Search

CN-122006057-A - Intelligent interaction training system and terminal based on electroencephalogram signals and attention mechanisms

CN122006057ACN 122006057 ACN122006057 ACN 122006057ACN-122006057-A

Abstract

The invention relates to the technical field of data processing, in particular to an intelligent interaction training system and a terminal based on an electroencephalogram signal and an attention mechanism, wherein the system comprises a full-flow technical chain for cooperative interaction of full roles through non-invasive electroencephalogram acquisition, multidimensional data processing, personalized model adaptation, ladder-type training feedback, solves the problems of lack of quantitative monitoring, insufficient individuation and single feedback in the traditional training, realizes scientific quantification, accurate intervention and long-acting migration of the attention training, and simultaneously guarantees non-invasive type no side effect in the whole course; meanwhile, the problem that the traditional attention training scheme is mostly generalized content and cannot provide personalized and sustainable training service according to individual characteristics of training objects is solved.

Inventors

  • GAO JIAHUA

Assignees

  • 东莞市彩田教育咨询服务有限公司

Dates

Publication Date
20260512
Application Date
20260202

Claims (10)

  1. 1. The intelligent interactive training system based on the electroencephalogram signal and the attention mechanism is characterized by comprising a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring multidimensional monitoring data; the data processing module is used for preprocessing the multi-dimensional monitoring data and extracting multi-dimensional characteristics to obtain a core characteristic vector and individual baseline data; The system comprises a model analysis module, an individual recommendation strategy, a personal recommendation strategy and a personal recommendation module, wherein the model analysis module is used for analyzing multidimensional monitoring data, core feature vectors and individual baseline data by adopting a neurophysiologic habit model to obtain individual core features and individual tag features; and the ladder feedback module is used for executing a personalized recommendation strategy according to a preset ladder training feedback rule.
  2. 2. The intelligent interactive training system according to claim 1, wherein the preprocessing and multi-dimensional feature extraction are performed on multi-dimensional monitoring data to obtain core feature vectors and individual baseline data, and the intelligent interactive training system comprises the steps of filtering electroencephalogram physiological data by adopting an IIR notch filter to obtain initial electroencephalogram data, filtering signals of a preset wave band in the initial electroencephalogram data by adopting a 4-order Butterworth band-pass filter to obtain second electroencephalogram data, performing artifact separation on the multi-dimensional monitoring data and the second electroencephalogram data by adopting an ICA algorithm to obtain pure electroencephalogram data, and performing feature extraction on the pure electroencephalogram data to obtain the core feature vectors and the individual baseline data.
  3. 3. The intelligent interactive training system according to claim 2, wherein the performing artifact separation on the multi-dimensional monitoring data and the second electroencephalogram data by using an ICA algorithm to obtain pure electroencephalogram data comprises performing blind source separation on the multi-dimensional monitoring data and the second electroencephalogram data by using the ICA algorithm, outputting independent component data, performing artifact identification on the independent component data based on a preset artifact feature library, outputting effective component data, reconstructing the effective component data into electroencephalogram data based on ICA inverse transformation to obtain initial pure electroencephalogram data, and performing baseline drift correction on the initial pure electroencephalogram data to obtain pure electroencephalogram data.
  4. 4. The intelligent interactive training system according to claim 2 is characterized in that the feature extraction is carried out on pure electroencephalogram data to obtain a core feature vector and individual baseline data, and the intelligent interactive training system comprises the steps of calculating average powers of theta waves, alpha waves, SMR waves and beta waves by adopting fast Fourier transformation, calculating key ratios according to the average powers to obtain frequency domain features, calculating the ratio of covariance of beta wave power of forehead F3/F4 sites to standard deviation product, sample entropy and standard deviation of each band power to obtain time domain features, carrying out superposition average and component extraction on electroencephalogram data corresponding to training actions to obtain event related potential features, integrating the frequency domain features, the time domain features and the event related potential features to obtain a core feature vector, and calculating individual baseline data according to resting state electroencephalogram data.
  5. 5. The intelligent interaction training system of claim 1, wherein the analysis of the multi-dimensional monitoring data, the core feature vector and the individual baseline data by using the neuro-physiological habit model to obtain individual core features and individual tag features comprises the steps of performing feature analysis on the multi-dimensional monitoring data, the core feature vector and the individual baseline data to obtain multi-dimensional input features, processing the electroencephalogram features in the multi-dimensional input features by using a combination mode of an attention mechanism and wavelet packet transformation to obtain enhanced electroencephalogram features, performing dimension reduction on the multi-dimensional input features and the enhanced electroencephalogram features by using a principal component analysis method to obtain individual core features, analyzing the individual core features by using an increment FTRL algorithm to optimize the neuro-physiological habit model parameters, outputting optimal parameter vectors, electroencephalogram baseline parameters and an associated rule base, and performing classification processing on the optimal parameter vectors, the electroencephalogram baseline parameters and the associated rule base by using a multi-dimensional decision tree to obtain the individual tag features.
  6. 6. The intelligent interactive training system of claim 5, wherein the objective function of the neurophysiologic habit model is: , wherein the objective function Representing model parameter vectors A minimization solution is performed to the extent that, In order to train the number of parameters, As a Logistic Loss function, Is the first The 64-dimensional individual core features of the individual training samples, Is the first The true attention state label of the individual samples, For the regularization coefficient of L1, Regularizing the coefficients for L2.
  7. 7. The intelligent interaction training system of claim 1, wherein the intelligent recommendation model for brain data is used for carrying out fusion analysis on individual core features and individual label features uploaded by each terminal and outputting an individualized recommendation strategy, the intelligent interaction training system comprises the steps of obtaining initial clustering center features generated randomly, splicing the initial clustering center features with the individual core features to obtain initial feature vectors, mapping the initial feature vectors into fusion feature vectors by using a multi-layer perceptron, clustering the fusion feature vectors by using a K-means++ algorithm to generate a group portrait, carrying out time sequence rule analysis on the fusion feature vectors by using a time sequence LSTM model to obtain an initial individualized dynamic recommendation strategy, screening strategies with the same rank from training strategies of a plurality of similar users as collaborative filtering strategies according to cosine similarity of the individual core features and the group portrait, and fusing the initial dynamic recommendation strategy and the collaborative filtering strategies to obtain the individualized recommendation strategy.
  8. 8. The intelligent interaction training system according to claim 7, wherein the clustering of the fused feature vectors by the K-means++ algorithm is performed to generate a group portrait, and the intelligent interaction training system comprises the steps of initializing a clustering center, calculating the distance between the feature vectors by cosine similarity, distributing the fused feature vector of each individual to the closest clustering center, calculating the feature mean value of each type of sample as a new clustering center, and stopping iteration when the offset of the clustering center is smaller than a preset offset value or the iteration number reaches a preset number, and performing feature extraction on the clustering result to generate the group portrait.
  9. 9. The intelligent interactive training system according to claim 1, wherein the step training feedback rule is used for executing the personalized recommendation policy, the step training feedback rule is used for executing training modes according to a preset training sequence, feedback data of each mode is obtained, and the training policy suitable for the current mode is matched for the feedback data based on the personalized recommendation policy.
  10. 10. The intelligent interactive training system according to claim 1, wherein the terminal comprises an electroencephalogram acquisition terminal for acquiring electroencephalogram physiological data; the training interaction terminal is used for training and acquiring interaction data, and completing the processes of feature extraction, label generation and personalized recommendation strategy generation according to the integrated neurophysiologic habit model and the brain electrical big data intelligent recommendation model; and the multi-mode feedback terminal generates and executes the intervention action according to the personalized recommendation strategy.

Description

Intelligent interaction training system and terminal based on electroencephalogram signals and attention mechanisms Technical Field The invention relates to the technical field of data processing, in particular to an intelligent interaction training system and terminal based on an electroencephalogram signal and an attention mechanism. Background Teenagers (6-12 years old) are in the rapid attention development stage and show remarkable age stratification characteristics, namely the 6-8 years old children are mainly careless, the attention duration is short (only 10-15 minutes), the children are easily disturbed by external stimulus, the 9-12 years old children are mainly cared, the attention stability and the distribution capacity are gradually improved, and the problems of large individual difference, weak anti-interference capacity and the like still exist. The attention development in this stage directly affects cognitive ability, learning efficiency and behavioral habit development, and is the golden period of intervention training. At present, there are a number of training problems in attention training, especially for teenager children. The existing attention training process lacks scientific quantitative real-time monitoring means, can not timely sense the state of inattention of a trainer, and is difficult to accurately intervene in the training process. The traditional attention training scheme is mostly generalized content, personalized and sustainable training service cannot be provided according to individual characteristics (such as age, occupation, attention base level and the like) of a training object, and training difficulty and frequency cannot be dynamically adjusted to match the capability lifting rhythm of a trainer. Third, the feedback intervention mode of the existing training system is single, and the existing training system lacks hierarchical intervention logic, so that when the trainer has a distraction sign, efficient and humanized attention pull-back guidance cannot be realized. Therefore, the invention provides an intelligent interactive training system and a terminal based on an electroencephalogram signal and an attention mechanism so as to solve the problems. Disclosure of Invention Aiming at the situation, in order to overcome the defects of the prior art, the invention provides an intelligent interaction training system and a terminal based on an electroencephalogram signal and an attention mechanism, so as to solve the problem that the traditional attention training scheme is mostly generalized content and cannot provide personalized and continuous training service according to individual characteristics of a training object. In order to achieve the above purpose, the technical scheme adopted by the invention is as follows: The invention provides an intelligent interaction training system based on an electroencephalogram signal and an attention mechanism, which comprises a data acquisition module, a data processing module, a model analysis module, a ladder feedback module and a ladder feedback module, wherein the data acquisition module is used for acquiring multi-dimensional monitoring data, the data processing module is used for preprocessing the multi-dimensional monitoring data and extracting multi-dimensional characteristics to obtain a core characteristic vector and individual baseline data, the model analysis module is used for analyzing the multi-dimensional monitoring data, the core characteristic vector and the individual baseline data by adopting a neurophysiologic habit model to obtain individual core characteristics and individual tag characteristics, the intelligent electroencephalogram data recommendation model is used for carrying out fusion analysis on the individual core characteristics and the individual tag characteristics uploaded by each terminal to output an individual recommendation strategy, and the ladder feedback module is used for executing the individual recommendation strategy according to a preset ladder training feedback rule. The method comprises the steps of preprocessing multi-dimensional monitoring data and extracting multi-dimensional characteristics to obtain core characteristic vectors and individual baseline data, filtering electroencephalogram physiological data by adopting an IIR notch filter to obtain initial electroencephalogram data, screening signals of preset wave bands in the initial electroencephalogram data by adopting a 4-order Butterworth band-pass filter to obtain second electroencephalogram data, carrying out artifact separation on the multi-dimensional monitoring data and the second electroencephalogram data by adopting an ICA algorithm to obtain pure electroencephalogram data, and carrying out characteristic extraction on the pure electroencephalogram data to obtain the core characteristic vectors and the individual baseline data. The method comprises the steps of performing artifact separation on multi-dimensional