The multi-dimensional training data stream, collaborative interaction data stream, and personalized training data packet are encrypted separately to generate corresponding data fingerprints, which are stored through distributed nodes. The storage nodes are configured with network topology and redundant backup, and a role-based access control mechanism is established to finely manage the data access rights of different users, and finally build a security management database.

[0016] Finally, the effect evaluation is conducted based on the data in the collaborative interaction data stream and the security management database. The action norm data, emotion change data, and interaction feedback data are extracted from the collaborative interaction data stream, and time series analysis and feature extraction are performed respectively. The ability change trend and the emotion feature sequence are time-aligned and correlated, and then merged with the interaction quality score to obtain comprehensive evaluation data. Through comparative analysis with historical data, the degree of progress is calculated and quantified according to the evaluation dimension, and finally an intervention effect data set is generated.

[0017] Taking the exercise intervention process of a 6-year-old autistic child as an example, the multi-source sensing device collected the sequence of facial expression changes, voice emotion characteristics, heart rate variability indicators and joint movement trajectory data during the exercise. After time alignment, these data detected the emotional fluctuation points and movement incoordination points during the training process at the millisecond level, analyzed the changing rules of their training participation and movement standardization at the minute level, and found their long-term improvement trend in specific training projects at the day level. Based on these analysis results, the training content suitable for their ability level was selected from the sports project library, and the training difficulty and rhythm were dynamically adjusted according to their emotional change characteristics. During the training process, their movement standardization, emotional state and interaction quality were analyzed in real time, and the training parameters were adjusted in time when fatigue or distraction occurred. All collected and analyzed data were encrypted and stored through distributed nodes, and corresponding access rights were set according to different roles such as therapists, teachers, and parents. Through long-term tracking and analysis, quantitative evaluations were conducted on their improvement in motor ability, improvement in emotional stability, and improvement in social interaction ability, providing a basis for the optimization of subsequent intervention plans.

[0018] In the embodiment of the present application, the facial expressions, sounds, physiological and behavioral data of special children in the process of movement are collected and extracted in an all-round time series through multi-source sensing equipment, which realizes the accurate perception and quantitative characterization of the movement state of special children, and provides a rich data basis for subsequent personalized intervention. Based on the multidimensional training data stream, hierarchical analysis and processing are carried out, and the millisecond-level response data, minute-level adjustment data and day-level optimization data are integrated to build a hierarchical interaction intensity evaluation system, which can accurately grasp the immediate needs and long-term development trends of special children from different time scales. On this basis, the scheme intelligently screens the sports project library through an adaptive screening algorithm, dynamically optimizes and configures parameters such as training difficulty, rhythm and duration, realizes accurate matching and personalized adjustment of training content, and effectively improves the pertinence and adaptability of sports intervention. At the same time, the scheme also establishes a real-time interaction mechanism based on multimodal analysis, matches and coordinates the action standardization, emotional changes and interactive feedback data in real time, which can not only timely discover and respond to the abnormal state of special children in the training process, but also dynamically adjust the training strategy to ensure the continuous optimization of the training effect. In terms of data security, the solution uses distributed nodes for encrypted storage and implements fine-grained access control, effectively ensuring the security and privacy of data related to special children. Finally, the solution builds a complete intervention effect evaluation system. Through multi-dimensional correlation analysis and quantitative evaluation, it can objectively reflect the actual effect of exercise intervention and provide a scientific basis for the continuous improvement of training programs. Overall, this solution innovatively applies artificial intelligence algorithms in the field of sports intervention for special children, including adaptive screening algorithms, multimodal analysis algorithms, and collaborative interaction algorithms. The organic combination of these algorithms not only improves the intelligence level and personalization of sports intervention, but also significantly enhances the real-time and accuracy of the intervention process, providing more scientific and efficient technical support for the rehabilitation training of special children. Through the deep integration and optimized application of algorithm features, the solution successfully solves the problems of insufficient personalization, poor real-time performance, and weak security in traditional sports intervention methods, greatly improving the overall effect of sports intervention for special children.

[0019] In a specific embodiment, the process of executing step S101 may specifically include the following steps: (1) The facial expression data stream is sampled at 60fps for data collection, and the collected facial expression data is labeled with feature points to obtain facial feature data; the facial feature data is analyzed in the frequency domain by discrete Fourier transform to extract the facial micro-expression feature sequence to obtain the facial feature data; (2) The sound data is collected in all directions through an 8-channel microphone array, and the collected sound data is subjected to noise reduction and signal enhancement processing to obtain pure speech data; the pure speech data is analyzed in the time and frequency domain to extract the acoustic characteristic parameters of pitch, speech rate, and volume to obtain sound characteristic data; (3) The physiological data is collected in real time through wearable sensors, and the heart rate variability and galvanic skin response physiological signals are filtered to obtain steady-state physiological data; the steady-state physiological data is divided into time windows, and the characteristics of physiological state changes are extracted to obtain physiological characteristic data; (4) The behavior data is collected in three-dimensional space through a depth camera, and the spatial coordinates of the joint position data are mapped to obtain the motion trajectory data; the motion trajectory data is analyzed in time series to extract the motion amplitude, speed, and acceleration motion characteristics to obtain the behavior feature data; (5) Time-align the facial feature data, sound feature data, physiological feature data, and behavioral feature data, and perform feature fusion on the aligned feature data to obtain a feature vector sequence; sort the feature vector sequence according to the sampling timestamp to obtain a multi-dimensional training data stream.

[0020] Specifically, the high-speed camera collects the facial expressions of special children at a sampling rate of 60 frames per second, ensuring that the facial area is fully captured during the collection process. Feature points are annotated for each frame of the image collected. The annotated feature points include the positioning points of key areas such as eyebrows, eyes, nose, and mouth. The facial feature data is constructed through the spatial distribution of these feature points. The facial feature data is then converted to the frequency domain space through discrete Fourier transform, and the changing characteristics of facial micro-expressions are extracted in the frequency domain space. The discrete Fourier transform decomposes the time domain signal into a superposition of sine waves of different frequencies, thereby identifying the main frequency components of facial expression changes. These frequency components reflect the speed and amplitude characteristics of expression changes, forming expression feature data.

[0021] The sound data is collected using an 8-channel circular microphone array, which is evenly distributed in a circular shape to achieve omnidirectional sound collection. The collected original sound signal is first subjected to noise reduction processing, and an adaptive filter is used to remove environmental noise and interference signals. Then the signal is enhanced, and the effective components of the voice signal are selectively amplified to obtain pure voice data. The pure voice data is analyzed in the time and frequency domain to extract the pitch characteristics (fundamental frequency change), speech rate characteristics (phoneme duration) and volume characteristics (energy envelope). These acoustic feature parameters together constitute the sound feature data. Physiological data collection relies on wearable sensors such as smart bracelets to record heart rate variability indicators and galvanic skin response signals in real time. Heart rate variability data reflects the changes in the time interval between consecutive heartbeats, while galvanic skin response data records the changes in skin conductivity. These original signals are processed by bandpass filters to filter out high-frequency noise and baseline drift to obtain steady-state physiological data. The steady-state physiological data is divided into fixed time windows (such as 30 seconds), and statistical features (mean, variance, peak, etc.) and change features (rise rate, decline rate, etc.) are extracted in each time window to form physiological feature data.

[0022] The behavioral data is collected using a depth camera, which obtains three-dimensional depth information during the movement process through the principle of infrared structured light. The depth camera captures the spatial position information of the human skeleton joints, maps these position data to the standard three-dimensional space through the spatial coordinate system transformation, and obtains the motion trajectory data. The motion trajectory data is analyzed in time series to calculate the motion amplitude (displacement), instantaneous velocity (first-order derivative of displacement with respect to time), and acceleration (first-order derivative of velocity with respect to time) of the joints. These kinematic parameters constitute the behavioral characteristic data.

[0023] The facial feature data, voice feature data, physiological feature data, and behavioral feature data are time-aligned, and the data from different sources are mapped to the same time point based on the timestamp of data collection. The aligned feature data are fused at the feature level, and the feature vectors of each data source are connected into a unified feature vector. These feature vectors are arranged in timestamp order to form a multi-dimensional training data stream that represents the movement status of special children.

[0024] For example, in a sports intervention training, the acquisition equipment synchronously recorded the multimodal data of special children completing simple sports tasks. The high-speed camera captured subtle changes in facial expressions, and identified changes in facial expressions such as raised eyebrows and raised corners of the mouth through feature point annotation. After frequency domain analysis, it was found that the changes in facial expressions had regular frequency characteristics. At the same time, the voice signal collected by the microphone array was noise-reduced and enhanced, and the acoustic characteristics of slow speech but rich intonation were analyzed. The heart rate variability data recorded by the wearable device showed a regular fluctuation pattern, and the galvanic skin response changed with the difficulty of the task. The depth camera tracked and recorded the joint movement trajectory during the exercise, and analyzed the changes in the amplitude, speed and acceleration of the movement. These data from different sensors were time-aligned and feature-fused, and finally formed a multidimensional feature sequence that can fully reflect the movement and emotional state of special children.

[0025] In a specific embodiment, the process of executing step S102 may specifically include the following steps: (1) Segment the multi-dimensional training data stream according to the time dimension, divide the sampled data into millisecond time windows, and obtain millisecond data segments; perform mutation detection on the millisecond data segments, extract the emotional fluctuations and abnormal action sudden state data, and obtain millisecond response data; (2) Accumulate multi-dimensional training data streams at the minute level, perform sliding window analysis on the accumulated data, and obtain training process data; perform trend analysis on the training process data, extract training participation, emotional stability, and action normative state change data, and obtain minute-level adjustment data; (3) Summarize the multi-dimensional training data stream at the daily level, perform periodic analysis on the summarized data, and obtain training pattern data; perform statistical analysis on the training pattern data, extract the long-term performance data of training effects, emotional changes, and physical condition, and obtain daily optimization data; (4) Weights are assigned to millisecond-level response data, and the impact of sudden states is quantified as interactive adjustment parameters to obtain rapid response weight data; minute-level adjustment data is normalized, and the impact of state changes is quantified as interactive correction parameters to obtain medium-term adjustment weight data; day-level optimization data is standardized, and the impact of long-term performance is quantified as interactive optimization parameters to obtain long-term adjustment weight data; (5) Perform weighted combination of the quick response weight data, the mid-term adjustment weight data, and the long-term adjustment weight data, and normalize the combined data to obtain the interaction intensity data set.

[0026] Specifically, Figure 2As shown, it is a timing diagram of hierarchical analysis and processing of interaction parameters according to multidimensional training data stream in the embodiment of the present application, and the figure shows the complete processing flow from raw data to the final interaction intensity data set. The figure contains four main participants: multidimensional training data stream (MDS), millisecond processing (MS), minute processing (MIN) and day processing (DAY), as well as the final interaction intensity data set (INT). In the millisecond processing link, the system performs segmented processing of the data in a 10 millisecond time window, identifies emotional fluctuations and abnormal movements through the mutation detection formula D (t), and generates rapid response weight data. In the minute-level processing link, the system uses a 60-second sliding window for analysis, and uses the trend analysis formula T (w) to evaluate training participation, emotional stability and action standardization to form medium-term adjustment weight data. In the day-level processing link, the system summarizes the data, uses the statistical analysis formula P (d) to analyze the training effect, emotional changes and physical condition, and generates long-term adjustment weight data. The weight data of these three time scales are integrated through the normalization processing formula W (t) to form a unified interaction intensity data set. The annotations in the figure clearly show the mathematical formulas and main analysis contents used in each processing stage, and intuitively reflect the hierarchy and systematic nature of the entire data processing.

[0027] In millisecond-level processing, the multidimensional training data stream is segmented into 10-millisecond time windows, and mutation detection is performed on the data in each time window. The detection formula is: in, represents the value of the i-th emotional feature (such as facial expression changes, voice emotion changes) at time t, represents the value of the j-th action feature (such as joint angle, movement speed) at time t, D(t): represents the mutation detection value at time t is the weight coefficient of the i-th emotional feature; is the weight coefficient of the jth action feature; It is the overall balancing factor for emotional mutations; is the overall balance factor of action mutation; n is the total number of emotion features; m is the total number of action features.

[0028] The sudden state is identified by calculating the difference between these feature values ​​and the previous moment or the standard reference value. For example, when a special child is detected doing a pitching action, the facial expression suddenly changes from calm to nervous ( The value suddenly changes, and the arm trajectory appears unstable and jittery ( Values ​​deviate from the standard value ), these changes will lead to The value increases, thus being recognized as an emergency state that requires an immediate response.

[0029] In minute-level data processing, a 60-second sliding window is used, updated every 10 seconds, and trend analysis is performed on the accumulated data in the window. The analysis formula is: In this formula, Represents the change gradient of the state parameters (such as attention concentration and action completion quality) within the window w, Indicator representing the stability of features (e.g. emotional stability, action coherence). T(w) is the trend analysis value within window w; is the gradient of the state parameter; is the standard deviation of the kth state parameter in window w; is the stability index value of the lth feature; is the maximum reference value of the lth feature; is the weight coefficient of the kth state parameter; is the weight coefficient of the lth stability index; p is the total number of state parameters; q is the total number of stability indicators.

[0030] By calculating the trend and stability of these parameters, the system can identify medium-term change patterns during training. For example, if a child's attention level gradually decreases within a one-minute window ( is negative), and the action accuracy also begins to decrease ( These changes will be reflected by changes in the T(w) value.

[0031] Daily data processing focuses on the analysis of long-term training performance and uses the following statistical analysis formula: In this formula, Represents the training effect indicators on day d (such as movement accuracy, completion time), Represents an indicator of mood change (such as the duration of positive mood), represents an indicator of physical condition (e.g., fatigue, concentration level). P(d) is the overall performance score on day d; is the weight of the rth training effect indicator; is the weight of the e-th sentiment change indicator; is the weight of the vth individual performance status indicator; is the overall weight of the training effect dimension; is the overall weight of the emotion change dimension; is the overall weight of the physical condition dimension; s is the total number of training effect indicators; t is the total number of emotion change indicators; u is the total number of physical condition indicators.

[0032] By weighting these indicators, the system can find the long-term changes in the training effect. For example, through the analysis of data for several consecutive days, it may be found that the training effect of children in the morning period (higher values) and emotional states (more stable values) are better than other periods.

[0033] In order to integrate the data of these three time scales into a unified interaction intensity indicator, the following normalization formula is used: Where M(t), N(t), and L(t) represent the weights of millisecond-level, minute-level, and day-level data at time t, respectively. , , These weight coefficients are used to balance the importance of data at different time scales. This normalization process ensures that data at different time scales can be compared and integrated within the same numerical range. W(t) is the normalized interaction intensity value at time t; For example, suppose a special child is doing pitching training. The system detects sudden changes in facial expressions and abnormal arm trajectories at the millisecond level, and immediately calculates the mutation detection value through the D(t) formula. At the same time, at the minute level, the T(w) formula is used to analyze the changing trends of attention and movement accuracy. At the day level, the P(d) formula finds that the child's training effect is better during a specific period of time. The analysis results of these different time scales are finally integrated into a unified interaction intensity dataset through the W(t) formula, providing a quantitative basis for training adjustments. Through weight allocation and normalization, an interaction intensity dataset that guides training adjustments is formed.

[0034] In a specific embodiment, the process of executing step S103 may specifically include the following steps: (1) Decomposing the sports performance data in the multi-dimensional training data stream, and conducting ability evaluation and analysis on the decomposed data to obtain sports ability benchmark data; matching and screening the sports ability benchmark data with the training items in the sports item library to obtain a set of candidate training items; (2) Perform correlation analysis on the emotion parameters in the interaction intensity data set and the candidate training item set, filter the correlation data by emotion threshold, and obtain emotion-compatible training data; perform secondary screening on the candidate training item set based on the emotion-compatible training data to obtain training item data; (3) Calculate the difficulty coefficient of the training movements in the training project data, compare the calculation results with the sports ability benchmark data, and obtain the training difficulty parameter; classify and quantify the training difficulty parameter to obtain dynamic difficulty data; (4) According to the reaction speed index in the interaction intensity data set, the action rhythm in the training project data is adjusted to obtain the training rhythm parameter; the training rhythm parameter is divided into intervals to obtain the rhythm control data; (5) Based on the persistence index in the interaction intensity data set, the training time in the training project data is segmented to obtain the training duration parameter; the training duration parameter is configured into time periods to obtain the duration control data; (6) Integrate the dynamic difficulty data, rhythm control data, and duration control data, arrange the integrated data into a training sequence, and obtain a personalized training data package.

[0035] Specifically, the sports performance data is subjected to principal component analysis and feature extraction, including quantitative indicators of sports stability, coordination, reaction speed and other dimensions. These decomposed features are combined through the ability assessment matrix to form benchmark data that reflects the current sports ability level of special children. The sports ability benchmark data is then feature matched with the sports project library, and the set of alternative training projects that meet the current ability level is screened out by calculating the similarity between feature vectors.

[0036] The emotional parameters in the interaction intensity dataset are correlated with the candidate training item set. Emotional parameters include indicators such as the frequency of facial expression changes, voice emotional feature values, and physiological signal fluctuations. These indicators are compared with the emotional activation threshold of each candidate training item. Through emotional threshold filtering, those training items with high emotional compatibility are retained to form emotional compatibility training data. Subsequently, the candidate training item set is screened again based on the emotional compatibility training data to obtain training item data.

[0037] For the calculation and adjustment of training difficulty, the following formula is used: Parameter meaning: is the difficulty coefficient of training action d; is the i-th complexity index of action d; is the benchmark value of the i-th complexity index; is the j-th accuracy requirement of action d; is the standard value of the jth accuracy requirement; is the mth motion parameter of action d; is the reference value of the mth motion parameter; is the complexity index weight; Requires weight for accuracy; is the motion parameter deviation weight; is the number of complexity indicators; Require quantity for accuracy; is the number of motion parameters.

[0038] In practical applications, complexity indicators It includes specific parameters such as throwing distance (2 meters), target area (circle with a diameter of 30 cm), throwing angle requirement (vertical 45 degrees), etc. The accuracy requirement Q_j(d) is quantified into indicators such as hit rate requirement and force control, while the motion parameters It includes real-time monitoring data such as wrist angular velocity, elbow flexion, and torso stability.

[0039] For the adjustment of action rhythm, the following formula is used: Parameter meaning: is the rhythm parameter at time t; is the pth reaction speed index; is the rth action coherence index; is the u-th rhythm stability index; is the weight of the reaction speed indicator; is the action coherence index weight; is the rhythm stability index weight; , , is the overall weight of the three types of indicators; is the maximum value of the reaction speed index; is the mean value of the coherence index; It is the reference value of rhythm stability.

[0040] In the specific implementation, the reaction speed index Indicates the time interval from receiving the command to starting the action (1.5 seconds on average), an indicator of action continuity The fluency score of the complete throwing action is reflected in the rhythm stability index. It reflects the stability of the time interval between two consecutive throws (the target is set to 3 seconds). Based on the persistence indicators in the interaction intensity data set, the training time is segmented. The persistence indicators include attention span, physical tolerance, emotional stability cycle, etc. According to the distribution characteristics of these indicators, the training time is divided into multiple segments, and each segment is configured with corresponding training duration parameters. These duration parameters are configured through time periods to form detailed duration control data.

[0041] The dynamic difficulty data, rhythm control data and duration control data are integrated. The integration process takes into account the mutual constraints between the three types of data, generates a training sequence arrangement plan through a multi-objective optimization algorithm, and forms a personalized training data package.

[0042] For example, in a pitching training for a 7-year-old autistic child, the child's hand-eye coordination ability, spatial perception ability and other motor features are first extracted from the multidimensional training data stream. After decomposition and evaluation, these features show that the child's hand-eye coordination ability is at a basic level, but the spatial perception ability is relatively good. Based on these motor ability benchmark data, alternative training items including fixed-point pitching and mobile pitching are selected from the sports project library. Combined with the emotional parameter analysis in the interactive intensity data set, it is found that the child's emotions are relatively stable when throwing at a fixed target, while the moving target training is prone to cause anxiety. Therefore, the fixed target training items are retained through emotional threshold filtering. In the calculation of the difficulty coefficient, considering the characteristics of the child, the parameters such as pitching distance, target size, and accuracy requirements are substituted into the difficulty coefficient formula for calculation. In terms of action rhythm, based on its reaction speed index, the rhythm parameters such as the preparation time and throwing interval of the pitching action are quantitatively adjusted. The training time is divided into multiple 15-20 minute training units according to its attention span and physical condition. Ultimately, these parameters are integrated into a personalized training data package with progressive difficulty, controllable rhythm, and reasonable duration.

[0043] In a specific embodiment, the process of executing step S104 may specifically include the following steps: (1) Map the standard motion data in the personalized training data package to the real-time collected motion data, calculate the spatial deviation of the mapped data, and obtain the motion error data; compare and analyze the motion error data with the training difficulty parameter to obtain the motion standardization data; (2) Synchronously compare the real-time emotion index in the interaction intensity data set with the training rhythm parameters, perform fluctuation detection on the comparison results, and obtain emotion difference data; perform time series correlation on the emotion difference data and the training duration parameters to obtain emotion change data; (3) Score the normative indicators in the action normativeness data in real time, pair the scoring results with the emotion change data in a time window to obtain the interaction state data; perform segmented statistics on the interaction state data to obtain the interaction feedback data; (4) Time-align the action norm data with the emotion change data, perform correlation analysis on the aligned data, and obtain action emotion correlation data; fuse the action emotion correlation data with the interaction feedback data to obtain multimodal feature data; (5) Segmentally label the training process based on the multimodal feature data, match the labeling results with the training parameters in the personalized training data package to obtain training matching data; arrange the training matching data in time sequence to obtain training sequence data; (6) Cross-validate the training sequence data with the interaction feedback data, integrate the verification results, and obtain the collaborative interaction data stream.

[0044] Specifically, standard motion data are extracted from the personalized training data package, which contains the three-dimensional coordinate sequence of the joint points of the standard motion. The motion data of special children are collected in real time through a depth camera, and the key joint points (such as shoulders, elbows, wrists, etc.) in these data are spatially mapped with the standard motion data. The Euclidean distance is calculated for the mapped data to obtain the spatial deviation value of each joint point, thereby forming the motion error data. These error data are then compared with the training difficulty parameters, and the motion standardization score is calculated according to the allowable error range of different difficulty levels to generate the motion standardization data.

[0045] In the emotion analysis phase, real-time emotion indicators are extracted from the interaction intensity data set, including the frequency of facial expression changes, voice emotion feature values, physiological signal fluctuations and other data. These emotion indicators are synchronously compared with the training rhythm parameters to detect whether the emotional state matches the current training rhythm. Emotion difference data is obtained by calculating the fluctuation amplitude and frequency of the emotion indicators. These difference data are then time-series associated with the training duration parameters to analyze the relationship between the emotion change trend and the training duration to form emotion change data. The normative indicators in the action normativeness data are scored in real time, and the scoring criteria are based on dimensions such as action accuracy, stability and coherence. The scoring results are paired with the emotion change data in the same time window to generate interaction state data reflecting the current training state. These state data are divided into time periods and statistically analyzed to obtain interaction feedback data that characterizes the training quality.

[0046] In order to deeply analyze the correlation between action performance and emotional state, it is necessary to time-align the action standardization data and emotional change data. The aligned data is analyzed through correlation to calculate the correlation between action performance and emotional change, and generate action-emotion correlation data. These correlation data are then fused with the interactive feedback data at the feature level to obtain multimodal feature data containing multi-dimensional information. Based on the multimodal feature data, the entire training process is marked with time periods, and the selection of marking points is based on significant changes in feature data. These marking results are matched and analyzed with the training parameters in the personalized training data package to evaluate the parameter fitness of each training stage and generate training matching data. These matching data are arranged in chronological order to form training sequence data.

[0047] The training sequence data is cross-validated with the interactive feedback data. The validation process mainly examines whether the rationality of the training sequence is supported by the interactive feedback. The validated data is integrated to eventually generate a collaborative interactive data stream that reflects the entire training process.

[0048] Take a special child's pitching training as an example. When the child starts to perform the pitching action, the depth camera collects the real-time coordinate data of the wrist, elbow, shoulder and other joints. These data are mapped and compared with the joint data of the standard pitching action to calculate the spatial position deviation. At the same time, the facial expression recognition system captures the child's emotional changes, such as nervousness (frown) in the pitching preparation stage and relaxation (smile) after pitching. These emotional indicators are compared with the preset training rhythm (such as 3 seconds between each pitch), and it is found that the emotional fluctuations increase when the rhythm is accelerated. The system aligns the action error data and the emotional change data in time and finds that whenever the emotion tends to be stable, the action accuracy increases accordingly. This correlation is recorded in the action emotion association data and matched with the training parameters, and finally forms a collaborative interactive data stream to guide the adjustment of subsequent training parameters.

[0049] In a specific embodiment, the process of executing step S105 may specifically include the following steps: (1) Classify and organize the multi-dimensional training data stream according to data type, encrypt and mark the organized data to obtain training encrypted data; generate a data summary of the training encrypted data through a hash function to obtain a training data fingerprint; (2) Divide the collaborative interaction data stream into blocks according to the time series, encrypt and package the block data to obtain interactive encrypted data; digitally sign the interactive encrypted data to obtain the interactive data fingerprint; (3) Classify the personalized training data package according to the training items, encrypt and encapsulate the classified data to obtain the encrypted data of the training package; store the encrypted data of the training package on the blockchain to obtain the data fingerprint of the training package; (4) Distribute and store the training data fingerprint, interaction data fingerprint, and training package data fingerprint, configure the network topology of the storage nodes, and obtain node distribution data; perform redundant backup of the node distribution data and obtain storage node mapping; (5) Divide the access levels of training encrypted data, interactive encrypted data, and training package encrypted data, configure the access rights of different roles in a matrix, and obtain permission allocation data; associate and bind the permission allocation data with the storage node mapping to obtain access control data; (6) Integrate the access control data and node distribution data, construct an index for the integrated data, and obtain a security management database.

[0050] Specifically, the multidimensional training data stream is classified according to the data type, and the expression data, sound data, physiological data and behavioral data are classified separately. A unique identifier and timestamp are added to each type of data, and the AES-256 encryption algorithm is used to encrypt and process the training encrypted data. The data summary of the training encrypted data is then calculated using the SHA-256 hash algorithm to generate a fixed-length training data fingerprint. This fingerprint value can uniquely identify the integrity of the data content. The processing of the collaborative interactive data stream adopts the time series block method to divide the continuous data stream into multiple data blocks according to a fixed time length (such as 10 minutes). Each data block contains the action standardization data, emotional change data and interactive feedback data within the time period. These block data are encrypted using the RSA asymmetric encryption algorithm to obtain interactive encrypted data. The encrypted data is digitally signed with a private key to generate an interactive data fingerprint for subsequent data verification.

[0051] The processing of personalized training data packets is first classified according to the type of training items, such as basic sports training, coordination training, sensory integration training, etc. Each type of training data is grouped and encrypted and encapsulated using a hybrid encryption method: first, the data content is encrypted with a symmetric key, and then the symmetric key is encrypted with an asymmetric public key to form encrypted training package data. These encrypted data are stored through the blockchain network, and each data packet is written into the block as a transaction record to generate an unalterable training package data fingerprint. The storage of the three types of data fingerprints (training data fingerprint, interaction data fingerprint, and training package data fingerprint) adopts a distributed architecture, and the data is dispersed and stored on multiple service nodes. The network topology between nodes adopts a P2P network structure, and each node maintains a complete routing table. The data is sharded and distributed through a consistent hashing algorithm to generate node distribution data. A multi-copy backup strategy is implemented for these distributed data, and copies are saved on nodes in different geographical locations to form a storage node mapping relationship.

[0052] Data access control is based on the RBAC (role-based access control) model, which divides users into different roles (such as administrators, therapists, teachers, parents, etc.). Different access levels are set for three types of encrypted data (training encrypted data, interactive encrypted data, and training package encrypted data). The access control matrix defines the operation permissions of each role on different data and generates permission allocation data. These permission data are associated with storage node mapping to ensure that users can only access data on authorized nodes, forming access control data.

[0053] Integrate access control data and node distribution data to build a distributed index structure. The index uses a B+ tree structure, with data identifiers as index keys, storage locations and access permission information as index values, to form a complete security management database. The database supports multi-dimensional queries, such as retrieval by time, type, user role, etc.

[0054] For example, during the sports intervention training of a special child, when a training data containing a pitching action is collected, the action data, expression data and physiological data in this data are first encrypted by AES and the corresponding hash value is generated. These encrypted data are divided into multiple time blocks, each of which is digitally signed. At the same time, the relevant training program data is encapsulated, encrypted and recorded through the blockchain. The storage location information of all these data is recorded in the distributed nodes, and the corresponding access rights are set. For example, the therapist can view the complete training data, while the parents can only access the summary training report. When the therapist needs to adjust the training plan, the identity is verified through the access control system, and then the required encrypted data is obtained from the distributed storage, and the authorized key is used for decryption and viewing. This multi-level security mechanism ensures that the privacy data of special children is effectively protected, while ensuring that relevant personnel can easily access the required information.

[0055] In a specific embodiment, the process of executing step S106 may specifically include the following steps: (1) Extract action standardization data from the collaborative interaction data stream, calculate the score of action completion quality, and obtain sports ability data; process the sports ability data into time series, perform trend analysis on sports performance, and obtain the ability change curve; (2) Extract emotion change data from the collaborative interaction data stream, perform numerical statistics on the amplitude of emotion fluctuations, and obtain emotion state data; perform periodic analysis on the emotion state data, extract features of emotion changes, and obtain an emotion feature sequence; (3) Extract interaction feedback data from the collaborative interaction data stream, perform data analysis on the interaction response degree, and obtain interaction effect data; segment the interaction effect data, perform score statistics on the interaction quality, and obtain an interaction score sequence; (4) Time-align the ability change curve with the emotion feature sequence, calculate the correlation of the aligned data, and obtain the ability-emotion correlation data; fuse the ability-emotion correlation data with the interaction score sequence to obtain the comprehensive evaluation data; (5) Search the historical data in the safety management database, compare and analyze the search results with the comprehensive evaluation data to obtain the progress data; perform dimensional statistics on the progress data to obtain intervention analysis data; (6) The intervention analysis data are quantified according to the evaluation dimensions, and an indicator system is constructed for the processing results to obtain an intervention effect data set.

[0056] Specifically, the action standardization data includes quantitative indicators such as angle deviation, force control, and speed stability when special children perform actions. By weighted calculation of these quantitative indicators, the motor ability data reflecting the quality of action completion is obtained. The motor ability data is arranged in chronological order to form a time series. Trend analysis is performed on this time series to calculate the change rate of action accuracy and the degree of improvement in stability, thereby generating an ability change curve.

[0057] The emotion analysis part extracts emotion change data from the collaborative interaction data stream, which includes information such as facial expression changes, voice emotion characteristics, and physiological signal fluctuations. The amplitude of emotion fluctuations is statistically analyzed, and the frequency, duration, and conversion rules of positive and negative emotions are calculated to obtain emotion state data. The emotion state data is periodically analyzed to identify the regular characteristics of emotion changes, such as emotional stability in a specific training stage, emotional response patterns to different training contents, etc., to form an emotion feature sequence. The processing of interactive feedback data focuses on analyzing the interaction quality of special children during the training process. The interactive feedback data extracted from the collaborative interaction data stream includes indicators such as instruction response speed, action correction effect, and attention maintenance time. These indicators are quantitatively analyzed to evaluate the timeliness and effectiveness of the interaction and generate interaction effect data. The interaction effect data is segmented according to the training stage, and the interaction quality of each stage is scored to form an interaction score sequence.

[0058] The correlation analysis of the ability change curve and the emotional feature sequence first needs to be time-aligned to ensure the correspondence between the two sets of data in the time dimension. The correlation coefficient is calculated for the aligned data, and the correlation pattern between the improvement of motor ability and emotional changes is analyzed to obtain the ability-emotion correlation data. These correlation data are then integrated with the interaction score sequence at the feature level, and through the combination of multi-dimensional features, comprehensive evaluation data that fully reflects the training effect is generated. The historical training data is retrieved from the safety management database and compared with the current comprehensive evaluation data. The comparison includes the improvement in movement accuracy, the improvement in emotional stability, the change trend of interaction quality, etc., and the progress degree data is obtained through this longitudinal comparison. The progress degree data is classified and counted according to different dimensions (such as motor ability, emotional regulation, social interaction, etc.) to form intervention analysis data.

[0059] The intervention analysis data is converted into standardized evaluation indicators. The evaluation dimensions include basic motor skills, coordination, attention, emotional regulation, social interaction, etc. The data of each dimension is quantified, and a multi-level indicator evaluation system is established to finally generate an intervention effect data set.

[0060] Taking the hand-eye coordination training of a 6-year-old special child as an example, the data such as the deviation of the hand movement trajectory and the stability of force control when grasping and placing objects were extracted from the action standardization data. After analysis, these data showed that the accuracy of hand movement gradually improved during continuous training, especially in the aspect of force control. At the same time, it was observed from the emotional change data that when the correct action was completed, the facial expression and voice characteristics showed positive emotional reactions, and this positive emotional state was positively correlated with the accuracy of subsequent actions. In terms of interactive feedback, the records showed that the child's response speed to instructions gradually improved with the progress of training, especially when the game-based training method was adopted, the enthusiasm for interaction was significantly enhanced. By comparing the historical training data, it was found that the accuracy of the child's hand-eye coordination ability was significantly improved during the three-month training period, and the emotional regulation ability was also significantly improved. These improvements were quantified through standardized evaluation indicators.

[0061] The above describes the special children's sports intervention information interaction method based on the cloud platform in the embodiment of the present application. The following describes the special children's sports intervention information interaction system based on the cloud platform in the embodiment of the present application. Figure 3 In the embodiment of the present application, an embodiment of the special children's sports intervention information interaction system based on the cloud platform includes: The acquisition module 201 is used to perform time-series acquisition and processing of the expression data stream, sound data, physiological data and behavioral data of the special children's movement process through multi-source sensing equipment, perform feature extraction and time synchronization processing on the multi-source data, and obtain a multi-dimensional training data stream; An analysis module 202 is used to perform hierarchical analysis on the interaction parameters according to the multi-dimensional training data stream, and to fuse the millisecond-level response data, minute-level adjustment data, and day-level optimization data to obtain an interaction intensity data set; A screening module 203 is used to adaptively screen the training content in the sports project library based on the multi-dimensional training data stream and the interactive intensity data set, dynamically configure the training difficulty parameter, the training rhythm parameter and the training duration parameter, and obtain a personalized training data packet; The processing module 204 is used to perform multimodal analysis and processing on the real-time motion data of special children according to the interaction intensity data set and the personalized training data set, and to match the action standardization data, the emotion change data and the interaction feedback data in real time to obtain a collaborative interaction data stream; An encryption module 205 is used to encrypt and store the multi-dimensional training data stream, the collaborative interaction data stream and the personalized training data packet through distributed nodes, and hierarchically configure data access permissions to obtain a security management database; The association module 206 is used to perform multi-dimensional association analysis on the athletic ability data, emotional state data and interaction effect data based on the collaborative interaction data stream and the security management database, and to quantitatively evaluate the analysis results to obtain an intervention effect data set.

[0062] Through the collaboration of the above components, multi-source sensing equipment is used to collect and extract features of the facial expressions, sounds, physiological and behavioral data of special children during exercise, achieving accurate perception and quantitative representation of the movement status of special children, providing a rich data basis for subsequent personalized intervention. Based on the hierarchical analysis and processing of multi-dimensional training data streams, the millisecond-level response data, minute-level adjustment data and day-level optimization data are integrated to build a hierarchical interaction intensity evaluation system, which can accurately grasp the immediate needs and long-term development trends of special children from different time scales. On this basis, the program uses an adaptive screening algorithm to intelligently screen the sports project library, dynamically optimizes and configures parameters such as training difficulty, rhythm and duration, and achieves accurate matching and personalized adjustment of training content, effectively improving the pertinence and adaptability of sports intervention. At the same time, the program also establishes a real-time interaction mechanism based on multimodal analysis, which matches and collaboratively analyzes the action standardization, emotional changes and interactive feedback data in real time, which can not only timely discover and respond to the abnormal state of special children during training, but also dynamically adjust the training strategy to ensure the continuous optimization of training effects. In terms of data security, the solution uses distributed nodes for encrypted storage and implements fine-grained access control, effectively ensuring the security and privacy of data related to special children. Finally, the solution builds a complete intervention effect evaluation system. Through multi-dimensional correlation analysis and quantitative evaluation, it can objectively reflect the actual effect of exercise intervention and provide a scientific basis for the continuous improvement of training programs. Overall, this solution innovatively applies artificial intelligence algorithms in the field of sports intervention for special children, including adaptive screening algorithms, multimodal analysis algorithms, and collaborative interaction algorithms. The organic combination of these algorithms not only improves the intelligence level and personalization of sports intervention, but also significantly enhances the real-time and accuracy of the intervention process, providing more scientific and efficient technical support for the rehabilitation training of special children. Through the deep integration and optimized application of algorithm features, the solution successfully solves the problems of insufficient personalization, poor real-time performance, and weak security in traditional sports intervention methods, greatly improving the overall effect of sports intervention for special children.

[0063] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0064] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A special children's sports intervention information interaction method based on a cloud platform, characterized in that: The cloud platform-based special children's sports intervention information interaction method includes: Through multi-source sensing equipment, the expression data stream, sound data, physiological data and behavioral data of special children during their movements are collected and processed in time series, and the multi-source data are subjected to feature extraction and time synchronization processing to obtain a multi-dimensional training data stream; Performing hierarchical analysis and processing on the interaction parameters according to the multi-dimensional training data stream, fusing the millisecond-level response data, minute-level adjustment data, and day-level optimization data to obtain an interaction intensity data set; Based on the multi-dimensional training data stream and the interactive intensity data set, adaptively screening the training content in the sports project library, dynamically configuring the training difficulty parameter, the training rhythm parameter and the training duration parameter, and obtaining a personalized training data packet; Based on the interaction intensity data set and the personalized training data packet, multimodal analysis and processing are performed on the real-time motion data of special children, and the action standardization data, emotion change data and interaction feedback data are matched in real time to obtain a collaborative interaction data stream; The multi-dimensional training data stream, the collaborative interaction data stream and the personalized training data packet are encrypted and stored through distributed nodes, and data access permissions are configured in a hierarchical manner to obtain a security management database; Based on the collaborative interaction data stream and the security management database, a multi-dimensional correlation analysis is performed on the athletic ability data, emotional state data and interaction effect data, and the analysis results are quantitatively evaluated to obtain an intervention effect data set.

2. The method for interacting with special children's sports intervention information based on a cloud platform according to claim 1, characterized in that: The multi-source sensing device is used to perform time-series acquisition and processing of the expression data stream, sound data, physiological data and behavioral data of the special children's movement process, and the multi-source data is subjected to feature extraction and time synchronization processing to obtain a multi-dimensional training data stream, including: The facial expression data stream is sampled at 60fps to collect data, and the collected facial expression data is marked with feature points to obtain facial feature data; the facial feature data is analyzed in the frequency domain by discrete Fourier transform to extract the facial micro-expression feature sequence to obtain the facial feature data; The sound data is collected in all directions through an 8-channel microphone array, and the collected sound data is subjected to noise reduction and signal enhancement processing to obtain pure speech data; the pure speech data is analyzed in the time-frequency domain to extract acoustic characteristic parameters of pitch, speech speed, and volume to obtain sound characteristic data; The physiological data is collected in real time through wearable sensors, and the heart rate variability and galvanic skin response physiological signals are filtered to obtain steady-state physiological data; the steady-state physiological data is divided into time windows, and the characteristics of physiological state changes are extracted to obtain physiological characteristic data; The behavior data is collected in three-dimensional space by a depth camera, and the joint point position data is mapped to spatial coordinates to obtain motion trajectory data; the motion trajectory data is analyzed in time series to extract motion amplitude, speed, and acceleration motion features to obtain behavior feature data; The expression feature data, the sound feature data, the physiological feature data and the behavior feature data are time aligned, and the aligned feature data are feature fused to obtain a feature vector sequence; the feature vector sequence is sorted according to sampling timestamps to obtain the multidimensional training data stream.

3. The method for interactive information of special children's sports intervention based on cloud platform according to claim 1, characterized in that: The interactive parameters are analyzed and processed in layers according to the multi-dimensional training data stream, and the millisecond-level response data, minute-level adjustment data and day-level optimization data are fused to obtain an interactive intensity data set, including: The multi-dimensional training data stream is segmented and sampled according to the time dimension, and the sampled data is divided into millisecond-level time windows to obtain millisecond-level data segments; mutation detection is performed on the millisecond-level data segments, and the emotional fluctuation and abnormal action sudden state data are extracted to obtain the millisecond-level response data; Accumulate the multidimensional training data stream at the minute level, perform sliding window analysis on the accumulated data, and obtain training process data; perform trend analysis on the training process data, extract training participation, emotional stability, and action normative state change data, and obtain the minute-level adjustment data; Summarize the multi-dimensional training data stream at the daily level, perform periodic analysis on the summarized data, and obtain training regularity data; perform statistical analysis on the training regularity data, extract the training effect, emotional changes, and long-term performance data of physical condition, and obtain the daily optimization data; The millisecond-level response data is weighted, and the impact of the sudden state is quantified as an interactive adjustment parameter to obtain rapid response weight data; the minute-level adjustment data is normalized, and the impact of the state change is quantified as an interactive correction parameter to obtain medium-term adjustment weight data; the day-level optimization data is standardized, and the impact of the long-term performance is quantified as an interactive optimization parameter to obtain long-term adjustment weight data; The quick response weight data, the mid-term adjustment weight data and the long-term adjustment weight data are weighted and combined, and the combined data are normalized to obtain the interaction intensity data set.

4. The method for interactive information of special children's sports intervention based on cloud platform according to claim 1, characterized in that: Based on the multi-dimensional training data stream and the interactive intensity data set, the training content in the sports project library is adaptively screened and processed, and the training difficulty parameter, the training rhythm parameter and the training duration parameter are dynamically configured to obtain a personalized training data packet, including: Decomposing the sports performance data in the multidimensional training data stream by features, and performing capability assessment and analysis on the decomposed data to obtain sports capability benchmark data; matching and screening the sports capability benchmark data with the training items in the sports item library to obtain a set of candidate training items; Performing correlation analysis on the emotion parameters in the interaction intensity data set and the candidate training item set, performing emotion threshold filtering on the correlation data, and obtaining emotion compatibility training data; performing secondary screening on the candidate training item set according to the emotion compatibility training data, and obtaining training item data; Calculating the difficulty coefficient of the training movements in the training project data, comparing the calculation result with the sports ability benchmark data to obtain the training difficulty parameter; grading and quantifying the training difficulty parameter to obtain dynamic difficulty data; According to the reaction speed index in the interaction intensity data set, the action rhythm in the training item data is adjusted to obtain the training rhythm parameter; the training rhythm parameter is divided into intervals to obtain rhythm control data; Based on the persistence index in the interaction intensity data set, the training time in the training project data is segmented to obtain the training duration parameter; the training duration parameter is configured for time periods to obtain duration control data; The dynamic difficulty data, the rhythm control data and the duration control data are integrated, and the integrated data are arranged into a training sequence to obtain the personalized training data packet.

5. The method for interactive information of sports intervention for special children based on cloud platform according to claim 1, characterized in that: The multimodal analysis and processing of the real-time motion data of special children is performed based on the interaction intensity data set and the personalized training data set, and the action standardization data, the emotion change data and the interaction feedback data are matched in real time to obtain a collaborative interaction data stream, including: Perform joint point mapping of the standard motion data in the personalized training data package and the real-time collected motion data, perform spatial deviation calculation on the mapped data, and obtain motion error data; compare and analyze the motion error data with the training difficulty parameter to obtain the motion standardization data; Synchronously comparing the real-time emotion index in the interaction intensity data set with the training rhythm parameter, performing fluctuation detection on the comparison result to obtain emotion difference data; performing time series association between the emotion difference data and the training duration parameter to obtain the emotion change data; Scoring the normative index in the action normativeness data in real time, pairing the scoring result with the emotion change data in a time window to obtain interaction state data; performing segmented statistics on the interaction state data to obtain the interaction feedback data; Time-aligning the action standardization data with the emotion change data, performing correlation analysis on the aligned data to obtain action emotion association data; fusing the action emotion association data with the interactive feedback data to obtain multimodal feature data; Based on the multimodal feature data, the training process is segmented and marked, and the marking results are matched with the training parameters in the personalized training data packet to obtain training matching data; the training matching data is arranged in time sequence to obtain training sequence data; The training sequence data and the interactive feedback data are cross-validated, and the validation results are integrated to obtain the collaborative interactive data stream.

6. The method for interacting with special children's sports intervention information based on a cloud platform according to claim 1, characterized in that: The method of encrypting and storing the multi-dimensional training data stream, the collaborative interaction data stream and the personalized training data packet through distributed nodes, and hierarchically configuring data access rights to obtain a security management database includes: Classify and sort the multidimensional training data stream according to data type, encrypt and mark the sorted data to obtain training encrypted data; generate a data summary of the training encrypted data through a hash function to obtain a training data fingerprint; The collaborative interaction data stream is divided into data blocks according to the time series, and the block data is encrypted and packaged to obtain interactive encrypted data; the interactive encrypted data is digitally signed to obtain an interactive data fingerprint; Classifying the personalized training data packet according to the training items, encrypting and encapsulating the classified data to obtain training package encrypted data; storing the training package encrypted data on a blockchain to obtain a training package data fingerprint; Distribute the training data fingerprint, the interaction data fingerprint and the training package data fingerprint, configure the network topology of the storage nodes to obtain node distribution data; perform redundant backup on the node distribution data to obtain storage node mapping; The training encrypted data, the interactive encrypted data and the training package encrypted data are divided into access levels, and the access rights of different roles are configured in a matrix to obtain permission allocation data; the permission allocation data is associated and bound with the storage node mapping to obtain access control data; The access control data and the node distribution data are integrated, and indexing is performed on the integrated data to obtain the security management database.

7. The method for interacting with special children's sports intervention information based on a cloud platform according to claim 1, characterized in that: Based on the collaborative interaction data stream and the security management database, a multi-dimensional correlation analysis is performed on the athletic ability data, the emotional state data, and the interaction effect data, and the analysis results are quantitatively evaluated to obtain an intervention effect data set, including: Extracting action standardization data from the collaborative interaction data stream, calculating the score of action completion quality, and obtaining the athletic ability data; performing time series processing on the athletic ability data, performing trend analysis on athletic performance, and obtaining an ability change curve; Extracting emotion change data from the collaborative interaction data stream, performing numerical statistics on the emotion fluctuation amplitude, and obtaining the emotion state data; performing periodic analysis on the emotion state data, performing feature extraction on the emotion change, and obtaining an emotion feature sequence; Extracting interaction feedback data from the collaborative interaction data stream, performing data analysis on the interaction response degree, and obtaining the interaction effect data; segmenting the interaction effect data, and scoring and counting the interaction quality to obtain an interaction scoring sequence; Time-aligning the capability change curve with the emotion feature sequence, performing correlation calculation on the aligned data, and obtaining capability emotion correlation data; fusing the capability emotion correlation data with the interaction score sequence, and obtaining comprehensive evaluation data; Retrieving historical data in the safety management database, comparing and analyzing the retrieval results with the comprehensive evaluation data to obtain progress data; performing dimensional statistics on the progress data to obtain intervention analysis data; The intervention analysis data is quantified according to the evaluation dimensions, and an indicator system is constructed for the processing results to obtain the intervention effect data set.

8. A cloud platform-based special children's sports intervention information interaction system, used to implement the cloud platform-based special children's sports intervention information interaction method according to any one of claims 1 to 7, characterized in that: The special children's sports intervention information interaction system based on the cloud platform includes: The acquisition module is used to perform time-series acquisition and processing of the expression data stream, sound data, physiological data and behavioral data of the special children's movement process through multi-source sensing equipment, perform feature extraction and time synchronization processing on the multi-source data, and obtain a multi-dimensional training data stream; An analysis module, used to perform hierarchical analysis on the interaction parameters according to the multi-dimensional training data stream, and to fuse the millisecond-level response data, minute-level adjustment data, and day-level optimization data to obtain an interaction intensity data set; A screening module, for adaptively screening the training contents in the sports project library based on the multi-dimensional training data stream and the interactive intensity data set, dynamically configuring the training difficulty parameter, the training rhythm parameter and the training duration parameter, and obtaining a personalized training data packet; A processing module, used to perform multimodal analysis and processing on the real-time motion data of special children according to the interaction intensity data set and the personalized training data packet, and to match the action standardization data, the emotion change data and the interaction feedback data in real time to obtain a collaborative interaction data stream; An encryption module is used to encrypt and store the multidimensional training data stream, the collaborative interaction data stream and the personalized training data packet through distributed nodes, and to configure data access permissions in a hierarchical manner to obtain a security management database; The association module is used to perform multi-dimensional association analysis on the athletic ability data, emotional state data and interaction effect data based on the collaborative interaction data stream and the security management database, and to quantitatively evaluate the analysis results to obtain an intervention effect data set.

Images