A method, system, and storage medium for processing autism rehabilitation training data

Abstract

The present invention discloses a method, a system and a storage medium for processing autism rehabilitation training data, belonging to the field of autism rehabilitation. Electrocardiogram, electroencephalogram, voice and eye movement trajectory signals during the rehabilitation training of autistic patients are collected; heart rate variability, P wave and QRS complex are extracted from the electrocardiogram signal, time domain, frequency domain and time-frequency domain features are extracted from the electroencephalogram signal, statistical features related to prosody, voice quality and emotion are extracted from the voice signal, and the trajectory of fixation, the duration of fixation point and the switching frequency feature of fixation point are extracted from the eye movement trajectory signal; the features are selected or dimensionally reduced; the features are matched with emotion labels; the signals after preprocessing and feature extraction are input into a trained classifier, and the classifier outputs the real-time emotion state; according to the real-time emotion recognition result, the rehabilitation training plan is adjusted or corresponding emotion support is given, and the emotion of autistic children can be automatically recognized through multiple dimensions, so as to achieve automatic interaction.

Description

Technical Field

[0001] The present invention relates to the field of autism rehabilitation, and in particular to an autism rehabilitation training data processing method, system and storage medium. Background Art

[0002] Early intervention in developmental behavioral disorders in autistic children has been proven to have significant effects on improving the prognosis of children with these diseases. Patent CN104874087B discloses a handheld terminal system for rehabilitation training of autistic children, which discloses a rehabilitation training module and a picture operation module for autistic children to conduct cognitive communication training. During rehabilitation training, the children can express their ideas in combination with the operation of the picture operation module; the network module is used to upload the user's information, training plan and training progress to the server and then interact with the data. However, the above-mentioned handheld terminal system for rehabilitation training of autistic children must express its ideas through the single dimension of autistic children operating the picture operation module. When autistic children cannot operate the picture operation module, the rehabilitation training system cannot interact. Summary of the invention

[0003] In order to overcome the deficiencies of the prior art, one of the objectives of the present invention is to provide an autism rehabilitation training data processing method that can automatically identify the emotions of autistic children through multiple dimensions, thereby achieving automatic interaction.

[0004] In order to overcome the deficiencies of the prior art, a second object of the present invention is to provide an autism rehabilitation training data processing system that can automatically identify the emotions of autistic children through multiple dimensions, thereby achieving automatic interaction.

[0005] In order to overcome the deficiencies of the prior art, the third object of the present invention is to provide a storage medium that can automatically identify the emotions of autistic children through multiple dimensions, thereby realizing automatic interactive autism rehabilitation training data processing.

[0006] One of the purposes of the present invention is achieved by the following technical solution:

[0007] A method for processing autism rehabilitation training data comprises the following steps:

[0008] Data collection: Collect ECG signals, EEG signals, speech signals and eye movement trajectory signals of autistic patients during rehabilitation training. The ECG signals include P waves and QRS wave groups, the EEG signals include α waves, β waves, θ waves and δ waves, and the speech signals include the pitch, volume, speech speed and sound quality of the sound.

[0009] Data preprocessing: filtering, denoising, and removing baseline drift on the ECG signal; filtering, removing electrooculographic interference, and removing myoelectric interference on the EEG signal; and pre-emphasis, framing, and windowing on the speech signal;

[0010] Feature extraction: extracting heart rate variability, P wave, QRS complex from the ECG signal, extracting time domain features, frequency domain features and time-frequency domain features from the EEG signal, extracting rhythm features, sound quality features and emotion-related statistical features from the speech signal, and extracting the patient's gaze trajectory features, gaze point duration features, and gaze point switching frequency features from the eye movement trajectory signal;

[0011] Feature selection or dimensionality reduction: Select or reduce the extracted features to reduce redundant information;

[0012] Emotion model training: Use the labeled emotion dataset as training samples and match the extracted features with the emotion labels;

[0013] Real-time emotion recognition: The real-time ECG, EEG, speech and eye movement trajectory signals after preprocessing and feature extraction are input into the trained classifier, and the classifier outputs the real-time emotional state;

[0014] Feedback and adjustment: According to the real-time emotion recognition results, timely adjust the rehabilitation training plan or provide corresponding emotional support.

[0015] Furthermore, the feedback and adjustment steps are specifically as follows: when the patient's emotional state is low, the training and accompanying virtual digital human is called, and the digital human has an intelligent dialogue with the patient. At the same time, the digital human shows corresponding expressions and actions. During the intelligent dialogue between the digital human and the patient, the patient's language content, tone and speaking speed are analyzed to detect the patient's emotional state. When it is detected that the patient shows negative emotions such as depression, anxiety or uneasiness, the feedback mechanism of the hug vest is triggered to provide tactile feedback simulating a hug; when the patient's emotional state is high, the training and accompanying virtual digital human is turned off.

[0016] Furthermore, the hug vest can be personalized according to the patient's individual needs and preferences by adjusting the tightness, duration and frequency of the hug.

[0017] Furthermore, in the data collection step, the ECG signal is collected by contacting the skin with a physiological bracelet, the EEG signal is collected by wearing an EEG cap to make the electrodes conductive to the scalp, and the voice signal is collected by recording the patient's voice with a pickup or a wireless microphone.

[0018] Furthermore, in the feature extraction step, the time domain features are mean and variance, the frequency domain features are power spectral density and frequency band energy, the time and frequency domain features are wavelet transform coefficients, the rhythmic features are fundamental frequency, energy and duration, the sound quality features are resonance peaks and glottal parameters, and the emotion-related statistical features are the spectral distribution corresponding to the emotion label.

[0019] Furthermore, in the feature selection or dimensionality reduction step, principal component analysis or linear discriminant analysis is used for feature selection or dimensionality reduction.

[0020] Furthermore, in the real-time emotion recognition step, any one or more of a support vector machine, an artificial neural network, a decision tree or a deep learning model is used to train the classifier.

[0021] Furthermore, the autism rehabilitation training data processing method also includes an interest screening step, which is located after the feature selection or dimensionality reduction step. The interest screening step is specifically: identifying the patient's attention level and preference for different things based on the patient's gaze trajectory characteristics, gaze point duration characteristics, and gaze point switching frequency characteristics, so as to provide personalized recommendations for virtual interactive objects in the rehabilitation training process; the gaze trajectory characteristics, gaze point duration characteristics, and gaze point switching frequency characteristics can also be used to judge the real-time interest level in the rehabilitation training process and realize scene adaptive switching.

[0022] The second object of the present invention is achieved by adopting the following technical solution:

[0023] A system for processing autism rehabilitation training data is used to implement any one of the above-mentioned methods for processing autism rehabilitation training data. The system comprises a processor, a physiological bracelet, an electroencephalogram cap, and a voice acquisition device. The physiological bracelet, the electroencephalogram cap, and the voice acquisition device are communicatively connected with the processor. The physiological bracelet collects electrocardiogram signals, the electroencephalogram cap collects electroencephalogram signals, and the voice acquisition device collects voice signals. The processor analyzes the electrocardiogram signals, the electroencephalogram signals, and the voice signals to output a real-time emotional state and timely adjusts the rehabilitation training plan or provides corresponding emotional support according to the real-time emotional state recognition result.

[0024] The third object of the present invention is achieved by adopting the following technical solution:

[0025] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements any of the above-mentioned autism rehabilitation training data processing methods.

[0026] Compared with the prior art, the autism rehabilitation training data processing method of the present invention collects the electrocardiogram signal, electroencephalogram signal, voice signal and eye movement trajectory signal of the autism patient during the rehabilitation training process, wherein the electrocardiogram signal includes P wave and QRS wave group, the electroencephalogram signal includes α wave, β wave, θ wave and δ wave, and the voice signal includes the pitch, volume, speaking speed and sound quality of the sound; the electrocardiogram signal is filtered, denoised and baseline drift is removed, the electroencephalogram signal is filtered, electrooculogram interference is removed and electromyography interference is removed, and the voice signal is pre-emphasized, framed and windowed; heart rate variability, P wave and QRS wave group are extracted from the electrocardiogram signal, time domain features, frequency domain features and time-frequency domain features are extracted from the electroencephalogram signal, and the time domain features, frequency domain features and time-frequency domain features are extracted from the electrocardiogram signal. Extract rhythmic features, sound quality features and emotion-related statistical features from the speech signal; extract the patient's gaze trajectory features, gaze point duration features, and gaze point switching frequency features from the eye movement trajectory signal; select or reduce the extracted features to reduce redundant information; use the labeled emotion data set as a training sample to match the extracted features with the emotion label; input the real-time electrocardiogram, electroencephalogram, speech signal and eye movement trajectory signal after preprocessing and feature extraction into the trained classifier, and the classifier outputs the real-time emotional state; according to the real-time emotion recognition results, timely adjust the rehabilitation training plan or provide corresponding emotional support. Through the above steps, the emotions of autistic children can be automatically identified through multiple dimensions, thereby realizing automatic interaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 The figure is a flow chart of the autism rehabilitation training data processing method of the present invention. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] It should be noted that when a component is referred to as being "fixed to" another component, it may be directly on the other component or there may be another intermediate component through which it is fixed. When a component is considered to be "connected to" another component, it may be directly connected to the other component or there may be another intermediate component at the same time. When a component is considered to be "set on" another component, it may be directly set on the other component or there may be another intermediate component at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0031] See also Figure 1 , a method for processing autism rehabilitation training data, comprising the following steps:

[0032] Data collection: Collect ECG signals, EEG signals, speech signals and eye movement trajectory signals of autistic patients during rehabilitation training. ECG signals include P wave and QRS wave group, EEG signals include α wave, β wave, θ wave and δ wave, and speech signals include voice pitch, volume, speech speed and sound quality.

[0033] Data preprocessing: filtering, denoising, and removing baseline drift on ECG signals; filtering, removing electrooculographic interference, and removing myoelectric interference on EEG signals; and pre-emphasis, framing, and windowing on speech signals;

[0034] Feature extraction: extract heart rate variability, P wave, QRS complex from ECG signals; extract time domain features, frequency domain features, and time-frequency domain features from EEG signals; extract rhythmic features, sound quality features, and emotion-related statistical features from speech signals; extract the patient's gaze trajectory features, gaze point duration features, and gaze point switching frequency features from eye movement trajectory signals;

[0035] Feature selection or dimensionality reduction: Select or reduce the extracted features to reduce redundant information;

[0036] Emotion model training: Use the labeled emotion dataset as training samples and match the extracted features with the emotion labels;

[0037] Real-time emotion recognition: The real-time ECG, EEG, speech and eye movement track signal input signals after preprocessing and feature extraction are input into the trained classifier, and the classifier outputs the real-time emotional state;

[0038] Feedback and adjustment: According to the real-time emotion recognition results, timely adjust the rehabilitation training plan or provide corresponding emotional support.

[0039] In the data collection step, the ECG signal is collected by contacting the physiological bracelet with the skin, the EEG signal is collected by wearing an EEG cap to make the electrodes and scalp conductive, and the voice signal is collected by recording the patient's voice through a microphone or a wireless microphone. Specifically, autistic patients wear a physiological bracelet on their wrists, which contacts the patient's body surface through skin electrodes to collect electrical signals generated by ECG activity. ECG signals mainly include waveforms such as P waves and QRS complexes, which reflect the electrophysiological activity of the heart. The changes in ECG signals have a certain correlation with emotional states, such as heart rate variability (HRV), which can be used as an indicator of emotional state. Autistic patients wear EEG caps, which have multiple electrodes arranged on the scalp to capture the electrical activity of neurons in the cerebral cortex through electrodes. EEG signals contain rich information and can reflect the brain's advanced functions such as cognition and emotion. Under different emotional states, the frequency, amplitude and distribution pattern of brain waves will be different, such as α waves, β waves, θ waves, δ waves, etc. The microphone or wireless microphone collects the patient's voice signal to ensure that the recording is clear and noise-free. Speech signals include voice pitch, volume, speaking speed, sound quality and other characteristics. Speech signals contain rich emotional information, and different emotional states will lead to changes in speech characteristics. For example, when angry, the voice may be higher-pitched and the speaking speed may be faster, while when sad, the voice may be lower-pitched and the speaking speed may be slower.

[0040] The eye movement trajectory signal is achieved by wearing an eye tracker, which can capture the patient's eye movement trajectory in real time. As an important physiological parameter, eye movement trajectory data can reflect the patient's interest points in multiple dimensions. Through eye tracking technology, data such as the patient's gaze trajectory, the duration of the gaze point, and the switching frequency of the gaze point can be obtained. These data help to identify the patient's attention and preference for different things during the interest screening process, so as to provide personalized recommendations for virtual interactive objects during rehabilitation training. For example, if the patient shows a high interest in objects of a certain color or shape, then more such objects can be presented during rehabilitation training to improve the patient's participation and rehabilitation effect. Eye movement trajectory data can also be used to determine the patient's real-time interest level during rehabilitation training. By analyzing the patient's eye movement behavior, we can understand their level of involvement and interest in the current rehabilitation task. If the patient shows boredom or loses interest in the task, the system can adjust the rehabilitation scene or task in time to maintain the patient's enthusiasm and participation. This scene adaptive switching helps to improve the flexibility and effectiveness of rehabilitation training.

[0041] As an important physiological parameter, eye movement data can not only reflect the patient's interests in multiple dimensions, but also play a key role in emotional judgment. The data such as the patient's gaze trajectory, gaze point duration, and gaze point switching frequency obtained by eye tracking technology provide a strong basis for in-depth analysis of the patient's emotional state. Specifically, the gaze point duration in the eye movement data can reveal the patient's emotional response to specific stimuli. For example, when a patient shows a long gaze on an object or scene, this may indicate that they have a strong emotional resonance or interest in the object or scene. On the contrary, short gazes or frequent gaze point switching may indicate that the patient feels anxious, uneasy, or unable to concentrate. In addition, the overall pattern of eye movement trajectories can also provide important clues about the patient's emotional state. For example, a disordered and irregular eye movement pattern may indicate that the patient is in a tense, anxious, or confused emotional state. An orderly and smooth eye movement pattern may indicate that the patient feels relaxed, confident, or focused. Therefore, during rehabilitation training, by analyzing the characteristics of the patient's eye movement trajectory data, we can more accurately judge their emotional state and adjust the rehabilitation strategy in time. For example, when patients show anxiety or restlessness, we can provide a gentler and more soothing rehabilitation environment or tasks; and when patients show positive and focused emotions, we can appropriately increase the difficulty or complexity of rehabilitation training to promote further recovery of patients. This emotion judgment method based on the characteristics of eye movement data is expected to provide more personalized and accurate support for patients' rehabilitation training.

[0042] In the data preprocessing step, the EEG signal is filtered, the electrooculographic interference is removed, and the electromyographic interference is removed before segmentation and downsampling.

[0043] In the feature extraction step, the time domain features are mean and variance, the frequency domain features are power spectral density and frequency band energy, the time and frequency domain features are wavelet transform coefficients, the rhythmic features are fundamental frequency, energy, and duration, the sound quality features are resonance peaks and glottal parameters, and the emotion-related statistical features are the spectral distribution corresponding to the emotion label.

[0044] In the feature selection or dimensionality reduction step, the extracted features are selected or reduced in dimension to reduce redundant information and improve recognition efficiency. Principal component analysis or linear discriminant analysis is used for feature selection or dimensionality reduction.

[0045] In the real-time emotion recognition step, a classifier is trained using any one or more of a support vector machine, artificial neural network, decision tree, or deep learning model. The classifier outputs real-time emotional states such as happy, angry, sad, etc. The performance of the classifier is checked regularly, and the model is updated and optimized as needed.

[0046] The specific feedback and adjustment steps are as follows: when the patient's emotional state is low, the training and accompanying virtual digital human is called, and the digital human has an intelligent conversation with the patient. At the same time, the digital human shows corresponding expressions and actions. During the intelligent conversation between the digital human and the patient, the patient's language content, tone and speaking speed are analyzed to detect the patient's emotional state. When it is detected that the patient shows negative emotions such as depression, anxiety or uneasiness, the feedback mechanism of the hug vest is triggered to provide tactile feedback simulating a hug; when the patient's emotional state is high, the training and accompanying virtual digital human is turned off.

[0047] The Hug Vest is inflated to form an air bag that wraps around the patient, simulating the tactile feedback of a real hug. This design is intended to provide patients with a sense of warmth, comfort and soothing, especially when they are in a low emotional state or need emotional support. The Hug Vest is closely integrated with the intelligent dialogue system of the accompanying digital human, allowing the digital human to automatically trigger a hug feedback when it recognizes that the patient is in a low emotional state. This combination not only provides emotional support, but also enhances the interactivity and humanization between the digital human and the patient. The accompanying digital human monitors the patient's emotional state in real time through integrated sensors and algorithms. This includes facial expression analysis, voice recognition, and monitoring of physiological signals such as heart rate and skin conductance. When the accompanying digital human has an intelligent conversation with the patient, it analyzes the patient's language content, tone and speed of speech to assess his or her emotional state. If the robot detects that the patient is showing negative emotions such as depression, anxiety or uneasiness, it immediately triggers the feedback mechanism of the Hug Vest. The Hug Vest can be personalized according to the individual needs and preferences of the patient. For example, the patient can choose the tightness, duration and frequency of the hug. This personalized setting enables the Hug Vest to provide more intimate and effective emotional support. By providing tactile feedback that simulates hugging, the Hug Vest helps improve the patient's emotional state and mental health. It can reduce negative emotions such as anxiety and depression, and enhance the patient's sense of security and happiness. The application of the Hug Vest makes the companion robot not only a tool for providing information and guidance, but also a partner that can provide emotional support and companionship. This humanized design enhances the emotional connection and interaction between the robot and the patient. When patients feel understood and supported during rehabilitation training, they are more likely to actively participate in training and stick to it. The application of the Hug Vest helps create a more positive and warm rehabilitation environment, thereby promoting the patient's recovery process.

[0048] Through the above steps, the patient's real-time emotional state can be accurately identified based on ECG, EEG, voice and eye movement trajectory data signals, providing more personalized and scientific guidance for rehabilitation training.

[0049] The present application also discloses an autism rehabilitation training data processing system for implementing the above-mentioned autism rehabilitation training data processing method. The autism rehabilitation training data processing system includes a processor, a physiological bracelet, an EEG cap, and a voice acquisition device. The physiological bracelet, the EEG cap, and the voice acquisition device are communicatively connected to the processor. The physiological bracelet collects electrocardiogram (ECG) signals, the EEG cap collects electroencephalogram (EEG) signals, and the voice acquisition device collects voice signals. The processor analyzes the ECG signals, EEG signals, and voice signals to output the real-time emotional state and adjusts the rehabilitation training plan in a timely manner or provides corresponding emotional support according to the real-time emotional state recognition result. The voice acquisition device is a pickup or a wireless microphone.

[0050] The present application also discloses a computer-readable storage medium having a computer program stored thereon, and when the program is executed by a processor, it implements the above-mentioned autism rehabilitation training data processing method.

[0051] Compared with the prior art, an autism rehabilitation training data processing method of the present invention collects ECG signals, EEG signals, and voice signals during the rehabilitation training process of autistic patients. The ECG signals include P waves and QRS complexes. The EEG signals include alpha waves, beta waves, theta waves, and delta waves. The voice signals include pitch, volume, speech rate, and voice quality of the sound. The ECG signals are filtered, denoised, and baseline drift removed. The EEG signals are filtered, electrooculogram interference removed, and electromyogram interference removed. The voice signals are pre-emphasized, framed, and windowed. Heart rate variability, P waves, and QRS complexes are extracted from the ECG signals. Time-domain features, frequency-domain features, and time-frequency domain features are extracted from the EEG signals. Prosody features, voice quality features, and emotion-related statistical features are extracted from the voice signals. The extracted features are selected or dimensionally reduced to reduce redundant information. A labeled emotion data set is used as a training sample, and the extracted features are matched with emotion labels. The real-time ECG, EEG, and voice signals after preprocessing and feature extraction are input into a trained classifier, and the classifier outputs the real-time emotional state. According to the real-time emotion recognition result, the rehabilitation training plan is adjusted in a timely manner or corresponding emotional support is provided. Through the above steps, the emotions of autistic children can be automatically recognized through multiple dimensions, thereby realizing automatic interaction.

[0052] The above embodiments only represent several implementation manners of the present invention, and the description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several deformations and improvements can be made. These are all equivalent modifications and evolutions of the above embodiments based on the essence of the present invention, and all belong to the protection scope of the present invention.

Claims

1. A method for processing autism rehabilitation training data, characterized in that: The following steps are involved: Data collection: Collect ECG signals, EEG signals, speech signals and eye movement trajectory signals of autistic patients during rehabilitation training. The ECG signals include P waves and QRS wave groups, the EEG signals include α waves, β waves, θ waves and δ waves, and the speech signals include the pitch, volume, speech speed and sound quality of the sound. Data preprocessing: filtering, denoising, and removing baseline drift on the ECG signal; filtering, removing electrooculographic interference, and removing myoelectric interference on the EEG signal; and pre-emphasis, framing, and windowing on the speech signal; Feature extraction: extract heart rate variability, P wave, QRS complex from the ECG signal, extract time domain features, frequency domain features and time-frequency domain features from the EEG signal, extract rhythmic features, sound quality features and emotion-related statistical features from the speech signal, extract the patient's gaze trajectory features, the duration features of the gaze point, and the switching frequency features of the gaze point from the eye movement trajectory signal; the time domain features are mean and variance, the frequency domain features are power spectral density and frequency band energy, the time-frequency domain features are wavelet transform coefficients, the rhythmic features are fundamental frequency, energy and duration, the sound quality features are resonance peaks and glottal parameters, and the emotion-related statistical features are the spectrum distribution corresponding to the emotion label; Feature selection or dimensionality reduction: Select or reduce the extracted features to reduce redundant information; Emotion model training: Use the labeled emotion dataset as training samples and match the extracted features with the emotion labels; Real-time emotion recognition: The real-time ECG, EEG, speech and eye movement trajectory signals after preprocessing and feature extraction are input into the trained classifier, and the classifier outputs the real-time emotional state; Feedback and adjustment: According to the real-time emotion recognition results, the rehabilitation training plan is adjusted in time or corresponding emotional support is given. The feedback and adjustment steps are specifically as follows: when the patient's emotional state is low, the training and accompanying virtual digital human is called, and the digital human has an intelligent dialogue with the patient. At the same time, the digital human shows corresponding expressions and actions. During the intelligent dialogue between the digital human and the patient, the patient's language content, tone and speaking speed are analyzed to detect the patient's emotional state. When it is detected that the patient shows negative emotions of depression, anxiety or uneasiness, the feedback mechanism of the hug vest is triggered to provide tactile feedback of simulated hug; when the patient's emotional state is high, the training and accompanying virtual digital human is turned off. The hug vest can be personalized according to the patient's individual needs and preferences, which is achieved by adjusting the tightness, duration and frequency of the hug.

2. The autism rehabilitation training data processing method according to claim 1, characterized in that: In the data collection step, the ECG signal is collected by contacting the skin with a physiological bracelet, the EEG signal is collected by wearing an EEG cap to make the electrodes and scalp conductive, and the voice signal is collected by recording the patient's voice with a pickup or a wireless microphone.

3. The autism rehabilitation training data processing method according to claim 1, characterized in that: In the feature selection or dimensionality reduction step, principal component analysis or linear discriminant analysis is used for feature selection or dimensionality reduction.

4. The autism rehabilitation training data processing method according to claim 1, characterized in that: In the real-time emotion recognition step, any one or more of a support vector machine, an artificial neural network, a decision tree or a deep learning model is used to train the classifier.

5. The autism rehabilitation training data processing method according to claim 1, characterized in that: The autism rehabilitation training data processing method also includes an interest screening step, which is located after the feature selection or dimensionality reduction step. The interest screening step is specifically: identifying the patient's attention level and preference for different things based on the patient's gaze trajectory characteristics, gaze point duration characteristics, and gaze point switching frequency characteristics, so as to provide personalized recommendations for virtual interactive objects in the rehabilitation training process; the gaze trajectory characteristics, gaze point duration characteristics, and gaze point switching frequency characteristics can also be used to judge the real-time interest level in the rehabilitation training process and realize scene adaptive switching.

6. An autism rehabilitation training data processing system, used to implement the autism rehabilitation training data processing method according to any one of claims 1 to 5, characterized in that: The autism rehabilitation training data processing system includes a processor, a physiological bracelet, an EEG cap, and a voice acquisition device. The physiological bracelet, the EEG cap, and the voice acquisition device are communicatively connected to the processor. The physiological bracelet collects electrocardiogram signals, the EEG cap collects EEG signals, and the voice acquisition device collects voice signals. The processor analyzes the electrocardiogram signals, the EEG signals, and the voice signals to output a real-time emotional state and timely adjusts the rehabilitation training plan or provides corresponding emotional support according to the real-time emotional state recognition result.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the autism rehabilitation training data processing method as described in any one of claims 1 to 5 is implemented.

Images