Autistic children rehabilitation training management method and system based on artificial intelligence

Abstract

The present invention discloses an artificial intelligence-based rehabilitation training management method and system for autistic children, and relates to the technical field of autism rehabilitation training, including conducting rehabilitation training for autistic children through VR technology, collecting activity data of autistic children; judging the abnormal behavior of autistic children based on the collected data, setting an early warning mechanism, issuing a warning and terminating the training; calculating the matching degree between the attention of autistic children and the training content based on the collected data, and constructing a fusion feature matrix. The present invention uses VR technology to conduct rehabilitation training for autistic children, provides a richer and more interactive training experience for autistic children, uses a deep learning model, analyzes eye movement data and interaction point data, intuitively evaluates and judges the rehabilitation training effect of the current VR rehabilitation training on autistic children, and screens and adjusts the content of VR rehabilitation training, while making abnormal judgments.

Description

Technical Field

[0001] The present invention relates to the technical field of autism rehabilitation training, and in particular to an artificial intelligence-based rehabilitation training management method and system for autistic children. Background Art

[0002] Autism rehabilitation training aims to help autistic children improve their social, communication and behavioral abilities through a series of scientific and systematic training methods. Among them, the most advanced rehabilitation training method is to use VR technology for rehabilitation treatment. VR technology can simulate the real environment and provide autistic children with a richer and more interactive training experience.

[0003] However, the existing VR rehabilitation training process mostly relies on preset rehabilitation training content and lacks dynamic adjustment capabilities. It is impossible to optimize the training content according to the real-time performance of the children. After wearing the VR equipment, it is not easy to directly observe whether the children cooperate with the VR rehabilitation training program to interact. Therefore, there is a lack of effective observation and judgment methods for the cooperation of autistic children. In addition, many systems mainly focus on one-way visual and auditory stimulation, lack comprehensive analysis of multi-dimensional data, and cannot fully understand the training status and emotional changes of children. As a result, during the training process, it is impossible to timely identify and deal with children’s discomfort or abnormal reactions, which easily makes autistic children resist. Summary of the invention

[0004] In view of the problems existing in the above-mentioned existing artificial intelligence-based rehabilitation training management method and system for autistic children, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is that in the existing VR rehabilitation training process, most of them rely on preset rehabilitation training content, lack dynamic adjustment capabilities, and cannot optimize and adjust the training content according to the real-time performance of the children. After the VR device is worn, it is not easy to directly observe whether the children cooperate with the VR rehabilitation training program to interact, so there is a lack of effective observation and judgment methods for the cooperation of autistic children. In addition, many systems mainly focus on one-way stimulation of vision and hearing, lack comprehensive analysis of multi-dimensional data, and cannot fully understand the training status and emotional changes of children, resulting in the inability to timely identify and deal with children's discomfort or abnormal reactions during training, which easily causes autistic children to resist.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: an artificial intelligence-based rehabilitation training management method for autistic children, comprising:

[0007] Conduct rehabilitation training for autistic children through VR technology and collect activity data of autistic children;

[0008] Based on the collected data, we judge the abnormal behavior of autistic children, set up an early warning mechanism, issue a warning and terminate the training;

[0009] Based on the collected data, the matching degree between the attention of autistic children and the training content is calculated, and a fusion feature matrix is ​​constructed;

[0010] Set up a deep learning model to conduct comprehensive evaluation and abnormality judgment of VR rehabilitation training programs based on the fusion feature matrix, and adjust the VR rehabilitation training programs;

[0011] The collected data and judgment data of VR rehabilitation training are transmitted and saved.

[0012] As a preferred solution of the artificial intelligence-based rehabilitation training management method for autistic children of the present invention, the VR technology is used to conduct rehabilitation training for autistic children and collect activity data of autistic children, including:

[0013] Provide rehabilitation training for autistic children based on VR displays and supporting VR rehabilitation training programs;

[0014] Collect and identify eye movement data of children with autism through eye trackers;

[0015] Obtain the interaction point data of rehabilitation training content through the SDK of the VR display;

[0016] Collect facial images through facial acquisition equipment and identify facial expression change data of autistic children;

[0017] Collect heart rate and temperature data of autistic children through monitoring bracelets;

[0018] Collect children's voice data through microphones and voice recognizers;

[0019] The collected data are denoised and normalized.

[0020] As a preferred solution of the artificial intelligence-based rehabilitation training management method for autistic children of the present invention, wherein: the abnormal behavior of autistic children based on collected data includes:

[0021] Facial expression change data , heart rate data , body temperature data and sound feature data After standardization, we can obtain , , and ;

[0022] The eigenvalue calculation is expressed as:

[0023] ,

[0024] Where U is the total monitoring time, Expressed as a comprehensive eigenvalue;

[0025] Before conducting VR rehabilitation training, children's daily behavior data including facial expression change data, heart rate data, body temperature data and sound feature data are collected and feature value calculations are performed. Based on the professional judgment of the rehabilitation trainer, the analysis is performed and the threshold is set. ;

[0026] Based on the calculation of VR rehabilitation training program To judge abnormality, Less than threshold When the child's behavior is judged to be within the normal range, Greater than threshold When the child's behavior is judged to be abnormal.

[0027] As a preferred solution of the artificial intelligence-based rehabilitation training management method for autistic children of the present invention, the setting of an early warning mechanism, issuing a warning and terminating the training includes:

[0028] The system monitors comprehensive characteristic values ​​in real time , when it detects and determines that a child’s behavior is abnormal, the early warning mechanism is immediately triggered;

[0029] Sound an alarm through the speaker and display a red warning signal on the monitoring screen;

[0030] A vibration reminder is sent out through a vibration device worn by the guardian. After receiving the early warning signal, the guardian will check the child's condition in time to confirm whether it is necessary to adjust the training content or terminate the training.

[0031] As a preferred solution of the artificial intelligence-based rehabilitation training management method for autistic children of the present invention, wherein: the matching degree between the attention of autistic children and the training content is calculated based on the collected data, and the fusion feature matrix is ​​constructed, including:

[0032] The matching degree between eye movement data and VR interaction point data is calculated by the probability density function of two-dimensional Gaussian distribution, which is expressed as:

[0033] ,

[0034] in represents the matching degree at time step t, and is the eye movement data coordinates at time step t, and represents the coordinates of the interaction point at time step t, represents standard deviation;

[0035] Will With eye tracking data and interaction point data Combined with , forming a fusion feature matrix F.

[0036] As a preferred solution of the artificial intelligence-based rehabilitation training management method for autistic children described in the present invention, wherein: the setting of the deep learning model, the comprehensive evaluation and abnormality judgment of the VR rehabilitation training project based on the fusion feature matrix, and the adjustment of the VR rehabilitation training project include:

[0037] Build a deep learning model, including input layer, convolution layer, LSTM layer, attention mechanism layer, fully connected layer and output layer;

[0038] The input layer inputs the fused feature matrix F;

[0039] The convolution layer calculates the weighted sum through the sliding convolution kernel, introduces nonlinearity through the ReLU activation function, and outputs the output features of the tth time step;

[0040] The LSTM layer uses a gating mechanism to output the hidden state of the tth time step. ;

[0041] The attention mechanism layer calculates the attention weights for each time step ;

[0042] Comprehensive attention weights of fully connected layers , calculate the weighted hidden state sequence and determine the weight matrix of the fully connected layer;

[0043] Combining the above layers and the attention weight, the comprehensive evaluation output E is calculated and expressed as:

[0044] ,

[0045] in represents the sigmoid activation function, E represents the comprehensive evaluation output, T represents the total number of time steps, represents the bias of the fully connected layer, Represents the weight matrix of the fully connected layer;

[0046] Input the training set into the input layer of the deep learning model, use the mean square error as the loss function to calculate the loss, select the Adam optimizer to optimize the model parameters, use the back propagation algorithm and gradient descent method to update the model parameters, set the minimum value as the target threshold based on historical data, and stop the iteration when the iterative change is less than the target threshold;

[0047] Based on the VR rehabilitation training projects, the corresponding fusion feature matrix is ​​obtained to input the deep learning model, and a comprehensive evaluation output is performed on the VR rehabilitation training projects. A threshold is set based on the experience of child rehabilitation experts, and VR rehabilitation training projects below the threshold are marked as abnormal. VR rehabilitation training content with abnormal marks is removed from the VR rehabilitation training projects.

[0048] As a preferred solution of the artificial intelligence-based rehabilitation training management method for autistic children of the present invention, the data transmission and storage of the collected data and judgment data of VR rehabilitation training include:

[0049] The system records detailed data for each warning, including the comprehensive characteristic value that triggers the warning , trigger time, specific abnormal cause and warning type;

[0050] Children's daily behavior data, data collected from VR rehabilitation training programs, and comprehensive evaluation output data of the deep learning model are transmitted to the cloud server through the wireless network for storage and backup.

[0051] Another object of the present invention is to provide a system for autistic children rehabilitation training management method based on artificial intelligence, which comprises:

[0052] The data collection module collects various data of autistic children through a variety of sensors and devices;

[0053] Feature matrix building module, which fuses the collected data into a feature matrix for analysis and processing;

[0054] The deep learning model module uses a deep learning model to conduct comprehensive evaluation and abnormality determination of VR rehabilitation training programs based on the fusion feature matrix;

[0055] The training adjustment module adjusts the VR rehabilitation training program according to the output of the deep learning model;

[0056] The early warning mechanism module monitors the behavior of autistic children in real time, issues early warning signals and terminates training in a timely manner;

[0057] The data storage module transmits and saves the collected data and the judgment data.

[0058] A computer device comprises: a memory and a processor; the memory stores a computer program, and the processor implements the steps of the above-mentioned artificial intelligence-based rehabilitation training management method for autistic children when executing the computer program.

[0059] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned artificial intelligence-based rehabilitation training management method for autistic children.

[0060] The beneficial effects of the present invention are as follows: by adopting VR technology to carry out rehabilitation training for autistic children, a richer and more interactive training experience is provided for autistic children, and by using a deep learning model, the rehabilitation training effect of the current VR rehabilitation training content on autistic children is intuitively evaluated and judged through analysis of eye movement data and interaction point data, and the content of VR rehabilitation training is screened and adjusted, and abnormal judgments are made at the same time, and measures are taken in time to adjust the training content to avoid the occurrence of adverse reactions and prevent the rehabilitation training from causing excessive stimulation to children, and based on the early warning mechanism, a warning is issued and the training is terminated to prevent the autistic children from developing a resistance mentality. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] 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 only 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.

[0062] Figure 1 This is a flowchart of the rehabilitation training management method for autistic children based on artificial intelligence.

[0063] Figure 2 This is a structural diagram of the rehabilitation training management system for autistic children based on artificial intelligence. DETAILED DESCRIPTION

[0064] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0065] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0066] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0067] Example 1

[0068] Reference Figure 1, which is the first embodiment of the present invention, and provides a method for managing rehabilitation training for children with autism based on artificial intelligence. The method for managing rehabilitation training for children with autism based on artificial intelligence includes:

[0069] S1, conduct rehabilitation training for autistic children through VR technology and collect activity data of autistic children;

[0070] Preferably, the rehabilitation training of autistic children is carried out by VR technology, and the activity data of autistic children is collected, including:

[0071] Provide rehabilitation training for autistic children based on VR displays and supporting VR rehabilitation training programs;

[0072] Collect and identify eye movement data of children with autism through eye trackers;

[0073] Obtain the interaction point data of rehabilitation training content through the SDK of the VR display;

[0074] Through the facial acquisition device, the Intel RealSense D415 depth camera can be used to collect facial images and identify the facial expression change data of autistic children;

[0075] Collect heart rate and temperature data of autistic children through monitoring bracelets;

[0076] Collect children's voice data through microphones and voice recognizers;

[0077] The collected data are denoised and normalized.

[0078] By using VR displays and supporting VR rehabilitation training programs to conduct rehabilitation training for autistic children, we can simulate the real environment for autistic children based on cutting-edge VR technology, provide autistic children with a richer and more interactive training experience, and according to the symptoms of autistic children, rehabilitation experts can select the corresponding VR rehabilitation training content according to the condition of the disease to achieve customized treatment effects. Through such multi-dimensional data collection, it is possible to comprehensively and meticulously record the various behaviors and physiological reactions of children during the rehabilitation training process. Among them, the eye movement data and the interaction point data during VR interaction can be used to judge and analyze whether the corresponding VR rehabilitation training content attracts the attention of autistic children and generates interactive behaviors. Facial expression change data, heart rate and body temperature data, and sound data can also be used to judge the behavior of autistic children, providing a data basis for the subsequent adjustment of rehabilitation training content in combination with artificial intelligence.

[0079] S2, based on the collected data, determines the abnormal behavior of autistic children, sets up an early warning mechanism, issues a warning and terminates the training;

[0080] Preferably, judging the abnormal behavior of autistic children based on the collected data includes:

[0081] Facial expression change data , heart rate data , body temperature data and sound feature data After standardization, we can obtain , , and ;

[0082] The eigenvalue calculation is expressed as:

[0083] ,

[0084] Where U is the total monitoring time, Expressed as a comprehensive eigenvalue, standardization ensures that all data are compared and calculated in the same dimension, avoiding the error caused by direct comparison between data of different dimensions. ,Based on the large change range of sound feature data, logarithmic transformation can more effectively capture the change trend of data and reduce the interference of outliers. ,By introducing integral calculation, all data points in the entire training process are considered, so that the comprehensive feature value can fully reflect the performance of children in the training process. ,By introducing data fusion , so that the comprehensive eigenvalue can take into account different types of data and provide a holistic and comprehensive evaluation criterion;

[0085] Before conducting VR rehabilitation training, children's daily behavior data including facial expression change data, heart rate data, body temperature data and sound feature data are collected and feature value calculations are performed. Based on the professional judgment of the rehabilitation trainer, the analysis is performed and the threshold is set. ;

[0086] Based on the calculation of VR rehabilitation training program To judge abnormality, Less than threshold When the child's behavior is judged to be within the normal range, Greater than threshold When the child's behavior is judged to be abnormal.

[0087] By using multiple data collected in the early stage to calculate the characteristic values ​​and judge the current emotional state of autistic children in real time, the state of autistic children can be monitored in real time during VR rehabilitation training to prevent the content of VR rehabilitation training from stimulating the children without being noticed by caregivers in time, thereby preventing rehabilitation training from causing excessive stimulation to children, taking timely measures to adjust the training content, avoiding the occurrence of adverse reactions, and recording and analyzing the specific circumstances of abnormal behavior to provide a reference for subsequent training adjustments and optimizations, thereby achieving the beneficial effects of ensuring training safety and improving training results, making the rehabilitation training process safer, controllable and efficient.

[0088] Furthermore, an early warning mechanism is set up to issue a warning and terminate the training, including:

[0089] The system monitors comprehensive characteristic values ​​in real time , when it detects and determines that a child’s behavior is abnormal, the early warning mechanism is immediately triggered;

[0090] Sound an alarm through the speaker and display a red warning signal on the monitoring screen;

[0091] A vibration reminder is sent out through a vibration device worn by the guardian. After receiving the early warning signal, the guardian will check the child's condition in time to confirm whether it is necessary to adjust the training content or terminate the training.

[0092] By setting up an early warning mechanism, issuing warnings and terminating training, timely early warning and handling of potential risks in the training process can be achieved. Once abnormal behavior is detected or data exceeds the normal range, the system will immediately issue a warning signal and terminate training, realizing risk management and safety assurance in the training process. It can not only protect the safety of autistic children, but also improve the reliability and effectiveness of rehabilitation training, achieving the beneficial effects of ensuring children's safety and improving training effects, making the rehabilitation training process safer, more reliable and controllable, and ensuring that every training is conducted in a safe environment.

[0093] S3, based on the collected data, calculate the matching degree between the attention of autistic children and the training content, and construct a fusion feature matrix;

[0094] Preferably, the matching degree between the attention of autistic children and the training content is calculated based on the collected data, and the fusion feature matrix is ​​constructed, including:

[0095] The matching degree between eye movement data and VR interaction point data is calculated by the probability density function of two-dimensional Gaussian distribution, which is expressed as:

[0096] ,

[0097] in represents the matching degree at time step t, and is the eye movement data coordinates at time step t, and represents the coordinates of the interaction point at time step t, represents the standard deviation, describing the dispersion of the difference between the eye movement data and the VR interaction point coordinates around the mean;

[0098] Will With eye tracking data and interaction point data Combined with , forming a fusion feature matrix F.

[0099] Through the analysis of eye movement data and interaction point data, in-depth analysis and feature extraction of training data are achieved. In particular, for autistic children who lack social and communication skills, effective judgment can be made on whether VR rehabilitation training content attracts the attention of autistic children. Then, based on the rehabilitation training content that can attract autistic children, rehabilitation experts can further provide social and communication rehabilitation training for autistic children. This prevents the situation where experts cannot intuitively judge whether autistic children are attracted by the training content due to the wearing of VR display devices. In addition, the calculation of matching degree can be used as a basis for adjusting VR rehabilitation training content, which makes it convenient for experts to further select the same type of training content based on the rehabilitation training content that can attract children, thereby improving the efficiency of rehabilitation training.

[0100] S4, setting up a deep learning model, comprehensively evaluating and judging abnormalities of VR rehabilitation training programs based on the fusion feature matrix, and adjusting the VR rehabilitation training programs;

[0101] Preferably, a deep learning model is set up to comprehensively evaluate and determine abnormalities of VR rehabilitation training programs based on a fusion feature matrix, and the VR rehabilitation training programs are adjusted, including:

[0102] Build a deep learning model, including input layer, convolution layer, LSTM layer, attention mechanism layer, fully connected layer and output layer;

[0103] The input layer inputs the fusion feature matrix F;

[0104] The convolution layer calculates the weighted sum through the sliding convolution kernel and introduces nonlinearity through the ReLU activation function to output the output features of the tth time step. , expressed as:

[0105] ,

[0106] in represents the output feature of the kth convolution kernel at time step t, represents the i-th weight of the k-th convolution kernel, Represents the time step The fusion feature F, Represents the bias of the kth convolution kernel, 0 means that negative values ​​are set to 0 when the ReLU activation function is applied, and n represents the size of the convolution kernel;

[0107] The LSTM layer uses a gating mechanism, including input gate, forget gate, and output gate, to capture long-term and short-term feature changes and output the hidden state of the t-th time step. , expressed as:

[0108] ,

[0109] in represents the output at time step t, represents the sigmoid activation function, represents the weight matrix of the LSTM unit, represents the convolutional layer output features of LSTM input, represents the bias of the LSTM unit, Represents the hidden state at time step t-1. The hidden state contains the information summary of all time steps before the current time step.

[0110] The attention mechanism layer calculates the attention weights for each time step , expressed as:

[0111] ,

[0112] ,

[0113] in represents the attention weight at time step t, represents the attention score at time step t, represents the transpose of the attention mechanism weight vector, represents the weight matrix of the attention mechanism, represents the bias of the attention mechanism, T represents the total number of time steps, represents the attention score at time step k;

[0114] Comprehensive attention weights of fully connected layers , calculate the weighted hidden state sequence and determine the evaluation output of the fully connected layer, expressed as:

[0115] ,

[0116] Where y represents the evaluation output of the fully connected layer, represents the weight matrix of the fully connected layer, represents the bias of the fully connected layer;

[0117] Based on the weighted sum formula, the characteristic data at the time step is weighted and summed to obtain the calculated evaluation value, which is expressed as:

[0118] ,

[0119] in Represents the comprehensive evaluation value, represents the total number of time steps, represents the weight, Represents input features;

[0120] By introducing convolutional layers, LSTM layers and attention mechanisms, and combining activation functions and linear transformations, a new comprehensive evaluation formula E is constructed, which improves the complexity and accuracy of the model and can better reflect the effect of VR training content.

[0121] Combining the above layers and the attention weight, the comprehensive evaluation output E is calculated and expressed as:

[0122] ,

[0123] in represents the sigmoid activation function, E represents the comprehensive evaluation output, T represents the total number of time steps, represents the bias of the fully connected layer, Represents the weight matrix of the fully connected layer;

[0124] The output y of the fully connected layer is the result of linear transformation of the weighted hidden state sequence of all time steps. The comprehensive evaluation output E averages the results of each time step and introduces the attention weight , so that the characteristics of important time steps can be highlighted when calculating the comprehensive evaluation;

[0125] Input the training set into the input layer of the deep learning model, use the mean square error as the loss function to calculate the loss, select the Adam optimizer to optimize the model parameters, use the back propagation algorithm and gradient descent method to update the model parameters, set the minimum value as the target threshold based on historical data, and stop the iteration when the iterative change is less than the target threshold;

[0126] Based on the VR rehabilitation training projects, the corresponding fusion feature matrix is ​​obtained to input the deep learning model, and a comprehensive evaluation output is performed on the VR rehabilitation training projects. A threshold is set based on the experience of child rehabilitation experts, and VR rehabilitation training projects below the threshold are marked as abnormal. VR rehabilitation training content with abnormal marks is removed from the VR rehabilitation training projects.

[0127] By setting up a deep learning model, the VR rehabilitation training project is comprehensively evaluated and output and abnormal judgment is made based on the fusion feature matrix, and the training project is adjusted. Through model optimization and training, the model can intuitively evaluate and judge the rehabilitation training effect of the current VR rehabilitation training content on autistic children based on the collected data. The powerful data analysis and pattern recognition capabilities of the deep learning model are used to comprehensively and accurately evaluate the performance of children, and abnormal situations are discovered in a timely manner. Based on the comprehensive evaluation data, the content of VR rehabilitation training is quickly screened and adjusted based on the judgment of child rehabilitation experts. Based on VR rehabilitation training content with high comprehensive evaluation value, similar training content can be used for subsequent rehabilitation training projects. At the same time, the appearance of rehabilitation training content with low comprehensive evaluation output is reduced, further improving the rehabilitation training effect for autism.

[0128] S5, data transmission and storage of the collected data and judgment data of VR rehabilitation training;

[0129] Furthermore, data transmission and storage of the collected data and judgment data of VR rehabilitation training include:

[0130] The system records detailed data for each warning, including the comprehensive characteristic value that triggers the warning , trigger time, specific abnormal cause and warning type;

[0131] Children's daily behavior data, data collected from VR rehabilitation training programs, and comprehensive evaluation output data of the deep learning model are transmitted to the cloud server through the wireless network for storage and backup.

[0132] By transmitting the collected multi-dimensional data and analysis results and storing them securely on cloud servers, the integrity and security of the data are ensured. Its role is to ensure that all collected data and analysis results can be safely and completely saved through scientific data management, which is convenient for subsequent analysis, research and training adjustments. This can not only provide detailed data support for rehabilitation trainers, helping them to better formulate and adjust training plans, but also provide valuable data resources for scientific research, and promote the continuous optimization and improvement of autism rehabilitation training methods.

[0133] Example 2

[0134] Reference Figure 2 , which is the second embodiment of the present invention, and which is different from the previous embodiment, provides a system for autistic children rehabilitation training management method based on artificial intelligence, including:

[0135] The data collection module collects various data of autistic children through a variety of sensors and devices;

[0136] Feature matrix building module, which fuses the collected data into a feature matrix for analysis and processing;

[0137] The deep learning model module uses a deep learning model to conduct comprehensive evaluation and abnormality determination of VR rehabilitation training programs based on the fusion feature matrix;

[0138] The training adjustment module adjusts the VR rehabilitation training program according to the output of the deep learning model;

[0139] The early warning mechanism module monitors the behavior of autistic children in real time, issues early warning signals and terminates training in a timely manner;

[0140] The data storage module transmits and saves the collected data and the judgment data.

[0141] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0142] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0143] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0144] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit with a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit with a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0145] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for the management of rehabilitation training for children with autism based on artificial intelligence, characterized by: include, Conduct rehabilitation training for autistic children through VR technology and collect activity data of autistic children; Based on the collected data, we judge the abnormal behavior of autistic children, set up an early warning mechanism, issue a warning and terminate the training; Based on the collected data, the matching degree between the attention of autistic children and the training content is calculated, and a fusion feature matrix is ​​constructed; Set up a deep learning model to conduct comprehensive evaluation and abnormality judgment of VR rehabilitation training programs based on the fusion feature matrix, and adjust the VR rehabilitation training programs; Transmit and save the collected data and judgment data of VR rehabilitation training; The method of calculating the matching degree between the attention of autistic children and the training content based on the collected data and constructing a fusion feature matrix includes: The matching degree between eye movement data and VR interaction point data is calculated by the probability density function of two-dimensional Gaussian distribution, which is expressed as: ; in represents the matching degree at time step t, and is the eye movement data coordinates at time step t, and represents the coordinates of the interaction point at time step t, represents standard deviation; Will With eye tracking data and interaction point data Forming the fusion feature matrix ; The deep learning model is set up to comprehensively evaluate and determine abnormalities of VR rehabilitation training programs based on the fusion feature matrix, and the VR rehabilitation training programs are adjusted, including: Build a deep learning model, including input layer, convolution layer, LSTM layer, attention mechanism layer, fully connected layer and output layer; The input layer inputs the fusion feature matrix F; The convolution layer calculates the weighted sum of fusion features at different time steps through a sliding convolution kernel, introduces nonlinearity through the ReLU activation function, and outputs the output features of the tth time step; The LSTM layer uses a gating mechanism to input the output features of the convolutional layer into the LSTM layer and output the hidden state of the tth time step. ; The attention mechanism layer calculates the attention weights for each time step ; Comprehensive attention weights of fully connected layers , calculate the weighted hidden state sequence; Combining the above layers, the comprehensive evaluation output E is calculated and expressed as: ， in represents the sigmoid activation function, E represents the comprehensive evaluation output, T represents the total number of time steps, represents the bias of the fully connected layer, Represents the weight matrix of the fully connected layer; Input the training set into the input layer of the deep learning model, use the mean square error as the loss function to calculate the loss, select the Adam optimizer to optimize the model parameters, use the back propagation algorithm and gradient descent method to update the model parameters, set the minimum value as the target threshold based on historical data, and stop the iteration when the iterative change is less than the target threshold; Based on the VR rehabilitation training projects, the corresponding fusion feature matrix is ​​obtained to input the deep learning model, and a comprehensive evaluation output is performed on the VR rehabilitation training projects. A threshold is set based on the experience of child rehabilitation experts, and VR rehabilitation training projects below the threshold are marked as abnormal. VR rehabilitation training content with abnormal marks is removed from the VR rehabilitation training projects.

2. The artificial intelligence-based rehabilitation training management method for autistic children as claimed in claim 1, characterized in that: The method of conducting rehabilitation training for autistic children through VR technology and collecting activity data of autistic children includes: Provide rehabilitation training for autistic children based on VR displays and supporting VR rehabilitation training programs; Collect and identify eye movement data of children with autism through eye trackers; Obtain the interaction point data of rehabilitation training content through the SDK of the VR display; Collect facial images through facial acquisition equipment and identify facial expression change data of autistic children; Collect heart rate and temperature data of autistic children through monitoring bracelets; Collect children's voice data through microphones and voice recognizers; The collected data are denoised and normalized.

3. The artificial intelligence-based rehabilitation training management method for autistic children as claimed in claim 2, characterized in that: The determination of behavioral abnormalities of autistic children based on collected data includes: Facial expression change data , heart rate data , body temperature data and sound feature data After standardization, we can obtain , , and ; The eigenvalue calculation is expressed as: ， Where U is the total monitoring time, Expressed as a comprehensive eigenvalue; Before conducting VR rehabilitation training, children's daily behavior data including facial expression change data, heart rate data, body temperature data and sound feature data are collected and feature value calculations are performed. Based on the professional judgment of the rehabilitation trainer, the analysis is performed and the threshold is set. ; Based on the calculation of VR rehabilitation training program To judge abnormality, Less than threshold When the child's behavior is judged to be within the normal range, Greater than threshold When the child's behavior is judged to be abnormal.

4. The artificial intelligence-based rehabilitation training management method for autistic children as claimed in claim 3, characterized in that: The said setting of early warning mechanism, issuing warning and terminating training includes: The system monitors comprehensive characteristic values ​​in real time , when it detects and determines that a child’s behavior is abnormal, the early warning mechanism is immediately triggered; Sound an alarm through the speaker and display a red warning signal on the monitoring screen; A vibration reminder is sent out through a vibration device worn by the guardian. After receiving the early warning signal, the guardian will check the child's condition in time to confirm whether it is necessary to adjust the training content or terminate the training.

5. The artificial intelligence-based rehabilitation training management method for autistic children as claimed in claim 4, characterized in that: The data transmission and storage of the collected data and judgment data of VR rehabilitation training includes: The system records detailed data for each warning, including the comprehensive characteristic value that triggers the warning , trigger time, specific abnormal cause and warning type; Children's daily behavior data, data collected from VR rehabilitation training programs, and comprehensive evaluation output data of the deep learning model are transmitted to the cloud server through the wireless network for storage and backup.

6. An artificial intelligence-based rehabilitation training management system for autistic children, the system implements the artificial intelligence-based rehabilitation training management method for autistic children described in any one of claims 1-5, the system comprising: The data collection module uses a variety of sensors and devices to collect activity data of autistic children during their rehabilitation training using VR technology; Feature matrix building module, which fuses the collected data into a feature matrix for analysis and processing; The deep learning model module uses a deep learning model to conduct comprehensive evaluation and abnormality determination of VR rehabilitation training programs based on the fusion feature matrix; The training adjustment module adjusts the VR rehabilitation training program according to the output of the deep learning model; The early warning mechanism module monitors the behavior of autistic children in real time, issues early warning signals and terminates training in a timely manner; The data storage module transmits and saves the collected data and the judgment data.

7. A computer device comprising: Memory and processor; The memory stores a computer program, characterized in that when the processor executes the computer program, the steps of the artificial intelligence-based autism children rehabilitation training management method described in any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the artificial intelligence-based rehabilitation training management method for autistic children described in any one of claims 1 to 5 are implemented.

Images