A closed-loop system for attention regulation based on EEG signals

Abstract

The application discloses a closed-loop system for attention regulation based on electroencephalogram signals, which is composed of a hardware part and a software part. The hardware part includes a head-mounted device equipped with high-sensitivity EEG electrodes for collecting electroencephalogram signals of the frontal lobe, and the collected signals are transmitted in real time to a computer or a mobile device through wireless communication technology for further processing and analysis. A multi-lead programmable electrical stimulation device dynamically adjusts the parameters of electrical stimulation according to algorithm feedback and provides a personalized stimulation plan. The personalized attention improvement plan: the application can accurately capture the attention changes of individuals in different states by analyzing the electroencephalogram (EEG) signals of individuals in real time, especially the ratio of theta wave to beta wave (TBR). Using these data, the system can customize a personalized stimulation plan to adapt to the current needs of individuals, rather than using a universal method suitable for all people.

Description

Technical Field

[0001] The present invention belongs to the field of brain-computer interface and artificial intelligence technology, and specifically relates to an attention control closed-loop system based on electroencephalogram (EEG) signals. Background Art

[0002] Attention regulation is an important research area in cognitive science and neuroengineering, particularly in the treatment of attention deficits and cognitive disorders. Traditional methods for diagnosing and regulating attention, such as behavioral scales and clinical assessments, suffer from high misdiagnosis rates, strong subjectivity, and a lack of quantitative indicators. Brain-computer interface (BCI) technology, by collecting and analyzing electroencephalogram (EEG) signals, provides a new approach for objectively assessing and regulating attentional states. Compared to other methods, BCI devices are more portable, less expensive, and easier to operate, offering greater potential for application in clinical settings.

[0003] Although existing EEG attention regulation technology can monitor and regulate attention status to a certain extent, it still faces many challenges in practical application:

[0004] Real-time and accuracy of signal processing: Existing methods still lack high real-time and accuracy when processing EEG signals. Complex EEG signals are easily affected by environmental noise and motion artifacts, resulting in unstable processing results.

[0005] Limited personalized adjustment capabilities: Current EEG control systems are inadequate in addressing individual differences. Different users have different EEG signal characteristics, and fixed control schemes are difficult to meet the needs of each user, affecting the control effect.

[0006] There's significant room for optimization in feedback mechanisms: Existing feedback and stimulation methods are often fixed, making it difficult to dynamically adjust based on the user's real-time state. Stimulation methods with fixed parameters can lead to poor stimulation effects or adverse reactions. Summary of the Invention

[0007] The purpose of the present invention is to provide an attention control closed-loop system based on EEG signals to address the limitations of the traditional attention control methods proposed in the above background technology, such as low diagnostic accuracy, insufficient personalized adjustment, and poor real-time feedback effect.

[0008] To achieve the above purpose, the present invention provides the following technical solutions: a closed-loop system for attention control based on EEG signals, which consists of hardware and software parts, wherein

[0009] The hardware part includes:

[0010] The headset is equipped with highly sensitive EEG electrodes to collect EEG signals from the frontal lobe. The collected signals are transmitted in real time to a computer or mobile device via wireless communication technology for further processing and analysis.

[0011] Multi-channel programmable electrical stimulation device, which dynamically adjusts the parameters of electrical stimulation based on algorithm feedback to provide personalized stimulation plans;

[0012] The software part includes:

[0013] The signal preprocessing module filters and denoises the collected EEG signals to remove environmental noise and artifacts to ensure signal purity;

[0014] Feature extraction module, which extracts attention-related features from the preprocessed EEG signal;

[0015] The attention state classification module uses a residual domain generalization model to classify the extracted features and determine the user's attention state;

[0016] The feedback control module adjusts the parameters of the electrical stimulation device in real time according to the classification results to ensure the effectiveness and targeting of the stimulation.

[0017] As a preferred technical solution in the present invention, the residual domain generalization model includes:

[0018] Residual network: By stacking multiple residual blocks to extract deep features of the signal, the residual network can effectively capture the complex features in EEG signals;

[0019] Domain generalization model: By learning shared features across different domains to improve classification robustness, the domain generalization model can adapt to the EEG signal characteristics of different users and provide personalized attention assessment;

[0020] Algorithm design: Accurately evaluate and regulate the user's attention state through EEG signals; through real-time attention state evaluation and feedback control.

[0021] As a preferred technical solution in the present invention, the head-mounted device uses a residual domain generalization model to extract multi-scale features related to attention from the preprocessed EEG signal, and performs accurate classification to judge the user's attention state, thereby realizing multi-level feature extraction and classification; the multi-channel programmable electrical stimulation device adjusts the parameters of the electrical stimulation device in real time according to the classification results, including waveform, frequency, current size and stimulation duration, to provide a personalized electrical stimulation plan, ensure the effectiveness and pertinence of the stimulation, and realize real-time feedback and dynamic regulation.

[0022] As a preferred technical solution in the present invention, the head-mounted device is also provided with a wireless communication module connected to the multi-channel programmable electrical stimulation device and the host end, and the wireless communication module integrates the functions of data collection and data transmission.

[0023] As a preferred technical solution of the present invention, the specific steps of using the closed-loop system are as follows:

[0024] Signal acquisition:

[0025] By collecting 1000Hz EEG signals, the real-time EEG signals are sent to the host device via WiFi communication to obtain the original EEG data; each real-time judgment is based on 2 seconds.

[0026] Data preprocessing:

[0027] The data preprocessing process requires the use of bandpass filters and bandstop filters. The bandstop filter will filter out the 50Hz power frequency interference signal. At the same time, because the attention EEG signal is related to theta waves and beta waves, the bandpass filter is used to filter and retain the theta wave and alpha signal.

[0028] Feature extraction:

[0029] After data preprocessing, the Fourier transform algorithm will be used to calculate the frequency domain features as follows, and then converted into power spectrum features; the power spectrum features of EEG will be used to characterize the algorithm of EEG attention for training and detection;

[0030]

[0031] Where N is the number of sample points, x[n] is the discrete time signal, X[k] is the discrete frequency domain signal, (where The power spectrum P(f) represents the power of the frequency domain signal as it changes with frequency. The power spectrum can be obtained by calculating the square of the amplitude spectrum. The following formula is used: x(f) represents the amplitude of the corresponding frequency, and its square value is the power spectrum, P(f).

[0032] p(f)=|x(f)| 2

[0033] As above, the frequency domain signal of the EEG is processed to convert it into a power spectrum signal, which is used as the EEG state characteristic signal for algorithm processing;

[0034] Attention state assessment:

[0035] State evaluation is divided into two parts: the first is the model pre-training process; the second is the data evaluation process; the model pre-training process requires the collection of attention EEG data, and an attention state evaluation model is obtained through the algorithm model to classify the attention state and the non-attention state; the residual domain generalization model, the model formula is as follows:

[0036] y=f(x)+x

[0037] Feature y is the sum of the input feature x after passing the feature transformation function f and the original input feature xx;

[0038]

[0039] P(y|x) represents the probability that the output feature y belongs to category k when the input feature x is given; f k (x) is the feature transformation result for category k, and the denominator is for all possible categories j The sum of

[0040] Through different convolution kernels, the input EEG power spectrum density features are convolved to different degrees, using three layers of convolution. The final results are feature fused to obtain the classification features, that is, the EEG feature extraction model.

[0041] As a preferred technical solution in the present invention, when performing classification feature processing, due to the large individual differences, the model used is a domain generalization model to reduce the differences between individuals.

[0042] As a preferred technical solution in the present invention, the software part is an EEG feedback system built with Python as the core framework language; the system interface uses the pyqt5 library to build the software framework, and the interface development utilizes web technology. Specifically, JavaScript and HTML5 are used to build the system's user interface and its interactive functions. In order to achieve communication between Python and JavaScript, the system adopts JSBridge technology; after the data is collected, the EEG signal is transmitted to the computer in real time through the websocket communication protocol. On the computer side, a real-time model is responsible for processing these signals and performing judgment and analysis. The output results of the model are then fed back to the user interface in real time. Users can immediately see their EEG activity status and make corresponding adjustments based on the feedback provided by the system.

[0043] As a preferred technical solution of the present invention, the closed-loop system further includes:

[0044] Attention regulation

[0045] The design of the hardware and software system can dynamically adjust the parameters of electrical stimulation and judge attention in multiple modes. First, the evidence of attention shows that the ratio of theta waves to beta waves can reflect the degree of attention. The higher the degree of attention, the higher the ratio of theta to beta. During the attention regulation stage, EEG data with eyes open and relaxed is first collected, and the ratio of theta to beta is calculated as the attention baseline.

[0046] System design optimization description:

[0047] Baseline data collection: Before the experiment began, the subjects were placed in a relaxed state with their eyes open, and electroencephalogram (EEG) data were collected. The theta to beta wave ratio (TBR) was calculated as the baseline level of attention.

[0048] Basal stimulation setting: select one EEG channel in the prefrontal region as the basal channel and set 0.8 mA direct current stimulation as the initial stimulation mode;

[0049] Task and stimulation synchronization: subjects were asked to perform attention tasks and relaxation tasks alternately; during the attention tasks, basal direct current stimulation was applied synchronously;

[0050] Real-time TBR monitoring: During stimulation, TBR is calculated in real time and compared with the baseline TBR in the relaxed state;

[0051] Adaptive current regulation: If the subject is in an attentive state and the TBR is higher than the relaxed state, the system will automatically adjust the stimulation current according to the change in TBR; if the TBR does not rise, the current will be gradually increased in steps of 0.05 mA until it does not exceed 2 mA; if the TBR exceeds the relaxed state, the current intensity will be gradually reduced;

[0052] Relaxation stimulation: During the relaxation period, AC stimulation frequencies are used to induce alpha waves to promote rest and relaxation;

[0053] Feedback loop: The system continuously monitors TBR and dynamically adjusts stimulation parameters based on its changes, forming a closed-loop feedback mechanism;

[0054] Attention Improvement Assessment

[0055] This closed-loop system builds a more logical attention improvement assessment process, combining neural electrical stimulation with daily attention tasks, and using medical assessment scales for systematic evaluation:

[0056] First, a baseline assessment of the participants' attention level was conducted; then, a series of daily attention tasks were designed, which were required to be completed under controlled external stimuli to test and induce the participants' attention ability; during the execution of the tasks, the attention model monitored and intuitively displayed the participants' attention status in real time by calculating the TBR indicator; changes in this indicator directly guided the adjustment of the personalized electrical stimulation plan to adapt to the real-time needs of the participants; by tracking the participants' performance over a long period of time and regularly repeating the assessment scale, the improvement of attention was systematically evaluated.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] Personalized attention enhancement program: This invention can accurately capture the changes in an individual's attention under different states by analyzing the individual's electroencephalogram (EEG) signals in real time, especially the ratio of theta waves to beta waves (TBR). Using this data, the system can customize a personalized stimulation program to suit the individual's current needs, rather than adopting a one-size-fits-all approach. The implementation of a personalized program makes interventions to enhance attention more precise and effective, because each person's brain response and attention pattern are unique.

[0059] Dynamic Stimulation Adjustment: The closed-loop system design of this invention allows for dynamic adjustment of stimulation parameters, including intensity, frequency, and duration, based on real-time feedback from EEG signals. This dynamic adjustment mechanism ensures that stimulation is always synchronized with the individual's current brain state and attention needs, thereby maximizing the stimulation effect. Compared to traditional fixed-pattern stimulation, dynamic adjustment can more flexibly respond to various changes that may occur during task performance, such as fluctuations in attention or increases in task difficulty.

[0060] Real-time closed-loop monitoring and feedback: A key innovation of the present invention is the introduction of a real-time closed-loop monitoring and feedback mechanism, which is achieved by continuously tracking the ratio of theta waves to beta waves (TBR) in the electroencephalogram (EEG). The closed-loop system is able to respond instantly to monitored EEG changes and adjust the electrical stimulation parameters in real time through an automated algorithm to ensure that the stimulation always matches the individual's current state of attention. This real-time feedback not only improves the accuracy of the stimulation, but also allows the system to quickly adapt to any changes in attention levels, whether due to task difficulty, external interference or individual physiological changes. Participants can obtain immediate feedback through real-time data visualization, which helps to enhance awareness of attention status and make manual interventions or adjustments when necessary. The real-time closed-loop monitoring and feedback mechanism significantly improves the adaptability and flexibility of the system, enabling it to cope with a variety of complex and dynamic real-world conditions, thereby effectively improving attention in a variety of environments and situations.

[0061] By integrating real-time closed-loop detection feedback into the attention monitoring system, the present application not only provides continuous attention state evaluation, but also ensures that the implementation of the electrical stimulation program is always centered on the actual needs of the individual, thereby maximizing the effectiveness and efficiency of the intervention. This real-time, adaptive monitoring and feedback method provides a new standard for personalized cognitive intervention. BRIEF DESCRIPTION OF DRAWINGS

[0062] Figure 1 is a whole flowchart of the present application;

[0063] Figure 2 is a whole use flowchart of the present application;

[0064] Figure 3 is a residual model parameter diagram of the present application;

[0065] Figure 4 is a domain generalization model classification diagram of the present application. DETAILED DESCRIPTION

[0066] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0067] Please refer to Figures 1 to 4 The present application provides a technical solution: an attention regulation closed-loop system based on electroencephalogram signals, which is composed of a hardware part and a software part, wherein

[0068] The hardware part includes:

[0069] The head-mounted device is equipped with high-sensitivity EEG electrodes for collecting electroencephalogram signals of the prefrontal cortex. The collected signals are transmitted in real time to a computer or a mobile device for further processing and analysis through wireless communication technology.

[0070] The multi-lead programmable electric stimulation device dynamically adjusts the parameters of electric stimulation, such as current size, waveform frequency and stimulation duration, according to an algorithm, and provides a personalized stimulation scheme, and the model of the application provides multiple stimulation modes, control of each channel, current size, voltage size, waveform control, frequency control and other two parameters, realizes multi-channel and multi-mode control, and when stimulation starts, the device can select a commonly used basic stimulation mode, such as using a sine wave as a stimulation mode, and obtains a dynamically changing stimulation mode through EEG feedback; the purpose of the hardware design is to realize real-time monitoring and regulation of the user's attention state, and to improve the convenience and user experience of the device; in the circuit design of the head-mounted device for EEG acquisition, the first consideration is the parameter index of EEG, the amplitude of EEG is 5-200 microvolts, and the frequency is below 70Hz, in the design, the related characteristics of EEG are fully considered, the chip design scheme of ads1299 is adopted, and in the hardware design, a filtering circuit according to the characteristics of EEG is adopted, a low-pass filtering circuit below 70Hz is used, and since the EEG is disturbed, the problem of noise of the hardware circuit itself needs to be considered in the design. In the design, the circuit processing mode matched with the chip is adopted, which can effectively avoid the noise problem caused by the design of the circuit board itself; in the design of the multi-lead programmable electric stimulation device for EEG electric stimulation, the mode of EEG stimulation and the size of stimulation need to be fully considered, transcranial direct current stimulation generally uses a current stimulation below 2mA, transcranial alternating current stimulation generally uses sine wave, square wave related characteristics, the device selects a chip model that can output a maximum current mode of 67mA and multiple stimulation models, and the communication delay of the chip is less than 1ms, which can realize real-time and rapid data feedback;

[0071] The software part includes:

[0072] The signal preprocessing module filters and denoises the collected EEG signal, removes environmental noise and artifacts, and ensures the purity of the signal;

[0073] The feature extraction module extracts attention-related features from the preprocessed EEG signal, such as EEG band power;

[0074] The attention state classification module uses a residual domain generalization model to classify the extracted features and determine the user's attention state, and an advanced residual domain generalization model is used for EEG signal processing and attention state classification; the model learns the characteristics of residual network and domain generalization, can generalize between different users, and provides consistent and accurate classification results;

[0075] The feedback control module adjusts the parameters of the electrical stimulation device in real time based on the classification results to ensure the effectiveness and targeting of the stimulation;

[0076] The overall user interface is divided into two types: the acquisition interface and the stimulation interface. Data is collected through electrical sensors, collecting EEG signals to the wearable device. The wearable device then transmits the data to the host through its wireless transmission module. The host analyzes the data and uses the model's predictions to determine the state of attention. Based on this state, the host adjusts parameters such as current and waveform, and sends the corresponding parameters to the wearable module. After receiving the parameters, the module adjusts the hardware's electrical stimulation indicators, stimulating feedback changes to achieve closed-loop attention control.

[0077] In this embodiment, the residual domain generalization model includes:

[0078] Residual network: By stacking multiple residual blocks to extract deep features of the signal, the residual network can effectively capture the complex features in EEG signals;

[0079] Domain generalization model: By learning shared features across different domains to improve classification robustness, the domain generalization model can adapt to the EEG signal characteristics of different users and provide personalized attention assessment;

[0080] Algorithm design: Accurately evaluate and regulate the user's attention state through EEG signals; through real-time attention state evaluation and feedback control, the present invention constructs an efficient closed-loop system, significantly improving the accuracy and personalization level of attention state judgment.

[0081] In this embodiment, the head-mounted device uses a residual domain generalization model to extract multi-scale features related to attention from the preprocessed EEG signal, and performs accurate classification to determine the user's attention state, achieving multi-level feature extraction and classification; the multi-channel programmable electrical stimulation device adjusts the parameters of the electrical stimulation device in real time based on the classification results, including waveform, frequency, current magnitude, and stimulation duration, to provide a personalized electrical stimulation solution, ensure the effectiveness and pertinence of the stimulation, and achieve real-time feedback and dynamic regulation; the present invention, by integrating advanced hardware design and efficient algorithm models, provides a closed-loop system that can monitor and regulate the user's attention state in real time, significantly improving the effect of attention regulation and user experience. Through accurate attention assessment and dynamic feedback regulation, the present invention provides users with an efficient, convenient, and personalized attention regulation solution.

[0082] In this embodiment, the head-mounted device is also provided with a wireless communication module connected to the multi-channel programmable electrical stimulation device and the host end. The wireless communication module integrates the functions of data acquisition and data transmission. In this design, the data sampling rate must be considered to reach a sampling rate of 1000 Hz. Considering the power consumption of this system, it is determined to use the ESP32 Bluetooth low power mode. Through the design of the module, a high sampling rate of EEG can be achieved, and there is no delay in Bluetooth transmission.

[0083] In this embodiment, the specific steps of using the closed-loop system are as follows:

[0084] Signal acquisition:

[0085] By collecting 1000Hz EEG signals, the real-time EEG signals are sent to the host device via WiFi communication to obtain the original EEG data; each real-time judgment is based on 2 seconds.

[0086] Data preprocessing:

[0087] The data preprocessing process requires the use of bandpass filters and bandstop filters. The bandstop filter will filter out the 50Hz power frequency interference signal. At the same time, because the attention EEG signal is related to theta waves and beta waves, the bandpass filter is used to filter and retain the theta wave and alpha signal.

[0088] Feature extraction:

[0089] After data preprocessing, the Fourier transform algorithm will be used to calculate the frequency domain features as follows, and then converted into power spectrum features; the power spectrum features of EEG will be used to characterize the algorithm of EEG attention for training and detection;

[0090]

[0091] Where N is the number of sample points, x[n] is the discrete time signal, X[k] is the discrete frequency domain signal, (where The power spectrum P(f) represents the power of the frequency domain signal as it changes with frequency. The power spectrum can be obtained by calculating the square of the amplitude spectrum. The following formula is used: x(f) represents the amplitude of the corresponding frequency, and its square value is the power spectrum, P(f).

[0092] p(f)=|x(f)| 2

[0093] As above, the frequency domain signal of the EEG is processed to convert it into a power spectrum signal, which is used as the EEG state characteristic signal for algorithm processing;

[0094] Attention state assessment:

[0095] State evaluation is divided into two parts: the first is the model pre-training process; the second is the data evaluation process; the model pre-training process requires the collection of attention EEG data, and an attention state evaluation model is obtained through the algorithm model to classify the attention state and the non-attention state; the residual domain generalization model, the model formula is as follows:

[0096] y=f(x)+x

[0097] Feature y is the sum of the input feature x after passing the feature transformation function f and the original input feature xx;

[0098]

[0099] P(y|x) represents the probability that the output feature y belongs to category k when the input feature x is given; f k (x) is the feature transformation result for category k, and the denominator is for all possible categories j The sum of

[0100] Through different convolution kernels, the input EEG power spectrum density features are convolved to different degrees, using 3 layers of convolution, and the final result is feature fused to obtain the classification features. Figure 3 , that is, the EEG feature extraction model.

[0101] In this embodiment, when performing classification feature processing, due to the large individual differences, the model used is a domain generalization model to reduce the differences between individuals. Classification A and classification B are EEG classifications between different individuals. By reducing the loss of model training, individuals cannot be classified, thereby reducing individual differences and the model focuses more on attention features.

[0102] In this embodiment, the software part of the EEG feedback system is built with Python as the core framework language; the system interface uses the pyqt5 library to build the software framework, and the interface development utilizes web technology. Specifically, JavaScript and HTML5 are used to build the system's user interface and its interactive functions. In order to achieve communication between Python and JavaScript, the system adopts JSBridge technology; after the data is collected, the EEG signal is transmitted to the computer in real time through the websocket communication protocol. On the computer side, a real-time model is responsible for processing these signals and performing judgment and analysis. The output results of the model are then fed back to the user interface in real time. Users can immediately see their EEG activity status and make corresponding adjustments based on the feedback provided by the system.

[0103] In this embodiment, the closed-loop system further includes:

[0104] Attention regulation

[0105] The design of the hardware and software system can dynamically adjust the parameters of electrical stimulation and judge attention in multiple modes. First, the evidence of attention shows that the ratio of theta waves to beta waves can reflect the degree of attention. The higher the degree of attention, the higher the ratio of theta to beta. During the attention regulation stage, EEG data with eyes open and relaxed is first collected, and the ratio of theta to beta is calculated as the attention baseline.

[0106] System design optimization description:

[0107] Baseline data collection: Before the experiment began, the subjects were placed in a relaxed state with their eyes open, and electroencephalogram (EEG) data were collected. The theta to beta wave ratio (TBR) was calculated as the baseline level of attention.

[0108] Basal stimulation setting: select one EEG channel in the prefrontal region as the basal channel and set 0.8 mA direct current stimulation as the initial stimulation mode;

[0109] Task and stimulation synchronization: subjects were asked to perform attention tasks and relaxation tasks alternately; during the attention tasks, basal direct current stimulation was applied synchronously;

[0110] Real-time TBR monitoring: During stimulation, TBR is calculated in real time and compared with the baseline TBR in the relaxed state;

[0111] Adaptive current regulation: If the subject is in an attentive state and the TBR is higher than the relaxed state, the system will automatically adjust the stimulation current according to the change in TBR; if the TBR does not rise, the current will be gradually increased in steps of 0.05 mA until it does not exceed 2 mA; if the TBR exceeds the relaxed state, the current intensity will be gradually reduced;

[0112] Relaxation stimulation: During the relaxation period, AC stimulation frequencies are used to induce alpha waves to promote rest and relaxation;

[0113] Feedback loop: The system continuously monitors TBR and dynamically adjusts stimulation parameters based on its changes, forming a closed-loop feedback mechanism;

[0114] Attention Improvement Assessment

[0115] This closed-loop system builds a more logical attention improvement assessment process, combining neural electrical stimulation with daily attention tasks, and using medical assessment scales for systematic evaluation:

[0116] First, a baseline assessment of participants' attention levels was conducted. Subsequently, a series of daily attention tasks were designed, requiring completion under controlled external stimuli, to test and stimulate participants' attentional abilities. During these tasks, the attention model calculated the TBR metric, providing real-time monitoring and visual display of participants' attentional states. Changes in this metric directly guided adjustments to the personalized electrical stimulation protocol to suit participants' real-time needs.

[0117] Improvements in attention were systematically assessed by tracking participants' performance over time and using assessment scales that were repeated periodically. This approach, combining real-time monitoring, dynamic stimulation adjustments, and long-term assessment, not only validated the potential effects of electrical stimulation on improving attention but also activated participants' proactive attention and concentration abilities through the performance of attention tasks.

[0118] The methodological innovation of this study lies in that it no longer relies on a single fixed stimulation pattern, but provides each participant with a personalized attention enhancement program through real-time monitoring and dynamic adjustment; the implementation of this program is expected to significantly improve the performance of participants in long-term concentration tasks, and provide new perspectives and tools for the intervention of attention deficits and related cognitive disorders.

[0119] Although embodiments of the present invention have been shown and described (see the detailed description above for details), it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, and the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A closed-loop attention control system based on EEG signals, characterized by: The closed-loop system consists of hardware and software. The hardware part includes: The headset is equipped with highly sensitive EEG electrodes to collect EEG signals from the frontal lobe. The collected signals are transmitted in real time to a computer or mobile device via wireless communication technology for further processing and analysis. Multi-channel programmable electrical stimulation device, which dynamically adjusts the parameters of electrical stimulation based on algorithm feedback to provide personalized stimulation plans; The software part includes: The signal preprocessing module filters and denoises the collected EEG signals to remove environmental noise and artifacts to ensure signal purity; Feature extraction module, which extracts attention-related features from the preprocessed EEG signal; The attention state classification module uses a residual domain generalization model to classify the extracted features and determine the user's attention state; The feedback control module adjusts the parameters of the electrical stimulation device in real time based on the classification results to ensure the effectiveness and targeting of the stimulation; The specific steps for using the closed-loop system are as follows: Signal acquisition: By collecting 1000Hz EEG signals, the real-time EEG signals are sent to the host device via WiFi communication to obtain the original EEG data; each real-time judgment is based on 2 seconds. Data preprocessing: The data preprocessing process requires the use of bandpass filters and bandstop filters. The bandstop filter will filter out the 50Hz power frequency interference signal. At the same time, because the attention EEG signal is related to theta waves and beta waves, the bandpass filter is used to filter and retain the theta wave and alpha signal. Feature extraction: After data preprocessing, the Fourier transform algorithm will be used to calculate the frequency domain features as follows, and then converted into power spectrum features; the power spectrum features of EEG will be used to characterize the algorithm of EEG attention for training and detection; Where N is the number of sample points, x[n] is the discrete time signal, and X[k] is the discrete frequency domain signal. is a negative function, Σ is a summation function; the power spectrum P(f) represents the change of the power of the frequency domain signal with frequency. The power spectrum can be obtained by calculating the square of the amplitude spectrum. The following formula x(f) represents the amplitude of the corresponding frequency, and its square value is the power spectrum, namely P(f); p(f)=|x(f)| 2 As above, the frequency domain signal of the EEG is processed to convert it into a power spectrum signal, which is used as the EEG state characteristic signal for algorithm processing; Attention state assessment: State evaluation is divided into two parts: the first is the model pre-training process; the second is the data evaluation process; the model pre-training process requires the collection of attention EEG data, and an attention state evaluation model is obtained through the algorithm model to classify the attention state and the non-attention state; the residual domain generalization model, the model formula is as follows: y=f(x)+x Feature y is the sum of the input feature x after passing through the feature transformation function f and the original input feature xx; P(y|x) represents the probability that the output feature y belongs to category k when the input feature x is given; f k (x) is the feature transformation result for category k, and the denominator is e for all possible categories j fj(x) The sum of Through different convolution kernels, the input EEG power spectrum density features are convolved to different degrees, using three layers of convolution. The final results are feature fused to obtain the classification features, that is, the EEG feature extraction model.

2. The closed-loop attention control system based on EEG signals according to claim 1, characterized in that: The residual domain generalization model includes: Residual network: By stacking multiple residual blocks to extract deep features of the signal, the residual network can effectively capture the complex features in EEG signals; Domain generalization model: By learning shared features across different domains to improve classification robustness, the domain generalization model can adapt to the EEG signal characteristics of different users and provide personalized attention assessment; Algorithm design: Accurately evaluate and regulate the user's attention state through EEG signals; through real-time attention state evaluation and feedback control.

3. The EEG-based attention control closed-loop system according to claim 1, characterized in that: The head-mounted device uses a residual domain generalization model to extract multi-scale features related to attention from preprocessed EEG signals, and performs accurate classification to determine the user's attention state, thereby achieving multi-level feature extraction and classification; the multi-channel programmable electrical stimulation device adjusts the parameters of the electrical stimulation device in real time based on the classification results, including waveform, frequency, current size and stimulation duration, to provide a personalized electrical stimulation plan, ensure the effectiveness and pertinence of the stimulation, and achieve real-time feedback and dynamic regulation.

4. The EEG-based attention control closed-loop system according to claim 1, characterized in that: The head-mounted device is also provided with a wireless communication module connected to the multi-channel programmable electrical stimulation device and the host end, and the wireless communication module integrates the functions of data collection and data transmission.

5. The closed-loop attention control system based on EEG signals according to claim 4, characterized in that: When processing classification features, due to the large individual differences, the model used is the domain generalization model to reduce the differences between individuals.

6. The EEG-based attention control closed-loop system according to claim 1, characterized in that: The software part is an EEG feedback system built with Python as the core framework language; the system interface uses the pyqt5 library to build the software framework, and the interface development utilizes web technology. Specifically, JavaScript and HTML5 are used to build the system's user interface and its interactive functions. In order to achieve communication between Python and JavaScript, the system adopts JSBridge technology; after the data is collected, the EEG signal is transmitted to the computer in real time through the websocket communication protocol. On the computer side, a real-time model is responsible for processing these signals and performing judgment and analysis. The output results of the model are then fed back to the user interface in real time. Users can immediately see their EEG activity status and make corresponding adjustments based on the feedback provided by the system.

7. The EEG-based attention control closed-loop system according to claim 1, characterized in that: In a closed-loop system, it also includes: Attention regulation The design of the hardware and software system can dynamically adjust the parameters of electrical stimulation and judge attention in multiple modes. First, the evidence of attention shows that the ratio of theta waves to beta waves can reflect the degree of attention. The higher the degree of attention, the higher the ratio of theta to beta. During the attention regulation stage, EEG data with eyes open and relaxed is first collected, and the ratio of theta to beta is calculated as the attention baseline. System design optimization description: Baseline data collection: Before the experiment began, the subjects were placed in a relaxed state with their eyes open, and electroencephalogram (EEG) data were collected. The theta to beta wave ratio (TBR) was calculated as the baseline level of attention. Basal stimulation setting: select one EEG channel in the prefrontal region as the basal channel and set 0.8 mA direct current stimulation as the initial stimulation mode; Task and stimulation synchronization: subjects were asked to perform attention tasks and relaxation tasks alternately; during the attention tasks, basal direct current stimulation was applied synchronously; Real-time TBR monitoring: During stimulation, TBR is calculated in real time and compared with the baseline TBR in the relaxed state; Adaptive current regulation: If the subject is in an attentive state and the TBR is higher than the relaxed state, the system will automatically adjust the stimulation current according to the change in TBR; if the TBR does not rise, the current will be gradually increased in steps of 0.05 mA until it does not exceed 2 mA; if the TBR exceeds the relaxed state, the current intensity will be gradually reduced; Relaxation stimulation: During the relaxation period, AC stimulation frequencies are used to induce alpha waves to promote rest and relaxation; Feedback loop: The system continuously monitors TBR and dynamically adjusts stimulation parameters based on its changes, forming a closed-loop feedback mechanism; Attention Improvement Assessment This closed-loop system builds a more logical attention improvement assessment process, combining neural electrical stimulation with daily attention tasks, and using medical assessment scales for systematic evaluation: First, a baseline assessment of the participants' attention level was conducted; then, a series of daily attention tasks were designed, which were required to be completed under controlled external stimuli to test and induce the participants' attention ability; during the execution of the tasks, the attention model monitored and intuitively displayed the participants' attention status in real time by calculating the TBR indicator; changes in this indicator directly guided the adjustment of the personalized electrical stimulation plan to adapt to the real-time needs of the participants; by tracking the participants' performance over a long period of time and regularly repeating the assessment scale, the improvement of attention was systematically evaluated.

Images