Learning system and method based on brain-computer interface and AI mode

Abstract

The invention provides a learning system and method based on a brain-computer interface and an AI mode, and the system carries out the reinforcement and upgrading of an old person or infant learning machine based on the brain-computer interface and the AI mode, monitors the activity information of an old person or an infant in combination with the brain-computer interface, analyzes the electroencephalogram information of the old person or the infant in combination with an AI algorithm, and recognizes the activity demands of the old person or the infant. The interaction tasks and activities of the elderly or the children are recommended, so that a humanized interaction and accompanying relationship with the elderly or the children is achieved; the system can analyze senior citizens or children's emotions to provide corresponding emotional values and psychological tutoring analysis recommendation conclusions, realizes senior citizens activities and ecological prevention in combination with a brain-computer interface, realizes interaction ability improvement and optimization, upgrades the learning machine functions, and makes the learning machine more intelligent.

Description

Technical Field

[0001] This invention relates to the field of brain-computer interface and artificial intelligence technology, and in particular to a learning system and method, electronic device and computer-readable storage medium based on brain-computer interface and AI mode. Background Technology

[0002] The core challenge for paralyzed elderly or young children in recognizing needs lies in their loss of primary communication channels (verbal expression and independent movement), and the fact that their alternative signals are ambiguous, variable, and easily misinterpreted. Specifically, this includes: I. Core Dilemma: Complete or partial blockage of communication channels Language loss / lack: Paralyzed elderly people may completely lose their language ability due to stroke, neurodegenerative diseases (such as late-stage ALS, aphasia, etc.). Even if they retain some language, articulation disorders may lead to unclear pronunciation, incoherent speech, and difficulty in being understood.

[0003] Infants and toddlers: They have not yet mastered language skills, or are in the early stages of language development (such as under 1 year old or with developmental delays), and can only express themselves through basic means (crying, humming, simple syllables).

[0004] Severe limitations in nonverbal communication: Loss of physical activity: Unable to clearly indicate the desired object or direction through conventional means such as pointing, gestures, nodding / shaking the head, or physical contact.

[0005] Interpreting facial expressions is difficult: paralysis or damage to the nervous system may lead to abnormal facial muscle control, and young children's expressions are relatively simple and basic, so interpreting their meaning relies heavily on guesswork and experience.

[0006] Limited eye contact: Physical reasons may prevent you from making stable eye contact or following someone.

[0007] II. Specific challenges in identification: the ambiguity and complexity of demand signals Diverse demands, homogeneous signals: Whether it's physical discomfort (pain, itching, cold, heat, thirst, hunger, need to eliminate), psychological discomfort (fear, anxiety, irritability, boredom, loneliness), or simply a desire for companionship, the initial expressions (such as crying, restless moaning, or unusual quiet) can be very similar. It can be difficult to distinguish whether it's a "stomach ache" or a "wanting to look out the window."

[0008] The concealment and delay of vital signs: Internal needs are difficult to detect: internal sensations such as thirst, mild hunger, early constipation, localized pain (such as early pressure sores), and low-grade fever may not be obvious or may be delayed in external manifestation.

[0009] Slowed bodily feedback: Paralysis can damage sensory nerves (especially in diabetic paralysis patients), causing changes in pain threshold; infants and young children are also unable to accurately perceive or locate discomfort. Discomfort may only manifest itself in a stronger reaction after it has worsened.

[0010] Fluctuations in expressive ability: The strength and clarity of the signals emitted may vary depending on factors such as mental state, level of fatigue, stage of illness, and medication. What can be expressed with a specific sound today may be silent again tomorrow.

[0011] Dependence on "proxy communicators" and deviations in the communication chain: For individuals who are completely unable to express their needs proactively, their communication relies on caregivers' active and continuous observation, interpretation, and feedback confirmation. This process is lengthy, with significant signal attenuation, making misunderstandings highly likely. A toddler's crying might be interpreted as "hunger," when in reality it could be due to "an uncomfortable diaper" or "a room that is too hot."

[0012] III. Subjective Interference Factors in the Identification Process Differences in caregiver experience and knowledge: Different caregivers may interpret the same signal (such as a gasp or a slight twitch of facial muscles) differently. Lack of expertise or knowledge of a particular individual's habits can increase the risk of misinterpretation.

[0013] Limitations of observation frequency and coverage: 24-hour uninterrupted observation is not possible. Early or weak signals are easily missed at night, during shift changes, and during busy periods. A single oversight could mean hours of patience and accumulated risk.

[0014] Easily confused with cognitive impairment / emotional problems: Elderly people with dementia and paralysis: Cognitive impairment itself can cause behaviors such as irritability and apathy, which are difficult to distinguish from behaviors caused by unmet needs (pain, discomfort). Painful reactions may be mistaken for "dementia-induced agitation".

[0015] Preschool children / children with special needs: Developmental delays or autistic tendencies may be accompanied by emotional and behavioral issues without obvious reasons, increasing the difficulty of recognizing effective signals.

[0016] IV. Consequences and risks amplified by failure to identify Physical risks: Untreated pain: Pain caused by bedsores, muscle spasms, urinary tract infections, gastrointestinal problems (such as intestinal obstruction) can persist for a long time and increase suffering.

[0017] Insidious disease development: Unrecognized discomfort may lead to the aggravation of symptoms such as pneumonia and infection when ignored.

[0018] Delay in meeting basic survival needs: hunger, dehydration, difficulty breathing, and obstructed excretion (severe constipation) are not identified and addressed, posing serious health threats.

[0019] Psychological and emotional harm: Collapse of security: Neglected or improperly met needs plunge individuals into a deep abyss of powerlessness and helplessness, resulting in a loss of a sense of existence and dignity.

[0020] Despair and behavioral withdrawal: Long-term inability to express oneself may lead to giving up on expressing oneself, turning into indifference, closure or extreme resistance, forming a "needs black hole" (i.e., becoming accustomed to a state of no response).

[0021] Increased anxiety and fear: Unpredictable needs-fulfilling processes can trigger persistent fear and unease.

[0022] Reduced quality of care: Errors or delays in treatment negatively impact recovery and quality of life.

[0023] Frequent misjudgments of needs can also lead to frustration and exhaustion for caregivers, creating a vicious cycle of mutual torment.

[0024] Therefore, understanding the needs of paralyzed elderly people / children is like deciphering a cryptic, multifaceted code in a dark room. Without direct verbal communication, the most convenient path, one must rely on identifying flickering, intermittent signals, searching for clues amidst numerous distractions. The core challenges can be summarized as follows: physical isolation of communication channels; diverse potential needs coupled with limited and singular modes of expression; fragile and volatile signals; and a highly subjective and lengthy identification process—each step potentially prone to misunderstanding and missed opportunities.

[0025] The emergence of brain-computer interfaces (BCIs) may solve the aforementioned difficulties; however, current BCI applications lack effective technical solutions for achieving effective interaction among the elderly or young children. Most existing BCIs are consciousness recognition devices, merely identifying the user's brainwave signal characteristics and feeding back potential needs, then notifying caregivers to prepare corresponding activities or items. This not only requires constant supervision but also fails to provide more precise activity recommendations, making caregiving neither effort nor time-saving.

[0026] Although learning machines can bring some interaction to the elderly or young children, existing learning machines cannot recognize the needs of the elderly or young children and actively recommend interactive tasks or activities. They do not have dynamic interactive functions, so they are not very user-friendly for users such as the elderly with paralysis / dementia and young children. Summary of the Invention

[0027] To address the technical problems existing in the prior art, the present invention provides the following technical solution: On the one hand, a learning system based on brain-computer interface and AI mode is provided, the system including an EEG module and a learning machine, wherein: The EEG module is used to collect EEG information from the elderly or young children based on the brain-computer interface and upload it to the learning machine. The learning machine is used to perform EEG feature recognition and analysis on the EEG information, output corresponding dimension demand information, and recommend interactive tasks with corresponding demand information; and / or, run interactive tasks with corresponding demand information based on the digital human to interact with the elderly or children; and / or, edit and generate speculative text with corresponding demand information based on AI automatic editing technology. The EEG module and the learning machine are communicatively connected.

[0028] Preferably, the EEG module uses the ErgoAI high-precision non-invasive EEG headband (8-channel EEG + near-infrared imaging), with a sampling rate of 256Hz, and supports wireless data transmission.

[0029] Preferably, the learning machine includes: A data preprocessing system is used to receive and preprocess the EEG information, including: (1) Signal cleaning process: 0.5-45Hz bandpass filter (FIR design) to remove electromyographic interference; Independent component analysis (ICA) separates eye movement artifacts; Automatic outlier removal based on ASR (Artifact Subspace Reconstruction); (2) Timing alignment: Dynamic time warping (DTW) is used to compensate for individual wear differences.

[0030] Preferably, the learning machine further includes: The EEG analysis module is used to extract time-frequency features from the EEG information through wavelet transform, including: energy ratios of the θ (4-8Hz), α (8-13Hz), or β (13-30Hz) bands; and to identify the time-frequency features and output corresponding dimension demand information through a pre-trained demand recognition model. The method for generating the demand identification model is as follows: Collect EEG information from several elderly people or young children, as well as demand information in corresponding dimensions based on the EEG information; The time-frequency features of the EEG information are extracted by wavelet transform and bound to the corresponding dimension of demand information. Analyze the time-frequency characteristics of each elderly person or child and construct a feature set; The feature set is divided into training set, test set and validation set according to the proportion; The training set is imported into the initial LSTM model for feature training and learning to generate the initial demand recognition model. The recognition performance of the requirement recognition model was tested and verified using a test set and a validation set, respectively. If the test passes, the requirement identification model will be deployed and applied.

[0031] Preferably, the learning machine further includes: The interactive task generation module is used to generate corresponding interactive tasks for the elderly or children based on the aforementioned requirements information. A digital twin system is used to provide digital human services based on digital twin technology and to generate a set of digital human interaction instructions adapted to the interactive tasks. The digital human projection module is used to project virtual images of digital humans, execute the digital human interaction instruction set, and drive the digital human to perform corresponding interactive actions to interact with the elderly or young children. A voice device is used to provide voice playback services for the digital human projection module during digital human operation; The facial expression control module is used to provide facial expression control services for the digital human projection module during digital human operation; The interactive sensing device is used to sense and feed back the interactive operation data of the elderly or young children to the digital twin system during the interaction between the elderly or young children and the digital human. The digital twin system then drives the digital human to dynamically adjust the corresponding interactive actions based on the interactive operation data.

[0032] Preferably, the learning machine further includes: The AI ​​editing module is used to identify the needs of the elderly or young children, and based on AI automatic editing technology, it generates predictive text messages corresponding to these needs, including: Memoir generation module: used to extract key memory points marked by alpha wave event-related potentials (ERPs) in elderly people, and use GPT-4o combined with spatiotemporal features to generate narrative text; Children's cognitive generation module: Visualizes gamma wave activity using a diffusion model and generates corresponding knowledge-related graphics and text.

[0033] Preferably, the learning machine further includes: The emotion intervention module is used to assess the corresponding anxiety index based on the time-frequency characteristics of the EEG information. Anxiety index = (β right frontal lobe - β left frontal lobe) / total β energy If the anxiety index is not within the preset threshold range, a corresponding emotion intervention strategy is generated and sent to the expression control module, which then regulates the digital human's expressions.

[0034] On the other hand, a learning method based on brain-computer interface and AI mode is provided. This method is used to implement the aforementioned learning system based on brain-computer interface and AI mode. The method includes: Wearing an EEG module, baseline tests are conducted to establish a personal EEG feature database in different dimensions; After the test is completed, the brain-computer interface is used to collect the electroencephalogram (EEG) information of the elderly or young children and upload it to the learning machine. The learning machine performs EEG feature recognition and analysis on the EEG information, outputs corresponding dimension demand information, and recommends interactive tasks with corresponding demand information; And / or, Based on the digital human's ability to perform interactive tasks that meet specific needs, it can interact with the elderly or young children. And / or, It uses AI-powered automatic editing technology to edit and generate speculative copy containing relevant demand information.

[0035] On the other hand, an electronic device is provided, comprising: a processor; and a memory storing computer-readable instructions, wherein when the computer-readable instructions are executed by the processor, the learning method based on brain-computer interface and AI mode described above is implemented.

[0036] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, the at least one instruction being loaded and executed by a processor to implement the above-described learning method based on brain-computer interface and AI mode.

[0037] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least: This invention enhances and upgrades learning machines for the elderly or young children based on brain-computer interfaces and AI. It combines brain-computer interface monitoring of activity information with AI algorithms to analyze brainwave information, identify activity needs, and associate these needs with the learning machine to recommend corresponding interactive tasks and activities. This fosters a humanized interaction and companionship with the elderly or children. Furthermore, it analyzes the emotions of the elderly or children, providing corresponding emotional value and psychological counseling recommendations. Combined with brain-computer interfaces, this achieves a positive ecosystem for elderly activity and the prevention of Alzheimer's disease. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used 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 those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a schematic diagram of the topology of a learning system based on brain-computer interface and AI mode provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the functional architecture of a learning machine provided in an embodiment of the present invention; Figure 3 This is a flowchart of a model training and generation process provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0040] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0041] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as an "exemplary" in the present invention should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of the word "exemplary" is intended to present concepts in a concrete manner. Furthermore, in the embodiments of the present invention, "and / or" can mean both or either of the two.

[0042] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0043] In this embodiment of the invention, sometimes a subscript such as W1 may be mistakenly written as a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0044] In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0045] This system is primarily intended for elderly or paralyzed patients, especially the elderly. It is also suitable for young children who are unable to express their needs.

[0046] This invention provides a learning system based on a brain-computer interface and AI mode. This method can be implemented by an electronic device, which can be a terminal or a server. Figure 1 The diagram shows the topology of a learning system based on a brain-computer interface and AI model. The system includes an EEG module and a learning machine, wherein: The EEG module is used to collect EEG information from the elderly or young children based on the brain-computer interface and upload it to the learning machine. The learning machine is used to perform EEG feature recognition and analysis on the EEG information, output corresponding dimension demand information, and recommend interactive tasks with corresponding demand information; and / or, run interactive tasks with corresponding demand information based on the digital human to interact with the elderly or children; and / or, edit and generate speculative text with corresponding demand information based on AI automatic editing technology. The EEG module and the learning machine are communicatively connected.

[0047] This invention, based on brain-computer interfaces (BCIs) and AI, enhances and upgrades learning devices for the elderly and young children. It combines BCI monitoring of activity information with AI algorithms to analyze brainwave information, identify activity needs, and associate these needs with the learning device, recommending corresponding interactive tasks and activities. It also provides digital human-like functionality, enabling human-to-human interaction. Users can create audio, video, and text content multiple times within the system using BCI and AI technology. For example, it can identify the elderly's BCI information and automatically edit and compose memories using AI, or identify children's cognitive information about the world and automatically edit cognitive text using AI. The system implements multiple interactive methods, promoting humanized interaction and companionship among the elderly and young children. It can also analyze the emotions of the elderly and young children, providing corresponding emotional value and psychological counseling analysis recommendations. This invention can effectively prevent age-related diseases, achieving a positive ecological prevention of geriatric diseases.

[0048] The system of this invention is a learning system built on brain-computer interface and AI mode, mainly composed of EEG module and learning mechanism, and its interaction principle and technical function are as follows: 1. Interaction Principle: EEG Data Acquisition: The EEG module utilizes brain-computer interface technology to collect EEG data from the elderly or young children. This is the source for the system to understand the user's inner needs. By capturing the electrical signals generated by brain activity, it provides raw data for subsequent analysis and processing.

[0049] Information Analysis and Task Recommendation: After receiving EEG information uploaded from the EEG module, the learning machine performs EEG feature recognition and analysis. It extracts key features from the EEG information and uses algorithms to interpret the needs of the elderly or young children in different dimensions. For example, it might identify the elderly's need to recall past events, or the young child's interest in a particular type of story. Subsequently, based on this information, it recommends corresponding interactive tasks, such as recommending nostalgic-themed interactions for the elderly and related story interactions for young children.

[0050] Digital Human Interaction: The learning machine can also perform interactive tasks corresponding to the above-mentioned needs based on the digital human. It can directly interact with the elderly or young children using the image and functions of a digital human. The digital human can communicate with the elderly or young children through various forms such as voice and gestures, meeting their interactive needs. For example, the digital human can tell stories to young children with a friendly image and voice, or accompany the elderly to recall past experiences, simulating interpersonal interaction scenarios.

[0051] Inference Text Generation: The learning machine uses AI-powered automatic editing technology to edit and generate inference text based on user needs. For example, if it detects a young child's interest in space exploration, it can generate simple science text about space exploration; or, for an elderly person's nostalgia for a certain period in the past, it can generate relevant reminiscing text, further enriching the user experience.

[0052] 2. Technological advantages: Personalized services: Through precise collection and analysis of EEG information from the elderly or young children, the system can provide highly personalized services to meet their needs. Whether it's interactive task recommendations or digital human interaction, it's all based on the unique EEG characteristics of each individual, meeting the specific needs of each user and enhancing the relevance and satisfaction of the user experience.

[0053] Intelligent Interactive Experience: Digital humans enable interactive tasks, creating a more vivid and natural interactive environment for the elderly and young children. The voice, movements, and other expressive forms of the digital humans enhance the fun and appeal of the interaction, helping to improve the emotional well-being of the elderly and the learning enthusiasm of young children.

[0054] Rich and Expanded Content: The predictive texts generated by AI-powered automatic editing technology not only provide users with additional information but also stimulate the thinking of the elderly and young children. For example, the science-related texts generated for young children can broaden their knowledge, while the reminiscent texts generated for the elderly can help deepen their emotional experience, enriching the user experience in the learning system from multiple perspectives.

[0055] The principle will be described in further detail below.

[0056] Preferably, the EEG module uses the ErgoAI high-precision non-invasive EEG headband (8-channel EEG + near-infrared imaging), which supports wireless data transmission, collects EEG information (sampling rate 256Hz) and uploads it to the learning machine (for example, receiving EEG information through the wireless port of the learning machine and sending it to the data storage system, and binding it to the current elderly or child's identity ID).

[0057] The EEG module mainly consists of the following parts and operates according to this principle: Signal acquisition equipment: The ErgoAI high-precision non-invasive EEG headband is used, which features 8-channel EEG (electroencephalography) and near-infrared imaging capabilities. The 8-channel EEG can acquire weak electrical signals generated by neuronal activity in the brain from multiple different locations, reflecting the activity of different functional areas of the brain. Near-infrared imaging uses near-infrared light to illuminate the brain, utilizing the differences in the absorption and scattering characteristics of near-infrared light by brain tissue to obtain physiological information such as changes in blood oxygenation within the brain. The combination of these two methods provides a more comprehensive reflection of the brain's activity state.

[0058] Data transmission module: Supports wireless data transmission. This allows data collected by the EEG headband to be transmitted wirelessly, freeing it from the constraints of traditional wired connections. This greatly facilitates the user's activities, improves the flexibility and convenience of device use, and makes it easy to collect EEG information in everyday life scenarios.

[0059] Signal Acquisition and Processing Flow: EEG information is acquired at a sampling rate of 256Hz. The sampling rate determines the number of times the EEG signal is acquired per unit time. A sampling rate of 256Hz can accurately capture changes in the EEG signal, ensuring that the acquired data has high temporal resolution, providing a foundation for subsequent accurate analysis of brain activity. The acquired EEG information is uploaded to the learning machine wirelessly, for example, the learning machine receives the EEG information through its own wireless port.

[0060] Data storage and identity binding: After receiving EEG information, the learning machine sends it to the data storage system and simultaneously binds it to the current elderly person's or child's identity ID. By binding the identity ID, the EEG data of different individuals can be distinguished and managed, facilitating subsequent data analysis for specific individuals, tracking changes in their brain function, and contributing to applications such as personalized health monitoring and cognitive ability assessment.

[0061] like Figure 2 The diagram shows the various functional systems / modules developed and deployed by this invention on an existing learning machine application system. The specific integration into the existing learning machine application system can be achieved through industry software upgrades, and necessary hardware configurations such as brain-computer interface, output, and interactive connection projection devices can be configured according to requirements.

[0062] Preferably, the learning machine includes: A data preprocessing system is used to receive and preprocess the EEG information, including: Signal cleaning process: 1) 0.5-45Hz bandpass filtering (FIR design) to remove electromyographic interference; The processing procedure is as follows: A finite impulse response (FIR) filter is designed, with its passband range set to 0.5-45Hz. In digital signal processing, this filter only allows signal components with frequencies between 0.5 and 45Hz to pass through, while attenuating signal components above 45Hz and below 0.5Hz. During bioelectrical signal acquisition, electromyographic interference often has high-frequency components. This bandpass filter effectively removes most of this interference, resulting in a purer target signal.

[0063] Advantages: FIR filters have linear phase characteristics, meaning that the phase relationship between frequency components remains unchanged after the signal passes through the filter, and phase distortion does not occur. This provides a more reliable base signal for subsequent analysis and processing based on signal phase information, such as feature extraction of certain neural electrical signals. Furthermore, by precisely setting the passband range, the target signal frequency band can be selectively preserved, minimizing the impact of electromyographic interference on subsequent analysis.

[0064] 2) Independent component analysis (ICA) to separate eye movement artifacts; Processing Procedure: ICA is a blind source separation technique that assumes the observed signal is a linear mixture of multiple independent source signals. When processing bioelectrical signals, the multi-channel signal containing eye-movement artifacts is input into the ICA algorithm as the observed signal. The ICA algorithm iteratively optimizes to find a set of separation matrices that maximizes the independence between the separated components. After ICA processing, the eye-movement artifact components originally mixed in the bioelectrical signal are separated and can thus be removed from the target signal.

[0065] Advantages: ICA effectively separates independent components from mixed signals without prior knowledge of the specific characteristics or distribution of the source signal. For eye-movement artifacts, which are complex in origin and difficult to completely remove using traditional filtering methods, ICA has a unique advantage. It adaptively separates eye-movement artifacts based on the statistical independence of the signal, effectively removing artifacts without losing key features of the target signal, improving signal quality, and facilitating accurate subsequent analysis of real physiological signals.

[0066] 3) Automatic outlier removal based on ASR (Artifact Subspace Reconstruction); Processing steps: First, feature extraction is performed on the acquired signal, such as calculating statistical characteristics like mean, variance, and peak value. Then, a subspace model of the normal signal is constructed based on these features. In actual signal processing, the real-time acquired signal segment is compared with the constructed normal signal subspace, and its projection error in the subspace is calculated. If the projection error exceeds a pre-set threshold, the signal segment is determined to be an abnormal segment. Once an abnormal segment is identified, it is automatically removed from the signal sequence to ensure that the signals processed subsequently are relatively normal and reliable.

[0067] Advantages: The ASR method can automatically detect and remove outliers, eliminating the need for manual signal inspection and significantly improving signal preprocessing efficiency. Furthermore, by constructing a subspace model based on signal features, it can adapt well to variations in individual signals, exhibiting strong robustness. Automatically removing outliers avoids misleading subsequent signal analysis results from anomalous data, ensuring that conclusions drawn from signal analysis are more accurate and reliable.

[0068] (2) Timing alignment: Dynamic time warping (DTW) is used to compensate for individual wear differences.

[0069] Processing Procedure: Dynamic Time Warping (DTW) is an algorithm used to compare the similarity of two time series along the time axis. When dealing with signal timing inconsistencies caused by individual wear differences, a reference individual's signal is used as the template sequence, and the signals of other individuals are used as sequences to be aligned. The DTW algorithm finds an optimal nonlinear mapping relationship on the time axis of the two sequences, minimizing the distance between them (usually measured by Euclidean distance). Specifically, a two-dimensional matrix is ​​constructed, where the rows and columns correspond to the time points of the two sequences. The value of each element in the matrix represents the distance between the signals at the corresponding time points. Then, starting from the beginning of the matrix, a path with the minimum cumulative distance is found using dynamic programming. This path determines the time warping relationship between the two sequences, thus achieving time alignment between the individual signal and the reference signal.

[0070] Advantages: The DTW algorithm effectively handles signal timing shifts caused by individual wearing differences. It is independent of specific frequency or phase characteristics of the signal and has good applicability to various signal types. By finding the optimal time warping path, it preserves the original feature information of the signal to the greatest extent, making the time-aligned signals from different individuals comparable in the time dimension. This provides a unified time benchmark for subsequent comprehensive analyses based on multiple individual signals, such as group physiological signal feature extraction and comparison of different individual physiological states, helping to improve the accuracy and reliability of the analysis results.

[0071] Preferably, the learning machine further includes: The EEG analysis module is used to extract time-frequency features from the EEG information through wavelet transform, including: energy ratios of the θ (4-8Hz), α (8-13Hz), or β (13-30Hz) bands; and to identify the time-frequency features and output corresponding dimension demand information through a pre-trained demand recognition model. The working principle of the EEG analysis module is as follows: Time-frequency feature extraction: Wavelet transform is used to process EEG information. Wavelet transform is a mathematical tool that can analyze signals simultaneously in the time and frequency domains, suitable for processing non-stationary signals, including EEG signals. This transform extracts time-frequency features of specific frequency bands from the EEG information, specifically involving the energy ratios of the theta (4-8Hz), alpha (8-13Hz), or beta (13-30Hz) bands. Different frequency bands of EEG signals are often associated with different brain activities or states. For example, theta waves are often associated with drowsiness and meditation, alpha waves with relaxation and wakefulness, and beta waves are more common during periods of focused attention and active thinking. Extracting the energy ratios of these frequency bands quantifies brain activity in different states, providing a data foundation for subsequent needs identification.

[0072] Demand Information Recognition: A pre-trained demand recognition model is used to analyze and process the extracted time-frequency features. This demand recognition model may be built based on technologies such as machine learning and deep learning (LSTM model is preferred in this invention). During training, the model learns a large number of mapping relationships between EEG time-frequency features and corresponding demand information. When the input is the time-frequency features extracted by wavelet transform, the model identifies these features based on its learned knowledge, and then outputs demand information of the corresponding dimension. This demand information can be information of various dimensions related to user behavior, intentions, emotions, etc., such as whether the user currently wants to rest, focus on work, or is in a certain emotional state, providing key decision-making basis for subsequent applications based on EEG signals.

[0073] The typical training and learning steps for an LSTM model are as follows: 1. Data preparation: Data Collection: Collect task-related sequence data. For example, in speech recognition tasks, collect a large number of speech samples and their corresponding text transcriptions; in stock price prediction, collect historical stock prices and related market indicator data. The scale and quality of the data are crucial to the model training effect; large-scale and high-quality data enable the model to learn richer and more accurate patterns.

[0074] Data preprocessing: Preprocessing the collected data. This typically includes data cleaning, removing noise, errors, or irrelevant data. For text data, word segmentation and word vector representation may also be required (e.g., using methods like Word2Vec or GloVe to convert words into vectors); for image sequence data, normalization and cropping may be necessary. The purpose of preprocessing is to transform the data into a format suitable for model input and reduce interference factors in the data, thereby improving the model's learning efficiency.

[0075] Dataset partitioning: Divide the preprocessed data into training, validation, and test sets. Generally, the training set is used to learn model parameters, the validation set is used to tune model hyperparameters (such as learning rate, number of layers, number of hidden units, etc.), and the test set is used to evaluate the model's performance on unseen data. A common partition ratio is 70% training set, 15% validation set, and 15% test set, but the specific ratio can be adjusted based on the amount of data and the characteristics of the task.

[0076] 2. Model Building: Determine the model architecture: Based on the task requirements, determine the specific architecture of the LSTM model. This includes determining the number of LSTM layers and the number of hidden units in each LSTM layer. For example, for simple time series prediction tasks, only one or two LSTM layers may be needed, with each layer containing tens to hundreds of hidden units; while for complex natural language processing tasks, multiple LSTM layers with a larger number of hidden units may be required. Additionally, consider adding other layers before and after the LSTM layers, such as an embedding layer in the input layer to process word vector representations of the text data, and an output layer that selects an appropriate activation function based on the task type (e.g., a linear activation function for regression tasks and a softmax activation function for classification tasks).

[0077] Initialize parameters: Initialize the weights and biases in the model. Common initialization methods include random initialization (such as using a Gaussian or uniform distribution), and there are also some more advanced initialization methods, such as Xavier initialization or Kaiming initialization. These methods help to better propagate gradients during training, avoid gradient vanishing or exploding problems, and thus accelerate model convergence.

[0078] 3. Model Training: Define the loss function: Choose an appropriate loss function based on the task type. For regression tasks, the mean squared error (MSE) loss function is commonly used, which measures the average squared error between the predicted and actual values. Its formula is: ,in It is the actual value. Here, (n) represents the predicted value, and (n) represents the number of samples. For classification tasks, the cross-entropy loss function is commonly used; it measures the difference between the predicted probability distribution and the true probability distribution. For example, in multi-class classification tasks, the formula for the cross-entropy loss function is: , where (C) is the number of categories, and (y{ij}) is the true probability (usually 0 or 1) that sample (i) belongs to category (j). It is the probability that the model predicts that sample (i) belongs to category (j).

[0079] Optimizer Selection: An optimizer is chosen to update the model's parameters. Common optimizers include Stochastic Gradient Descent (SGD) and its variants, such as Adagrad, Adadelta, RMSProp, and Adam. The Adam optimizer is a commonly used optimizer that combines the advantages of Adagrad and RMSProp, adaptively adjusting the learning rate for each parameter and converging relatively quickly during training. The optimizer calculates the gradient of the loss function with respect to the model parameters and updates the parameters based on the gradient and the learning rate to minimize the loss function.

[0080] Training Process: Iterative training is performed using the training set data. In each training step, a batch of data is input into the model. The model performs forward propagation based on the current parameters to calculate the predicted value, then calculates the loss between the predicted and the true values, followed by backpropagation to calculate the gradient of the loss with respect to the model parameters. Finally, the optimizer updates the model parameters based on the gradient. This process is repeated until the loss function converges (e.g., the loss value no longer decreases significantly) or the preset number of training epochs is reached. For example, in a training set containing 10,000 samples, with a batch size of 32, each training epoch requires... Secondary parameter update.

[0081] 4. Model Evaluation: Validation Set Evaluation: During training, the model is periodically evaluated using validation set data. By calculating the loss function value and other relevant metrics (such as accuracy, recall, and F1 score in classification tasks, and mean absolute error (MAE) in regression tasks) on the validation set, the model's performance changes are observed. Based on the validation set evaluation results, the model's hyperparameters are adjusted, such as increasing or decreasing the number of hidden units or adjusting the learning rate, to prevent overfitting or underfitting and improve the model's generalization ability. For example, if the loss on the validation set starts to increase after a certain training round while the loss on the training set continues to decrease, it may indicate that the model is starting to overfit. In this case, reducing model complexity or increasing regularization terms can be attempted.

[0082] Test Set Evaluation: After model training and hyperparameter tuning are complete, the model is evaluated using test set data. Test set data consists of data the model has never seen during training; therefore, the evaluation results on the test set accurately reflect the model's performance in real-world applications. Report the model's metrics on the test set, such as accuracy, F1 score, and MSE, to assess whether the model meets the task requirements. If the model's performance on the test set is unsatisfactory, it may be necessary to re-examine the data processing procedures, model architecture, or training methods for further improvements.

[0083] like Figure 3 As shown, the method for generating the demand identification model is as follows: Collect EEG information from several elderly people or young children, as well as demand information in corresponding dimensions based on the EEG information; The time-frequency features of the EEG information are extracted by wavelet transform and bound to the corresponding dimension of demand information. Analyze the time-frequency characteristics of each elderly person or child and construct a feature set; The feature set is divided into training set, test set and validation set according to the proportion; The training set is imported into the initial LSTM model for feature training and learning to generate the initial demand recognition model. The recognition performance of the requirement recognition model was tested and verified using a test set and a validation set, respectively. If the test passes, the requirement identification model will be deployed and applied.

[0084] 1. Collect training data Baseline testing can be performed after the Petty EEG module, collecting EEG information from the elderly or young children and simultaneously recording their needs tags (such as thirst, pain, etc.). After a series of preprocessing steps, the data enters the feature processing process.

[0085] Time-frequency feature extraction: A 6-level wavelet packet decomposition is performed using the db20 wavelet basis function to extract the energy ratios of the θ wave (4-8Hz), α wave (8-13Hz), and β wave (13-30Hz), forming a 3D feature vector [Eθ, Eα, Eβ]. Samples are generated by binding demand labels, in the format [time-frequency feature, demand category (one-hot encoding)].

[0086] 2. Model Building and Training Dataset partitioning: The dataset is divided into a training set (model learning), a validation set (hyperparameter tuning), and a test set (final evaluation) in a 7:2:1 ratio.

[0087] LSTM model architecture Input layer: Receives 3D time-frequency feature sequences with a time step of 10 (sliding window). Hidden layers: 2-layer bidirectional LSTM (64 units per layer), Dropout=0.3 to prevent overfitting; Output layer: The Softmax activation function outputs the required class probability.

[0088] Training parameter configuration: Loss function: Classification cross-entropy; Optimizer: Adam (lr=0.001); Batch size: 64; Early stop mechanism: Terminate training if the verification loss does not decrease for 5 consecutive rounds.

[0089] 3. Verification and Deployment Performance evaluation Test set metrics: A score of >85% accuracy and >0.8 is considered acceptable.

[0090] Confusion matrix analysis is used to identify the error rate of each category, and feature engineering is optimized accordingly.

[0091] Model Deployment The data is converted to ONNX format and embedded into the nursing system to receive EEG data uploaded from the wristband in real time and output strategies; the model is fine-tuned regularly with new data to maintain predictive accuracy.

[0092] For the LSTM architecture and training process described above, please refer to the general LSTM training steps outlined earlier. Training parameter configurations, such as the number of iterations and optimization functions, are selected by the user from existing functions in the library or can be customized. Model performance validation is the responsibility of the user; if validation fails, the training data should be changed and retraining should be performed.

[0093] Preferably, the learning machine further includes: 1. An interactive task generation module, used to generate corresponding interactive tasks for the elderly or children based on the aforementioned requirements information; The interactive task generation module uses input needs information as its basis. This information may encompass various factors such as the age, interests, cognitive level, and current context of the elderly or young children. By parsing this information and applying specific algorithms and preset rules, the module generates matching interactive tasks. For example, if the needs information indicates the target is a 3-5 year old child who loves animals, the module might generate interactive tasks such as imitating animal sounds or telling animal stories. If the target is an elderly person with limited mobility who enjoys chess, it might generate interactive tasks involving online chess games. By tailoring interactive tasks to the individual needs and characteristics of the elderly and young children, the module enhances their enthusiasm and initiative in participating in interactive activities. Through diverse and practical task generation, the module enriches their daily life experiences, promoting the preservation of cognitive abilities in the elderly and the cognitive development in young children.

[0094] 2. A digital twin system for providing digital human services based on digital twin technology and generating a set of digital human interaction instructions adapted to the interactive tasks; Digital twin systems utilize digital twin technology to construct virtual digital models (a known technology) that correspond to real-world scenarios or individuals. The system first analyzes the digital human's behavior, language, and other characteristics required for the interactive task, based on the interactive task generation module. Then, based on its self-constructed digital human model, it generates a set of digital human interaction instructions adapted to the task. For example, if the task is to teach young children to recognize fruits, the digital twin system will generate an instruction set including information such as the fruit's name, color, and taste, as well as corresponding action instructions, such as picking up a fruit model to demonstrate. Through digital twin technology, it provides highly realistic and accurately adapted digital human services for interactive tasks. The generated digital human interaction instruction set allows the digital human to participate in interactions in a natural and reasonable way, enhancing the fun and educational value of the interaction and increasing the immersion for the elderly and young children during the interaction process.

[0095] 3. A digital human projection module, used to project a virtual image of a digital human, execute the digital human interaction instruction set and drive the digital human to perform corresponding interactive actions, so as to interact with the elderly or young children; The digital human projection module receives the digital human interaction instruction set generated by the digital twin system. Using projection technology, it projects the virtual image of the digital human into a specific space. Based on the action instructions in the instruction set, it drives the digital human to perform corresponding interactive actions. For example, if the instruction set requires the digital human to smile or wave, the projection module will accurately present these actions by controlling projection parameters and related equipment, enabling visual interaction between the digital human and the elderly or young children. This presents the digital human to the elderly or young children in an intuitive way, allowing them to directly see the digital human's interactive actions, enhancing the realism and visualization of the interaction. By accurately executing the digital human interaction instruction set, it ensures that the digital human's interactive actions are consistent with the task requirements, improving the quality of the interactive experience.

[0096] 4. A voice device for providing voice playback services for the digital human projection module during digital human operation; The voice device works in conjunction with the digital human projection module. When the digital human projection module performs interactive actions, the voice device receives voice data sent from the digital twin system or related control unit. This voice data corresponds to the digital human's interactive command set, and the voice device plays the voice content according to the instructions. For example, when the digital human introduces fruit, the voice device simultaneously plays the corresponding fruit introduction audio. This adds a sound dimension to the digital human, making the interaction more vivid and realistic. The voice playback service, in conjunction with the digital human's actions, creates a comprehensive interactive scenario, allowing the elderly and young children to participate in the interaction through multiple senses, including hearing and sight, enhancing the attractiveness and appeal of the interaction.

[0097] 5. An expression control module, used to provide expression control services for the digital human projection module during digital human operation; The facial expression control module also works closely with the digital human projection module. Based on the interactive instruction set generated by the digital twin system and the real-time context of the interaction, it precisely controls the facial expressions projected by the digital human. For example, when playing games with young children, the module will display happy or approving expressions when the child performs well; during storytelling, it will adjust the digital human's expressions according to the story's atmosphere, such as sadness or surprise. This imbues the digital human with rich emotional expression, making it more human-like. Through precise facial expression control, it enhances the emotional resonance between the digital human and the elderly or young children, further improving the quality and experience of the interaction, making the process more human and engaging.

[0098] 6. A sensing device, used to sense and feed back the interactive operation data of the elderly or young children to the digital twin system during the interaction between the elderly or young children and the digital human, so that the digital twin system can drive the digital human to dynamically adjust the corresponding interactive actions based on the interactive operation data.

[0099] During interactions between the elderly or young children and the digital human, the interactive sensing device continuously monitors their interactive data in real time. This data includes various forms such as body movements, voice responses, and touch operations. The interactive sensing device transmits the collected data to the digital twin system. The digital twin system analyzes and processes this data, and drives the digital human to dynamically adjust its interactive actions based on the analysis results. For example, if a young child touches a virtual object in the digital human's hand, the interactive sensing device (such as a sensing finger sleeve) detects the touch action and location information, transmits it to the digital twin system, and the system then instructs the digital human to perform a corresponding action of handing over the object. This enables two-way interaction between the elderly, young children, and the digital human, changing the traditional one-way interaction mode. By sensing and feeding back interactive data, the digital human can dynamically adjust according to the actual interaction situation, making the interaction more natural and smooth, enhancing the fun and user engagement, and improving the overall intelligence level of the interactive system.

[0100] Preferably, the learning machine further includes: The AI ​​editing module is used to identify the needs of the elderly or young children, and based on AI automatic editing technology, it generates predictive text messages corresponding to these needs, including: Memoir generation module: used to extract key memory points marked by alpha wave event-related potentials (ERPs) in elderly people, and use GPT-4o combined with spatiotemporal features to generate narrative text; Children's cognitive generation module: Visualizes gamma wave activity using a diffusion model and generates corresponding knowledge-related graphics and text.

[0101] Preferably, the learning machine further includes: The emotion intervention module is used to assess the corresponding anxiety index based on the time-frequency characteristics of the EEG information. Anxiety index = (β right frontal lobe - β left frontal lobe) / total β energy If the anxiety index is not within the preset threshold range, a corresponding emotion intervention strategy is generated and sent to the expression control module, which then regulates the digital human's expressions.

[0102] The AI ​​editing module first uses advanced sensor technology and intelligent algorithms to accurately identify various signals emitted by the elderly or young children. These signals may include voice, body movements, and facial expressions, and then analyze the underlying needs. Subsequently, based on powerful AI automatic editing technology, it deeply understands and analyzes the identified needs, and uses natural language processing technology and text generation algorithms to generate inference texts based on the characteristics and background of the information, aiming to present content that matches the needs as accurately as possible. It can quickly and accurately interpret the relatively unique expressive needs of the elderly and young children, effectively bridging communication barriers caused by differences in expressive abilities. The generated inference texts possess high accuracy and logic, providing a reliable basis for related services and support, and greatly improving the intelligence level of care services for the elderly and young children. For example, in elderly care institutions, when an elderly person expresses a vague need to recount their past experiences through voice, the AI ​​editing module quickly identifies the need and generates guiding inference texts to help staff communicate more efficiently with the elderly to obtain detailed memories. In kindergarten settings, if a child expresses a desire for a certain type of story through actions or simple words, the AI ​​editing module generates inference texts to provide teachers with a reference for preparing suitable stories.

[0103] The AI ​​editing module can provide the following two application functions: 1. Memoir generation module Working Principle: This module utilizes professional EEG monitoring equipment to collect alpha wave event-related potentials (ERPs) in the elderly person's brain. These potential signals contain key information about the elderly person's memory activity. A specific algorithm analyzes the ERPs to identify key memory points. Then, leveraging the powerful language processing capabilities of GPT-4o and combining the spatiotemporal characteristics of the memory points, such as the time and place of the memory occurrence, a vivid and coherent narrative text is generated according to narrative logic, completing the creation of the memoir content. It can deeply explore the precious memories stored in the elderly person's brain, transforming scattered memory fragments into a systematic, complete, and emotionally rich memoir. The generated narrative text is of high quality, conforms to human language habits and narrative logic, and restores the elderly person's memories to the greatest extent possible, leaving a valuable spiritual legacy for the elderly person and their family.

[0104] For example, in a family setting, when children are creating memoirs for their elderly parents, they can use a portable EEG monitoring device to collect the elderly person's brain signals. A draft can then be generated using the memoir generation module, and subsequently adjusted and refined by the children or professionals to create a unique family memoir. In community senior activity centers, memoir-making activities can be organized for seniors who wish to participate. This module helps seniors record their life experiences and compile them into a book, enhancing communication and the transmission of memories among seniors in the community.

[0105] 2. Children's Cognitive Generation Module Working Principle: The children's cognitive generation module utilizes advanced EEG monitoring technology to monitor and collect gamma wave activity in children's brains in real time. Then, using a diffusion model, the information contained in the gamma wave activity is visualized and transformed into easily understandable images and text. Based on the laws and characteristics of children's cognitive development, the visualized information is further processed to generate corresponding cognitive graphics and text, such as simple science pictures paired with concise text, to match children's cognitive level and comprehension ability.

[0106] It can present the abstract cognitive activities in children's brains in an intuitive and vivid graphic format, providing parents and teachers with a clear basis for understanding children's cognitive status. The generated cognitive graphics are highly targeted, aligning with children's thinking patterns and interests, helping to stimulate children's learning interest and improve their cognitive abilities.

[0107] For example, in kindergarten teaching, teachers can use this module to understand children's cognitive feedback on a particular teaching content in real time, and adjust teaching methods and content accordingly based on the generated cognitive graphics and text. For instance, in early science education courses, graphics and text generated based on children's gamma wave activity can help teachers understand children's cognitive abilities regarding plants, animals, and natural phenomena, and adjust their explanations accordingly. At home, parents can use this module regularly to understand their children's cognitive development and select appropriate children's books or learning materials to promote their children's all-round development.

[0108] Preferably, it further includes: an emotion intervention module, used to assess the corresponding anxiety index based on the time-frequency characteristics in the EEG information: Anxiety index = (β right frontal lobe - β left frontal lobe) / total β energy If the anxiety index is not within the preset threshold range, a corresponding emotion intervention strategy is generated and sent to the expression control module, which then regulates the digital human's expressions.

[0109] The emotion intervention module first acquires EEG information, which contains rich data on human brain activity. Based on this, time-frequency feature analysis is performed on the EEG information to extract key features related to anxiety, namely the beta wave intensity of the right frontal lobe (β-right frontal lobe), the beta wave intensity of the left frontal lobe (β-left frontal lobe), and the total beta energy (which can be identified through feature engineering).

[0110] Anxiety Index Calculation: Anxiety levels are quantified using a specific formula: "Anxiety Index = (Right Frontal Lobe - Left Frontal Lobe) / Total Beta Energy". This formula is based on neurological principles of the brain, which state that differences in beta wave activity across different brain regions are related to anxiety. Through this mathematical calculation, complex EEG characteristics are transformed into a specific numerical value to characterize the anxiety state.

[0111] Threshold Judgment and Intervention Strategy Generation: Administrators can pre-set a threshold range based on user status. This threshold range, determined through extensive experimental data and clinical experience, defines the normal range of anxiety levels. When the calculated anxiety index falls outside this pre-set threshold range, it indicates an abnormal anxiety state. In this case, the emotion intervention module generates corresponding emotion intervention strategies based on a set of pre-defined rules and algorithms. These strategies aim to adjust the individual's emotional state, bringing their anxiety back to a normal range.

[0112] Strategy Transmission and Facial Expression Control: The generated emotion intervention strategy is sent to the facial expression control module. Upon receiving the strategy, the module adjusts the digital human's facial expressions according to the strategy's requirements. As a form of visual feedback, the digital human's facial expressions can intuitively demonstrate the intervention direction taken to alleviate anxiety. For example, by displaying relaxed and cheerful expressions, it can visually provide a certain psychological suggestion to the individual, helping to alleviate anxiety.

[0113] The effects of the emotion intervention module are as follows: Based on the time-frequency characteristics of EEG information and specific calculation formulas, an individual's anxiety index can be quantified relatively accurately. This physiological signal-based assessment method is more objective and accurate than traditional subjective self-assessment methods, such as questionnaires. It can capture changes in anxiety at the subconscious level, unaffected by the concealment or bias of individual subjective consciousness, providing a reliable basis for subsequent intervention measures.

[0114] When an abnormal anxiety level is detected, the module can generate targeted emotion intervention strategies. These strategies are not generalized but generated based on the specific deviation of the anxiety level and pre-set rules. For example, if the anxiety level is too high, strategies to guide relaxation and distraction may be generated; if the anxiety level is too low, which may indicate excessive depression, strategies to boost energy may be generated. This targeting makes the intervention measures more closely aligned with the individual's actual emotional state, improving the intervention's effectiveness.

[0115] This module can acquire EEG information in real time and calculate the anxiety index. As an individual's emotional state dynamically changes, it can promptly detect fluctuations in the anxiety index. Once the index exceeds a threshold range, a new intervention strategy is immediately generated and transmitted to the facial expression control module, enabling real-time adjustments to the digital human's facial expressions. This real-time dynamic adjustment mechanism ensures that emotional intervention closely follows changes in the individual's emotions, maintaining an effective intervention state and promptly alleviating abnormal anxiety.

[0116] By controlling the facial expressions of digital humans, emotion intervention strategies can be presented in an intuitive and visual way. This visual feedback allows individuals to more vividly understand their own emotional state and the corresponding intervention direction. For example, seeing a digital human display a smiling or relaxed expression may subconsciously provide an individual with a positive psychological suggestion, prompting them to proactively adjust their own emotions, cooperate with the intervention strategy, and further enhance the effectiveness of the emotion intervention.

[0117] On the other hand, a learning method based on brain-computer interface and AI mode is provided. This method is used to implement the aforementioned learning system based on brain-computer interface and AI mode. The method includes: Wearing an EEG module, baseline tests are conducted to establish a personal EEG feature database in different dimensions; After the test is completed, the brain-computer interface is used to collect the electroencephalogram (EEG) information of the elderly or young children and upload it to the learning machine. The learning machine performs EEG feature recognition and analysis on the EEG information, outputs corresponding dimension demand information, and recommends interactive tasks with corresponding demand information; And / or, Based on the digital human's ability to perform interactive tasks that meet specific needs, it can interact with the elderly or young children. And / or, It uses AI-powered automatic editing technology to edit and generate speculative copy containing relevant demand information.

[0118] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention, such as... Figure 4 As shown. Optionally, the electronic device 410 may include a first processor 2001.

[0119] Optionally, the electronic device 410 may also include a memory 2002 and a transceiver 2003.

[0120] The first processor 2001, memory 2002, and transceiver 2003 can be connected via a communication bus.

[0121] The following combination Figure 4 A detailed description of each component of electronic device 410 is provided below: The first processor 2001 is the control center of the electronic device 410. It can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs).

[0122] Optionally, the first processor 2001 can perform various functions of the electronic device 410 by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0123] In a specific implementation, as one example, the first processor 2001 may include one or more CPUs, for example... Figure 4 CPU0 and CPU1 are shown in the diagram.

[0124] In a specific implementation, as one example, the electronic device 410 may also include multiple processors, for example... Figure 4 The first processor 2001 and the second processor 2004 are shown in the diagram. Each of these processors can be a single-core processor or a multi-core processor. Here, a processor can refer to one or more devices, circuits, and / or processing cores used to process data (such as computer program instructions).

[0125] The memory 2002 is used to store the software program that executes the present invention, and is controlled by the first processor 2001 to execute it. The specific implementation method can be referred to the above method embodiment, and will not be repeated here.

[0126] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently and be connected via the interface circuit of the electronic device 410. Figure 4 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0127] The transceiver 2003 is used to communicate with network devices or with terminal devices.

[0128] Alternatively, transceiver 2003 may include a receiver and a transmitter. Figure 4 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function.

[0129] Optionally, the transceiver 2003 can be integrated with the first processor 2001, or it can exist independently and be connected via the interface circuit of the electronic device 410. Figure 4 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0130] It should be noted that, Figure 4 The structure of the electronic device 410 shown does not constitute a limitation on the router. Actual knowledge structure identification devices may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0131] Furthermore, the technical effects of the electronic device 410 can be referred to the technical effects of the learning system based on brain-computer interface and AI mode described in the above method embodiments, and will not be repeated here.

[0132] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0133] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0134] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable methods. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0135] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0136] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0137] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0138] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0139] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, methods, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0140] In the embodiments provided by this invention, it should be understood that the disclosed devices, methods, and approaches can be implemented in other ways. For example, the method embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, methods, or units, and may be electrical, mechanical, or other forms.

[0141] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0142] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0143] If the aforementioned functions are implemented as 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 this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0144] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A learning system based on brain-computer interface and AI mode, characterized in that, The system includes an EEG module and a learning machine, wherein: The EEG module is used to collect EEG information from the elderly or young children based on the brain-computer interface and upload it to the learning machine. The learning machine is used to perform EEG feature recognition and analysis on the EEG information, output corresponding dimension demand information, and recommend interactive tasks with corresponding demand information; and / or, run interactive tasks with corresponding demand information based on the digital human to interact with the elderly or children; and / or, edit and generate speculative text with corresponding demand information based on AI automatic editing technology. The EEG module and the learning machine are communicatively connected.

2. The learning system based on brain-computer interface and AI mode according to claim 1, characterized in that, The EEG module uses an ErgoAI high-precision non-invasive EEG headband (8-channel EEG + near-infrared imaging), with a sampling rate of 256Hz, and supports wireless data transmission.

3. The learning system based on brain-computer interface and AI mode according to claim 1, characterized in that, The learning machine includes: A data preprocessing system is used to receive and preprocess the EEG information, including: (1) Signal cleaning process: 0.5-45Hz bandpass filter (FIR design) to remove electromyographic interference; Independent component analysis (ICA) separates eye movement artifacts; Automatic outlier removal based on ASR (Artifact Subspace Reconstruction); (2) Timing alignment: Dynamic time warping (DTW) is used to compensate for individual wear differences.

4. The learning system based on brain-computer interface and AI mode according to claim 3, characterized in that, The learning machine also includes: The EEG analysis module is used to extract time-frequency features from the EEG information through wavelet transform, including: energy ratios of the θ (4-8Hz), α (8-13Hz), or β (13-30Hz) bands; and to identify the time-frequency features and output corresponding dimension demand information through a pre-trained demand recognition model. The method for generating the demand identification model is as follows: Collect EEG information from several elderly people or young children, as well as demand information in corresponding dimensions based on the EEG information; The time-frequency features of the EEG information are extracted by wavelet transform and bound to the corresponding dimension of demand information. Analyze the time-frequency characteristics of each elderly person or child and construct a feature set; The feature set is divided into training set, test set and validation set according to the proportion; The training set is imported into the initial LSTM model for feature training and learning to generate the initial demand recognition model. The recognition performance of the requirement recognition model was tested and verified using a test set and a validation set, respectively. If the test passes, the requirement identification model will be deployed and applied.

5. The learning system based on brain-computer interface and AI mode according to claim 4, characterized in that, The learning machine also includes: The interactive task generation module is used to generate corresponding interactive tasks for the elderly or children based on the aforementioned requirements information. A digital twin system is used to provide digital human services based on digital twin technology and to generate a set of digital human interaction instructions adapted to the interactive tasks. The digital human projection module is used to project virtual images of digital humans, execute the digital human interaction instruction set, and drive the digital human to perform corresponding interactive actions to interact with the elderly or young children. A voice device is used to provide voice playback services for the digital human projection module during digital human operation; The facial expression control module is used to provide facial expression control services for the digital human projection module during digital human operation; The interactive sensing device is used to sense and feed back the interactive operation data of the elderly or young children to the digital twin system during the interaction between the elderly or young children and the digital human. The digital twin system then drives the digital human to dynamically adjust the corresponding interactive actions based on the interactive operation data.

6. The learning system based on brain-computer interface and AI mode according to claim 5, characterized in that, The learning machine also includes: The AI ​​editing module is used to identify the needs of the elderly or young children, and based on AI automatic editing technology, it generates predictive text messages corresponding to these needs, including: Memoir generation module: used to extract key memory points marked by alpha wave event-related potentials (ERPs) in elderly people, and use GPT-4o combined with spatiotemporal features to generate narrative text; Children's cognitive generation module: Visualizes gamma wave activity using a diffusion model and generates corresponding knowledge-related graphics and text.

7. The learning system based on brain-computer interface and AI mode according to claim 6, characterized in that, The learning machine also includes: The emotion intervention module is used to assess the corresponding anxiety index based on the time-frequency characteristics of the EEG information. Anxiety index = (β right frontal lobe - β left frontal lobe) / total β energy If the anxiety index is not within the preset threshold range, a corresponding emotion intervention strategy is generated and sent to the expression control module, which then regulates the digital human's expressions.

8. A learning method based on brain-computer interface and AI mode, wherein the learning method based on brain-computer interface and AI mode is used to implement the learning system based on brain-computer interface and AI mode as described in any one of claims 1-7, characterized in that, The method includes: Wearing an EEG module, baseline tests are conducted to establish a personal EEG feature database in different dimensions; After the test is completed, the brain-computer interface is used to collect the electroencephalogram (EEG) information of the elderly or young children and upload it to the learning machine. The learning machine performs EEG feature recognition and analysis on the EEG information, outputs corresponding dimension demand information, and recommends interactive tasks with corresponding demand information; And / or, Based on the digital human's ability to perform interactive tasks that meet specific needs, it can interact with the elderly or young children. And / or, It uses AI-powered automatic editing technology to edit and generate speculative copy containing relevant demand information.

9. An electronic device, characterized in that, The electronic device includes: processor; A memory storing computer-readable instructions that, when executed by the processor, implement the method as described in claim 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that can be invoked by a processor to execute the method as described in claim 8.

Images