Language rehabilitation training interaction system for children with autism spectrum disorder

Abstract

The invention relates to the technical field of rehabilitation medical treatment and digital health, in particular to an autism spectrum disorder child language rehabilitation training interaction system which comprises a user terminal and a server, and the user terminal comprises a parent terminal, a child interaction terminal and an expert management terminal. The server comprises a data acquisition module, a data verification module, a personalized training module and an evaluation module. The system receives a child training video from the parent terminal, and receives voice, operation behaviors and response duration data from the child interaction terminal; performing logic association verification on the multi-terminal association data, and marking verified training data; child ability evaluation data is updated based on the last verified training data, and a personalized training task package is generated and pushed through a personalized training module; the children's language ability, social interaction behaviors and compliance indexes are quantitatively evaluated based on verified training data in a preset evaluation period, a visual evaluation report is generated, and the system further comprises an early warning and prompting module which can push early warning information in time.

Description

Technical Field

[0001] This invention relates to the fields of rehabilitation medicine and digital health technology, specifically to an interactive system for language rehabilitation training of children with autism spectrum disorder. Background Technology

[0002] Autism spectrum disorder (ASD) is a common neurodevelopmental disorder characterized by core clinical features including impaired social communication, delayed language development, and repetitive, stereotyped behaviors. Clinical studies have confirmed that early, intensive, and personalized language intervention is a key approach to improving the prognosis of children with ASD.

[0003] However, traditional language rehabilitation intervention models for children with ASD face many practical difficulties: First, professional therapists are scarce and unevenly distributed, making it difficult for children in remote areas to access high-quality intervention services. Face-to-face intervention is also costly, and long-term intervention places a heavy economic burden on families. Second, during family intervention, training data relies on parents' self-reporting, lacking an objective verification mechanism, which leads to doubts about the authenticity of the data and affects the accuracy of efficacy assessment. Third, the training content of existing intervention systems is mostly static and pre-set, making it difficult to dynamically adjust according to the child's real-time training performance. This results in insufficient personalization and adaptability, and the interactive forms are monotonous and boring, making it difficult to maintain the training interest of children with ASD. Family intervention compliance is generally low, and the final effect assessment also relies heavily on subjective scales, lacking quantitative analysis of multimodal data during the training process, making it difficult to form accurate closed-loop feedback.

[0004] In recent years, parent-led intervention models based on mobile devices have gradually emerged, but existing technologies have not yet solved the aforementioned problems. Therefore, there is an urgent need for an interactive language rehabilitation training system for children with autism spectrum disorder that can ensure the authenticity of training data, achieve dynamic and personalized intervention, improve children's participation, and conduct multi-dimensional intelligent assessments. Summary of the Invention

[0005] In view of the shortcomings of existing technologies, this invention aims to provide an interactive language rehabilitation training system for children with autism spectrum disorder, which can improve the effectiveness, accessibility and accuracy of rehabilitation training.

[0006] The basic solution provided by this invention is: an interactive language rehabilitation training system for children with autism spectrum disorder, including a user terminal and a server. The user terminal includes a parent terminal, a child interaction terminal, and an expert management terminal. The server includes a data acquisition module, a data verification module, a personalized training module, and an evaluation module; The data acquisition module is used to receive children's training videos from the parent's terminal; it is also used to receive voice data, operation behavior data and reaction time data from the children's interactive terminal. The voice data includes the pronunciation and sentence expression data of the children when completing interactive training tasks. The operation behavior data includes the screen clicks and gesture interaction data of the children when completing interactive training tasks. The reaction time data includes the task response delay and single-stage completion time data of the children when completing interactive training tasks. The data verification module is used to integrate children's training videos, voice data, operational behavior data, and reaction time data within the same standardized training period to obtain training data, and to perform logical correlation verification. The training data that passes the verification is marked as verified training data. The standardized training period refers to a training period that is carried out at a preset frequency and lasts for a preset duration each time. The personalized training module is used to output a personalized training task package based on the child's ability assessment data and the most recently validated training data. The personalized training task package includes a guide video, interactive levels, and reinforcer animations. The personalized training module is used to push the personalized training task package to the parent's terminal and the child's interactive terminal. The child's ability assessment data includes a language clarity score, cumulative vocabulary size, and instruction compliance accuracy rate. The child's ability assessment data is collected through a standardized scale and dynamically updated based on validated training data after an intervention event occurs. The intervention event is the first time the system pushes the training task package to the child's interactive terminal and initiates data collection. The language clarity score represents the recognizability of the child's pronunciation, the cumulative vocabulary size represents the total number of words that the child can actively pronounce and understand the meaning of, and the instruction compliance accuracy rate represents the child's correct execution rate of instructions. The assessment module is used to quantitatively assess children's language abilities, social interaction behaviors, and compliance index based on verified training data within a preset assessment period, and generate a visual assessment report.

[0007] The principle of this invention is as follows: Video and multimodal interaction data during the training process are collected through parent terminals and child interactive terminals respectively, achieving comprehensive coverage of training scenario data; the server's data acquisition module integrates multi-terminal data into training data, and then the data verification module filters out invalid data through logical correlation verification to ensure the authenticity and reliability of the training data; the personalized training module, based on the child's ability assessment data and the most recently verified training data, accurately matches training content that matches the child's current ability level and outputs personalized training task packages, avoiding inefficient intervention caused by training difficulty being too high or too low; finally, the evaluation module, based on continuously accumulated verified training data, periodically quantitatively evaluates the child's language ability, social interaction behavior, and compliance index, generating a visual report to provide an objective basis for subsequent intervention adjustments, thus achieving a closed-loop process of collection, verification, intervention, and evaluation.

[0008] The beneficial effects of this invention are as follows: By collecting data from multiple terminals and verifying logical correlations, it solves the problem of data reliance on parental self-reporting and difficulty in ensuring authenticity in traditional family interventions, laying a reliable data foundation for personalized intervention and effect evaluation; By combining children's ability assessment data and the most recently verified training data, a personalized training module dynamically generates an appropriate training task package, avoiding the limitations of static preset content and ensuring that the intervention content always matches the child's ability level; By relying on parent terminals and children's interactive terminals to build family intervention scenarios, it eliminates the need for a professional therapist to be present in real time, reducing geographical and cost limitations and enabling more children with autism spectrum disorders to receive continuous intervention treatment; Periodic quantitative evaluation based on continuous verified training data can more objectively and comprehensively reflect changes in children's abilities and provide reliable feedback on intervention effects.

[0009] Furthermore, the data verification module includes a spatiotemporal correlation verification unit, a behavioral result logic verification unit, and a data integrity check unit; The spatiotemporal correlation verification unit is used to verify whether the timestamp and location information of the child training video reported by the parent terminal match the task start time and device identifier recorded by the child interactive terminal and whether they are within the preset tolerance range. The behavior result logic verification unit is used to determine whether the key behavior frames in the children's training video correspond to the corresponding task nodes recorded by the children's interactive terminal. The judgment criteria include whether the deviation between the time of the action occurrence of the key behavior frame and the timestamp of the operation behavior data meets the preset time deviation requirements, whether the collection time period of the child's pronunciation action and the voice data in the behavior frame meets the preset overlap conditions, and whether the difference between the reaction time data and the task execution time in the behavior frame is within the preset difference range. The data integrity checking unit is used to check whether the difference between the training duration corresponding to the child's training video in a single training data session and the total time recorded by the child's interactive terminal meets the preset duration difference requirement. The training data is marked as verified training data only if the verification results of the spatiotemporal correlation verification unit, the behavior result logic verification unit, and the data integrity check unit all meet the preset rules.

[0010] By constructing a multi-dimensional data verification system, the credibility of verified training data is improved, providing high-quality data support for subsequent personalized recommendation and evaluation modules, and strengthening the reliability of the intervention loop. Among them, the spatiotemporal correlation verification unit ensures that data from multiple terminals comes from the same training scenario, avoiding data confusion; the behavior result logic verification unit verifies the rationality of data through the correspondence between behavior and data, filtering out false data; and the data integrity check unit ensures that the training data is complete and that the data is comprehensive.

[0011] Furthermore, the personalized training module includes a child's ability information unit, a task difficulty setting model, and a content generation unit; The child ability information unit is used to dynamically update the child ability assessment data based on the most recently verified training data; The task difficulty setting model takes children's ability assessment data as input and outputs the optimal challenge difficulty of the recommended task. The optimal challenge difficulty includes vocabulary difficulty level, sentence complexity level, and interaction intensity. The vocabulary difficulty level is divided into three levels (Level 1, Level 2, and Level 3) based on language clarity score and preset score threshold. The sentence complexity level is divided into three levels (Level 1, Level 2, and Level 3) based on cumulative vocabulary size and preset vocabulary size threshold. The interaction intensity is divided into three levels (Level 1, Level 2, and Level 3) based on instruction compliance accuracy rate and preset accuracy rate threshold. The content generation unit is used to automatically assemble and generate a personalized training task package, including a guide video, interactive levels, and reinforcement animations, from a multimedia resource library based on the optimal challenge difficulty and preset training objectives.

[0012] The child ability information unit continuously updates child ability assessment data based on the most recently validated training data, enabling dynamic tracking of changes in children's abilities and reflecting their real-time levels. The task difficulty setting model calculates the optimal challenge difficulty through collaborative analysis of ability information and historical task completion rates, ensuring that the training content is neither too simple, leading to stagnation, nor too difficult, causing frustration, and always within the range of children's ability improvement. The content generation unit automatically assembles training task packages from a multimedia resource library, combining guiding videos, interactive levels, and reinforcer animations to enhance the fun and interactivity of the training content, helping to maintain the training interest of children with autism spectrum disorder and improve intervention adherence.

[0013] Furthermore, the evaluation module includes a voice feature analysis unit, a social behavior quantification unit, and a progress tracking unit; The speech feature analysis unit is used to automatically analyze the speech data in the verified training data, extract features such as speech rate, average sentence length, and accuracy of specific phonemes, and output speech ability change trend data. The social behavior quantification unit is used to analyze children's training videos in the validated training data using computer vision technology, quantify the frequency of eye contact, the number of times of joint attention and the number of times of emotional response, and output a social interaction ability score. The progress tracking unit is used to calculate and output a compliance index based on the completion frequency, task interruption rate, and preset training plan achievement rate in the verified training data. The completion frequency is the ratio of the number of standardized training periods actually completed within a preset period to the preset number of standardized training periods. The task interruption rate is the ratio of the number of task interruptions in a single standardized training period to the total number of tasks. The preset training plan achievement rate is the ratio of the amount of training content actually completed to the preset amount of training content.

[0014] The assessment is broken down into three core modules: language ability, social interaction behavior, and compliance index, which achieves a more refined assessment: the speech feature analysis unit quantifies language ability from core dimensions such as pronunciation and expression, providing accurate feedback for language intervention; the social behavior quantification unit uses computer vision technology to automate the quantification of social behavior, solving the shortcomings of traditional social assessment that relies on subjective observation; and the compliance and progress tracking unit objectively reflects the implementation of family interventions, providing a basis for adjusting the intervention plan. The three work together to form a multi-dimensional assessment system, making the assessment of intervention effectiveness more targeted and persuasive.

[0015] Furthermore, the child interactive terminal is a tablet computer or robot equipped with a dedicated interactive application, and integrates a microphone and a front-facing camera for presenting the content of the training task package, as well as a data acquisition module for collecting children's voice data, operational behavior data, and reaction time data and synchronizing them to the server.

[0016] The microphone and front-facing camera ensure the collection of voice data, facial expressions, and behavioral data, providing raw data for subsequent verification and evaluation modules; the two implementation forms, tablet and robot, adapt to different family scenarios and children's needs, improving the system's applicability and flexibility.

[0017] Furthermore, the parent terminal is a smartphone with a parent application installed, whose functions include at least receiving and viewing personalized training task packages, shooting and uploading children's training videos, receiving visual assessment reports and system prompts; the system prompts include training-related reminders, data upload-related prompts, assessment report update prompts, and training operation guidance prompts.

[0018] Receiving personalized training task packages provides parents with clear intervention guidance, lowering the operational threshold for family intervention; shooting and uploading children's training videos enables convenient collection of training scenario data; receiving visual assessment reports and system prompts allows parents to keep abreast of changes in their children's abilities and key intervention points, enhancing parents' sense of participation and confidence in the intervention process, and further ensuring the continuity and effectiveness of the intervention.

[0019] Furthermore, the expert management terminal is used by rehabilitation therapists or researchers to view in batches the dynamic ability assessment data, verified training data, and visualized assessment reports of the children under their management. It can also manually calibrate parameters such as the difficulty matching coefficient and data weight of the personalized recommendation model, and supplement or replace and update the guidance videos, interactive levels, and reinforcement animations in the multimedia resource library.

[0020] The batch viewing function improves the work efficiency of professionals, enabling them to manage the intervention process of multiple children simultaneously; the functions of manually calibrating the parameters of the personalized recommendation model and updating the resource library can optimize the intervention plan based on professional experience, avoid the limitations of algorithm recommendations, make the intervention content more in line with the special needs of children with autism spectrum disorder, and promote the continuous iteration and optimization of the system's intervention plan.

[0021] Furthermore, it also includes an early warning and prompting module, which automatically generates early warning information and pushes it to the expert management terminal when any indicator in the voice ability change trend data or social interaction ability score stagnates or declines for a preset number of consecutive evaluation cycles, or when the compliance index is lower than a preset compliance threshold.

[0022] A risk warning mechanism for the intervention process is established to address the problem of untimely problem detection in traditional family interventions: by continuously monitoring changes in ability indicators and compliance indices, situations where the intervention is ineffective or poorly implemented can be quickly identified; early warning information is promptly pushed to the expert management terminal, which facilitates the timely intervention of professionals to adjust the intervention plan, avoids delays in children's rehabilitation process due to continuous ineffective intervention, and improves the accuracy and timeliness of intervention. Attached Figure Description

[0023] Figure 1 This is a system module diagram of Embodiment 1 of the interactive language rehabilitation training system for children with autism spectrum disorder of the present invention. Detailed Implementation

[0024] The following detailed description illustrates the specific implementation method: Example 1 is basically as shown in the appendix. Figure 1As shown: An interactive language rehabilitation training system for children with autism spectrum disorder, including a user terminal and a server; the user terminal includes a parent terminal, a child interaction terminal, and an expert management terminal; the server includes a data acquisition module, a data verification module, a personalized training module, and an assessment module, and the system also includes an early warning and prompting module.

[0025] I. Definition of Core Preset Parameters Standardized training sessions: The preset frequency is once a day, and the preset duration is 40 minutes (which can be adjusted to 30-60 minutes through the expert management terminal).

[0026] Data verification thresholds: The preset tolerance range is ±5 minutes, the preset time deviation requirement is less than or equal to 1 second, the preset overlap condition is that the child's vocalization action and the audio data collection period overlap, the non-overlapping part is less than or equal to 1 second, the preset difference range is less than or equal to 2 seconds, and the preset duration difference requirement is less than or equal to 3 minutes.

[0027] Children's ability assessment grading thresholds: Language clarity score (maximum 100 points) is divided into three intervals based on preset threshold 1 (60) and preset threshold 2 (80) to determine vocabulary difficulty level. Level 1 corresponds to a language clarity score less than 60, Level 2 corresponds to a language clarity score greater than or equal to 60 and less than 80, and Level 3 corresponds to a language clarity score greater than or equal to 80 and less than or equal to 100. Cumulative vocabulary size is divided into three intervals based on preset vocabulary threshold 1 (50) and preset vocabulary threshold 2 (100) to determine sentence complexity, etc. Level 1 corresponds to a cumulative vocabulary of less than 50 words; Level 2 corresponds to a cumulative vocabulary of 50 or more words but less than 100 words; and Level 3 corresponds to a cumulative vocabulary of 100 or more words. The instruction compliance accuracy is divided into three intervals based on two preset accuracy thresholds: 70% and 90%, to determine the intensity of the interaction. Level 1 corresponds to an instruction compliance accuracy of less than 70%; Level 2 corresponds to an instruction compliance accuracy of 70% or more words but less than 90%; and Level 3 corresponds to an instruction compliance accuracy of 90% or more words but less than or equal to 100%.

[0028] Assessment and early warning parameters: The preset assessment period is 7 days, the preset number of assessments is 3, and the preset compliance threshold is 60 points (out of 100).

[0029] Standardized scales: The "Language Ability Assessment Scale for Children with Autism" and the "Social Communication Ability Assessment Scale" are used. Parents fill them out through a parent terminal, and the initial child ability assessment data is generated after review and confirmation by the expert management terminal. The scales need to complete basic information (child's age, ASD diagnosis level, intervention history) and specific sub-items (vocabulary recognition, pronunciation accuracy, instruction response, etc.). After expert review and approval, the initial child ability assessment data (language clarity score, cumulative vocabulary size, instruction compliance accuracy rate) is generated.

[0030] II. Interaction Flow of Each Module Data acquisition module: Receives training videos of children collected from the parent terminal and voice data, operation behavior data and reaction time data collected from the child interactive terminal; Parental terminal data collection: Parents register and log in via the parental terminal application on their smartphones (requires binding the child's identity information and uploading an ASD diagnostic report). During training, clicking the "Shoot Training Video" button automatically turns on the camera. Recording can be paused (pause time not exceeding 5 minutes, automatically ending recording after the timeout). After recording, the system automatically marks the timestamp (accurate to the second) and GPS location information, providing cropping and editing functions (keeping only valid training segments and deleting irrelevant parts) and compression upload options. If upload fails, the system displays a "Network error, please try again" message and supports resume upload (automatically resuming upload after network disconnection and reconnection). Data collection for children's interactive terminals: A tablet computer with a dedicated interactive application is used, integrating a high-definition microphone and a front-facing camera. Training tasks are presented with gamified themes such as "Animal Paradise" and "Fruit and Vegetable Recognition." Children participate in the interaction by touching the screen (supporting multi-touch) and responding with voice. Among them, voice data is collected in real time by the microphone to capture children's pronunciation and speech, and ambient noise is filtered by the built-in noise reduction algorithm. The data is stored in WAV format. Operation behavior data is recorded by recording screen click coordinates, gesture type (swipe, long press, double tap), and operation timestamp. The data is stored in JSON format. Reaction time data is the time from the completion of the task interface to the child's first response (click / pronunciation) (task response delay) and the time taken from the start to the completion / failure of a single level (time taken to complete a single stage).

[0031] Data validation module: Integrates and validates multi-terminal data from the same standardized training period (e.g., 15:00-15:40 on June 10, 2025). The validation process is executed in the order of spatiotemporal correlation validation, behavioral result logic validation, and data integrity validation. Spatiotemporal correlation verification: Verify whether the timestamp of the video uploaded by the parent terminal (15:00:03-15:40:21) and the task start time (15:00:00) of the child interactive terminal are within the preset tolerance range of ±5 minutes, whether the location information (GPS coordinate error less than or equal to 10 meters) is consistent with the device location of the child interactive terminal, and whether the device identifier (terminal serial number and application registration ID) matches. Behavioral result logic verification: Key behavioral frames (such as the child clicking the screen at 15:05:10 and the child making a sound at 15:10:30) in the video are extracted using a video frame extraction algorithm (using the OpenCV library) for verification: The deviation between the click action time (15:05:10) and the timestamp of the operation behavior data (15:05:10.8) is less than or equal to 1 second; the speech action period (15:10:30-15:10:32) overlaps with the speech data acquisition period (the difference between the beginning and end of the timestamp is less than or equal to 1 second); the difference between the reaction time data (task response delay of 1.2 seconds) and the execution time in the behavioral frame (1.5 seconds) is less than or equal to 2 seconds; Data integrity check: The difference between the actual video duration (40 minutes and 18 seconds) and the total time recorded by the child interactive terminal (40 minutes and 21 seconds) is less than or equal to 3 minutes; Training data that meets the above preset rules after verification is marked as verified training data and stored in the server database; if the data fails verification (such as spatiotemporal mismatch or excessive data disconnection), the system will push a "Training data verification failed, please re-shoot and upload" prompt to the parent terminal, and indicate the reason for the failure (such as "video timestamp does not match the training period").

[0032] Personalized training module: The Children's Ability Information Unit dynamically updates children's ability assessment data based on the most recently validated training data. The update rules include: Language clarity score is calculated by analyzing pronunciation accuracy through AI speech recognition API and combining it with a scoring model trained on manually labeled samples. The average score of 3 valid pronunciations is used to update the score after each training session; Cumulative vocabulary is calculated by counting the total number of words that children can pronounce voluntarily and correctly match the meaning. New words must be verified and confirmed in 2 different training sessions before being included; Instruction compliance accuracy is calculated by the ratio of the number of successfully executed instructions to the total number of instructions. The standard for successfully executing an instruction is completing the specified action or response within 10 seconds. The task difficulty setting model calculates the optimal challenge difficulty by inputting updated children's ability assessment data (language clarity 62 points, cumulative vocabulary of 50 words, and instruction compliance accuracy of 75%) and using a score mapping method. Vocabulary difficulty level: 62 points corresponds to level 2 (common basic vocabulary, such as "apple", "puppy", "cup"); Sentence complexity level: 50 corresponding intervals are level 2 (simple short sentences, such as "I want an apple" or "This is a puppy"); Interactive intensity: 75% corresponds to level two (medium-frequency interaction, with one interactive instruction every 5 minutes, including pronunciation repetition, action imitation, etc.). Training task package generation: The multimedia resource library is stored locally on the server (main storage) and in cloud backup, including a tutorial video (30-60 seconds long, MP4 format), interactive level materials (images, audio, animation), and reinforcement animations (5-10 seconds long). The content generation unit assembles the task package according to the following rules: Guided Videos: Matching vocabulary difficulty levels, including pronunciation demonstrations and mouth shape demonstrations (e.g., mouth shape for pronouncing the initial "p" in "apple"); Interactive Levels: 3-5 tiered levels, from low to high difficulty (e.g., Level 1: Voice repetition, Level 2: Picture and pronunciation matching, Level 3: Short sentence completion, Level 4: Contextual expression); Reinforcement Animations: Set according to children's preferences (parents fill in their child's favorite cartoon character during registration), automatically playing after completing a level, with hidden animations unlocked after accumulating 3 level completions; Task package push: After generation, it is pushed to the parent terminal (including training steps and precautions) and the child interactive terminal (directly loaded to start training). If the push fails, the system reserves the right to re-push within 24 hours.

[0033] Evaluation module: Speech Ability Assessment: MFCC (Mel-frequency cepstral coefficients) feature extraction and SVM (Support Vector Machine) classification algorithms were used to analyze speech data from the validated training dataset. Specifically, speech rate was measured by counting the number of effective words pronounced per minute (excluding repetitive, meaningless pronunciations, and pauses longer than 2 seconds), increasing from an initial 30 words / minute to 38 words / minute; average sentence length was calculated by measuring the average number of words in each complete expression (a complete expression is defined as a sentence containing a subject and predicate / object), increasing from an initial 1.5 words to 2.3 words; specific phoneme accuracy was measured by calculating the percentage of correct pronunciations for basic phonemes such as "p," "m," and "b" (referencing the *Standard for Chinese Phonetic Alphabets*), increasing from 60% to 72%; output results included speech ability trend data (weekly percentage improvement and cumulative improvement), embedded in a visual assessment report as a line graph. Social Interaction Ability Assessment: Training videos were analyzed using computer vision technology (face detection and key point tracking algorithms based on the OpenCV library). The assessment included: Eye Contact Frequency: the ratio of the duration the child gazed at the virtual character's eye area on the terminal screen to the total training time; a gaze duration greater than 0.5 seconds was considered valid contact. Joint Attention Count: the number of times the child followed an object pointed to or prompted by the virtual character and responded with gaze, touch, etc.; a response delay of less than or equal to 3 seconds was considered valid. Emotional Response Count: the number of times the child responded positively to reinforcement animations and virtual character interactions, such as smiling, clapping, and nodding, determined by a facial expression recognition algorithm. Output: Social interaction ability score (out of 100), calculated using a weighted average of eye contact (30%), joint attention (40%), and emotional response (30%). Compliance Index Calculation: Compliance Index = Completion Frequency * 40% + (1 - Task Interruption Rate) * 30% + Preset Training Plan Achievement Rate * 30%. The definitions of each indicator are as follows: Completion Frequency: The ratio of the number of standardized training sessions actually completed per week to the number of preset training sessions. The preset evaluation period is 7 days (one week), with a total of 7 preset training sessions. Task Interruption Rate: The ratio of the number of interruptions (voluntary withdrawal, prolonged distraction (greater than or equal to 10 seconds of no operation / response)) in a single training session to the total number of tasks. Preset Training Plan Achievement Rate: The ratio of the amount of training content actually completed to the amount of preset training content. The amount of training content actually completed is obtained by multiplying the number of levels by the completion rate. For example, if the completion frequency is 100% (7 / 7), the task interruption rate is 20% (2 / 10), and the preset training plan achievement rate is 80% (8 / 10), then the compliance index will be calculated as 88 points.

[0034] Early Warning and Notification Module: When the early warning trigger conditions are met (ability indicators stagnate / decline for 3 consecutive assessment cycles, or compliance index is below 60 points), the system generates and pushes early warning information according to the following process: Early Warning Information Generation: Includes abnormal indicator name, current value, screenshot of historical change curve, cause analysis (e.g., "stagnation in speech ability may be related to excessive training difficulty or insufficient family guidance"), and specific adjustment suggestions (e.g., "reduce vocabulary difficulty level to level one, add parent demonstration and guidance sessions, and watch a 5-minute parent guidance video before each training session"); Push Method: Prioritizes notification via application pop-up on the expert management terminal, and simultaneously sends SMS notification (linked to the rehabilitation therapist's mobile phone number); Intervention Follow-up: After the expert adjusts the parameters of the personalized recommendation model (e.g., threshold division) or the content of the training task package, the system pushes a "Intervention plan has been updated, please check the new training task" notification to the parent terminal, and focuses on tracking the changes in the adjusted indicators within the subsequent assessment cycle.

[0035] The difference between Example 2 and Example 1 is that the child interactive terminal uses an intelligent robot. The intelligent robot's mobility and multimodal interaction functions can simulate more realistic social scenarios, making the training process more immersive, and is especially suitable for children with autism spectrum disorder who have a need for physical interaction.

[0036] The above are merely embodiments of the present invention. Commonly known structures and characteristics are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. An interactive language rehabilitation training system for children with autism spectrum disorder, characterized by: It includes user terminals and servers, wherein the user terminals include parent terminals, child interaction terminals and expert management terminals; The server includes a data acquisition module, a data verification module, a personalized training module, and an evaluation module; The data acquisition module is used to receive children's training videos from the parent's terminal; It is also used to receive voice data, operation behavior data and reaction time data from children's interactive terminals. The voice data includes the pronunciation and sentence expression data of children when completing interactive training tasks. The operation behavior data includes the screen click and gesture interaction data of children when completing interactive training tasks. The reaction time data includes the task response delay and single-stage completion time data of children when completing interactive training tasks. The data verification module is used to integrate children's training videos, voice data, operational behavior data, and reaction time data within the same standardized training period to obtain training data, and to perform logical correlation verification, marking the verified training data as verified training data. The standardized training period refers to a training period that is conducted at a preset frequency and lasts for a preset duration each time. The personalized training module is used to output a personalized training task package based on the child's ability assessment data and the most recently validated training data. The personalized training task package includes a guide video, interactive levels, and reinforcer animations. The personalized training module is used to push the personalized training task package to the parent's terminal and the child's interactive terminal. The child's ability assessment data includes a language clarity score, cumulative vocabulary size, and instruction compliance accuracy rate. The child's ability assessment data is collected through a standardized scale and dynamically updated based on validated training data after an intervention event occurs. The intervention event is the first time the system pushes the training task package to the child's interactive terminal and initiates data collection. The language clarity score represents the recognizability of the child's pronunciation, the cumulative vocabulary size represents the total number of words that the child can actively pronounce and understand the meaning of, and the instruction compliance accuracy rate represents the child's correct execution rate of instructions. The assessment module is used to quantitatively assess children's language abilities, social interaction behaviors, and compliance index based on verified training data within a preset assessment period, and generate a visual assessment report.

2. The interactive language rehabilitation training system for children with autism spectrum disorder according to claim 1, characterized in that: The data verification module includes a spatiotemporal correlation verification unit, a behavior result logic verification unit, and a data integrity check unit; The spatiotemporal correlation verification unit is used to verify whether the timestamp and location information of the child training video reported by the parent terminal match the task start time and device identifier recorded by the child interactive terminal and whether they are within the preset tolerance range. The behavior result logic verification unit is used to determine whether the key behavior frames in the children's training video correspond to the corresponding task nodes recorded by the children's interactive terminal. The judgment criteria include whether the deviation between the time of the action occurrence of the key behavior frame and the timestamp of the operation behavior data meets the preset time deviation requirements, whether the collection time period of the child's pronunciation action and the voice data in the behavior frame meets the preset overlap conditions, and whether the difference between the reaction time data and the task execution time in the behavior frame is within the preset difference range. The data integrity checking unit is used to check whether the difference between the training duration corresponding to the child's training video in a single training data session and the total time recorded by the child's interactive terminal meets the preset duration difference requirement. The training data is marked as verified training data only if the verification results of the spatiotemporal correlation verification unit, the behavior result logic verification unit, and the data integrity check unit all meet the preset rules.

3. The interactive language rehabilitation training system for children with autism spectrum disorder according to claim 2, characterized in that: The personalized training module includes a child's ability information unit, a task difficulty setting model, and a content generation unit. The child ability information unit is used to dynamically update the child ability assessment data based on the most recently verified training data; The task difficulty setting model takes children's ability assessment data as input and outputs the optimal challenge difficulty of the recommended task. The optimal challenge difficulty includes vocabulary difficulty level, sentence complexity level, and interaction intensity. The vocabulary difficulty level is divided into three levels (Level 1, Level 2, and Level 3) based on language clarity score and preset score threshold. The sentence complexity level is divided into three levels (Level 1, Level 2, and Level 3) based on cumulative vocabulary size and preset vocabulary size threshold. The interaction intensity is divided into three levels (Level 1, Level 2, and Level 3) based on instruction compliance accuracy rate and preset accuracy rate threshold. The content generation unit is used to automatically assemble and generate a personalized training task package, including a guide video, interactive levels, and reinforcement animations, from a multimedia resource library based on the optimal challenge difficulty and preset training objectives.

4. The interactive language rehabilitation training system for children with autism spectrum disorder according to claim 3, characterized in that: The evaluation module includes a voice feature analysis unit, a social behavior quantification unit, and a progress tracking unit; The speech feature analysis unit is used to automatically analyze the speech data in the verified training data, extract features such as speech rate, average sentence length, and accuracy of specific phonemes, and output speech ability change trend data. The social behavior quantification unit is used to analyze children's training videos in the validated training data using computer vision technology, quantify the frequency of eye contact, the number of times of joint attention and the number of times of emotional response, and output a social interaction ability score. The progress tracking unit is used to calculate and output a compliance index based on the completion frequency, task interruption rate, and preset training plan achievement rate in the verified training data. The completion frequency is the ratio of the number of standardized training periods actually completed within a preset period to the preset number of standardized training periods. The task interruption rate is the ratio of the number of task interruptions in a single standardized training period to the total number of tasks. The preset training plan achievement rate is the ratio of the amount of training content actually completed to the preset amount of training content.

5. The interactive language rehabilitation training system for children with autism spectrum disorder according to claim 4, characterized in that: The child interactive terminal is a tablet computer or robot equipped with a dedicated interactive application, and integrates a microphone and a front-facing camera to present the content of the training task package, as well as a data acquisition module to collect children's voice data, operation behavior data and reaction time data and synchronize them to the server.

6. The interactive language rehabilitation training system for children with autism spectrum disorder according to claim 5, characterized in that: The parent terminal is a smartphone with a parent terminal application installed. Its functions include at least receiving and viewing personalized training task packages, shooting and uploading children's training videos, receiving visual assessment reports and system prompts. The system prompts include training-related reminders, data upload-related prompts, evaluation report update prompts, and training operation guidance prompts.

7. The interactive language rehabilitation training system for children with autism spectrum disorder according to claim 6, characterized in that: The expert management terminal is used by rehabilitation therapists or researchers to view in batches dynamic ability assessment data, verified training data, and visualized assessment reports of the children under their management. It can also manually calibrate parameters such as the difficulty matching coefficient and data weight of the personalized recommendation model, and supplement or replace and update the guidance videos, interactive levels, and reinforcement animations in the multimedia resource library.

8. The interactive language rehabilitation training system for children with autism spectrum disorder according to claim 7, characterized in that: It also includes an early warning and prompting module, which automatically generates early warning information and pushes it to the expert management terminal when any indicator in the voice ability change trend data or social interaction ability score stagnates or declines for a preset number of consecutive evaluation cycles, or when the compliance index is lower than a preset compliance threshold.

Images