Transcranial electrical stimulation control method and system for regulating mental fatigue

Abstract

The invention discloses a transcranial electrical stimulation control method and system for regulating mental fatigue, and belongs to the technical field of mental state regulation. The system comprises an EEG collecting and processing module and an electrical stimulation module which are in wireless communication to form a closed loop. The EEG collecting and processing module collects a forehead lobe three-lead EEG signal, after preprocessing (baseline drift and high-frequency interference are removed through FIR filtering, and eye movement artifacts are removed through combination of wavelet transform and Kalman filtering), the mental fatigue state is classified through an onboard lightweight convolutional neural network (CNN) algorithm, and a control command is generated; the electrical stimulation module receives the command, generates bipolar pulse width modulation transcranial direct current stimulation (tPWMDCs) current, and stimulates the prefrontal lobe. Local low-power-consumption reasoning of EEG signals is achieved, tPWMDCs avoid skin injury, a closed-loop system dynamically adapts to individual differences, portability, safety and accuracy are achieved, and the system is suitable for driving, office and other scenes.

Description

Technical Field

[0001] This invention relates to the field of mental state regulation technology, and in particular to a transcranial electrical stimulation control method and system for regulating mental fatigue. Background Technology

[0002] Real-time monitoring and control of mental fatigue is of great significance in scenarios such as driving safety, improving mental work efficiency, and clinical rehabilitation. Currently, the industry mainly uses two types of technologies to achieve this function, but both have significant drawbacks:

[0003] 1. Deficiencies in EEG signal-based mental fatigue detection technology

[0004] Existing technologies largely rely on multi-lead EEG devices to collect data and combine it with machine learning or deep learning algorithms to determine fatigue status (such as patents CN119745389A and CN116870321A), but they have the following problems:

[0005] High equipment cost and poor portability: Multi-lead EEG equipment is bulky and expensive, requires professional personnel to operate, and cannot meet the real-time detection needs of daily scenarios (such as outdoors and office);

[0006] High algorithm complexity and high resource consumption: To ensure detection accuracy, existing algorithm models have complex structures, requiring a large amount of computing power, storage space and power consumption. They cannot be deployed locally on small hardware and can only rely on host computers for data transmission and inference, resulting in high detection latency and making it difficult to achieve the core requirements of "portability + real-time".

[0007] 2. Limitations of transcranial electrical stimulation-based techniques for regulating mental fatigue

[0008] Existing transcranial electrical stimulation protocols are mainly divided into transcranial direct current stimulation (tDCs), transcranial alternating current stimulation (tACs), and transcranial random noise stimulation (tRNs) (such as patent CN114712706A), but they have the following key problems:

[0009] tDCs have safety concerns: Although tDCs can relieve fatigue through mechanisms such as membrane potential polarization and synaptic plasticity induction, prolonged stimulation can lead to increased skin resistance and poses a risk of skin burns.

[0010] The system lacks closed-loop feedback: existing solutions are all "open-loop control", which cannot detect changes in the fatigue state of the subjects in real time. They can only stimulate according to preset parameters, resulting in low control accuracy, unstable effects, and difficulty in adapting to individual differences.

[0011] To address the aforementioned shortcomings, this invention designs a lightweight onboard algorithm, a portable EEG acquisition module, and a novel transcranial electrical stimulation scheme, constructing an integrated closed-loop system of "detection-feedback-regulation" to solve the problems of portability, low power consumption, safety, and closed-loop regulation in existing technologies. Summary of the Invention

[0012] The purpose of this invention is to provide a transcranial electrical stimulation control method and system for regulating mental fatigue, which solves the problems of poor portability of existing EEG detection equipment, high algorithm resource consumption, insufficient safety of transcranial electrical stimulation, and lack of closed-loop control, and achieves the goal of "real-time detection, precise control, safety and portability" in regulating mental fatigue.

[0013] To achieve the above objectives, the present invention provides a transcranial electrical stimulation control system for regulating mental fatigue, comprising an EEG acquisition and processing module and an electrical stimulation module, wherein the EEG acquisition and processing module and the electrical stimulation module are wirelessly connected to form a closed-loop control.

[0014] The EEG acquisition and processing module is used to acquire EEG signals from the prefrontal cortex leads. After preprocessing the EEG signals, the module uses an onboard lightweight convolutional neural network (CNN) algorithm to classify mental fatigue states and generates control commands based on the classification results, which are then sent to the electrical stimulation module.

[0015] The electrical stimulation module is used to receive the control command, generate bipolar pulse width modulated transcranial direct current (tPWMDCs) to electrically stimulate the F3 and F4 EEG electrode points, thereby regulating mental fatigue.

[0016] Preferably, the EEG acquisition and processing module includes an analog front-end, a signal processing system, and a power management system;

[0017] The analog front end includes a data acquisition electrode, an electrostatic impedance device, an anti-aliasing circuit, and an analog-to-digital converter (ADC) connected in sequence, used to convert the EEG signal into a digital EEG signal;

[0018] The signal processing system includes a microcontroller, which is connected to the analog-to-digital converter and the wireless communication module, and is used to execute the preprocessing, lightweight CNN algorithm and control command generation.

[0019] The power management system includes an isolation circuit, a charging circuit, a power-on / off circuit, and a power module, which provides a stable voltage for the EEG acquisition and processing module.

[0020] Preferably, the electrostatic impedance device is model ESD9B3.3ST5G, used to achieve electrostatic protection; the analog-to-digital converter is model ADS1299, used to convert analog EEG signals into high-precision digital EEG signals.

[0021] Preferably, the electrical stimulation module includes a programmable constant current source, a signal processing system, and a power management system;

[0022] The signal processing system includes a microcontroller, which is connected to a digital-to-analog converter (DAC) and a wireless communication module, respectively, and is used to receive the control commands and generate DAC control commands and pulse width modulation (PWM) parameters.

[0023] The programmable constant current source is connected to the digital-to-analog converter and is used to generate the tPWMDCs current according to the DAC control command, PWM parameters and stimulation direction.

[0024] The power management system is used to provide a stable voltage for the electrical stimulation module.

[0025] Preferably, the programmable constant current source includes a differential operational amplifier, a general-purpose operational amplifier, a sampling resistor, and a MOS array;

[0026] The analog voltage generated by the digital-to-analog converter is superimposed on the transcranial voltage acquired by the general-purpose operational amplifier and then input to the differential operational amplifier. After being amplified 1:1, the voltage is output to the sampling resistor.

[0027] The sampling resistor has an accuracy of 0.1% and a resistance of 1KΩ;

[0028] The MOS array is used to control the output direction, frequency, and duty cycle of the tPWMDCs current.

[0029] Preferably, the wireless communication module is model HC-42; the microcontroller in the EEG acquisition and processing module is model STM32U575VGT6; and the microcontroller in the electrical stimulation module is model STM32F1038T6.

[0030] The present invention also provides a transcranial electrical stimulation control method for regulating mental fatigue, comprising the following steps:

[0031] S1. Collect EEG signals from the frontal lobe leads of the subject;

[0032] S2. Preprocess the acquired EEG signal to remove interference signals;

[0033] S3. Input the preprocessed EEG signal into the onboard lightweight convolutional neural network (CNN) algorithm to classify mental fatigue state and obtain the classification result.

[0034] S4. Generate an electrical stimulation control command based on the classification result, and wirelessly transmit the control command to the electrical stimulation execution unit;

[0035] S5. The electrical stimulation execution unit generates a bipolar pulse width modulated transcranial direct current (tPWMDCs) according to the control command, and performs electrical stimulation on the frontal lobe of the subject to achieve closed-loop regulation of mental fatigue.

[0036] Preferably, the model structure of the lightweight CNN algorithm includes:

[0037] Four convolutional layers are connected in sequence, with a dropout layer following each convolutional layer;

[0038] The four convolutional layers are followed by an average pooling layer;

[0039] The average pooling layer is followed by the first fully connected layer and the second fully connected layer in sequence.

[0040] The output of the second fully connected layer is connected to the Softmax function, which is used to output the classification probability of mental fatigue state.

[0041] Preferably, the deployment process of the lightweight CNN algorithm includes:

[0042] The trained CNN model is lightweighted using the TensorFlowLite framework to obtain the model parameters;

[0043] The model parameters are compiled in the Arduino development environment and burned into the ROM of the microcontroller;

[0044] The model parameters are loaded by the TensorFlowLite interpreter in the microcontroller, and the preprocessed EEG signal is inferred and classified.

[0045] Preferably, the preprocessing in step S2 includes:

[0046] A finite impulse response (FIR) filter is used to remove baseline drift and high-frequency interference. The FIR filter has an order of 500 and a bandwidth of 0.5-45Hz.

[0047] Wavelet transform combined with Kalman filtering is used to remove eye movement artifacts.

[0048] Therefore, the transcranial electrical stimulation control method and system for regulating mental fatigue using the above-described structure of the present invention have the following beneficial effects:

[0049] (1) Enhanced safety: The new tPWMDCs solution avoids the risk of increased skin impedance and burns caused by prolonged tDCs stimulation by dynamically adjusting PWM parameters (frequency, duty cycle, direction);

[0050] (2) Portability and low power consumption: The three-lead EEG design of the frontal lobe + the onboard lightweight CNN algorithm enable the EEG acquisition and processing module to complete inference locally without the need for a host computer. The device is small in size and has low power consumption, meeting the daily portability needs.

[0051] (3) Closed-loop precise control: Real-time closed loop of "EEG detection - fatigue classification - electrical stimulation feedback" can be achieved. Stimulation parameters can be dynamically adjusted according to the fatigue state of the subject to adapt to individual differences. The control precision is significantly higher than that of the open-loop scheme.

[0052] (4) Algorithm resource balance: Lightweight CNN reduces computing power and space requirements while ensuring classification accuracy. It can be deployed on microcontrollers, solving the pain point of "high resource consumption" of existing algorithms.

[0053] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0054] Figure 1 This is a diagram of the overall system structure.

[0055] Figure 2 This is a system architecture diagram of the EEG acquisition and processing module;

[0056] Figure 3 This is a schematic diagram of an electrostatic discharge (ESD) protection and anti-aliasing circuit.

[0057] Figure 4 This is a schematic diagram of an analog-to-digital converter;

[0058] Figure 5 A schematic diagram of the STM32U575VGT6 microcontroller;

[0059] Figure 6 This is a schematic diagram of the CNN model structure;

[0060] Figure 7 This is a diagram of the electrical stimulation hardware structure.

[0061] Figure 8 Schematic diagram of voltage-controlled constant current source 1;

[0062] Figure 9 This is a schematic diagram of voltage-controlled constant current source 2. Detailed Implementation

[0063] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0064] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0065] Example

[0066] This invention provides a transcranial electrical stimulation control system for regulating mental fatigue, such as... Figure 1 As shown, the mental fatigue relief system based on real-time feedback transcranial electrical stimulation (TCS) consists of two parts: an EEG acquisition and processing module and an electrical stimulation module. The EEG acquisition and processing module acquires EEG signals, preprocesses them, and then classifies mental fatigue using an onboard convolutional neural network algorithm. Based on the classification results, the EEG acquisition and processing module generates control commands for the electrical stimulation module, which are transmitted via Bluetooth. The electrical stimulation module generates corresponding pulse-width modulation (PWM) demodulation stimulation currents based on the received control commands. These stimulation currents are then delivered to the subject via electrodes attached to the prefrontal cortex of the brain.

[0067] like Figure 2 As shown, the EEG acquisition and processing module includes an analog front-end, a signal processing system, and a power management system. The power management system generates the various voltages required for the entire module. First, the Type-C charging port passes through an isolation circuit to isolate interference from discharge through the skin surface during electrical stimulation. Then, the battery is charged via a charging circuit, followed by device power-on / off control via a power-on / off circuit, and finally, the power module generates the necessary voltages for the device. In the analog front-end, the EEG signal sequentially passes through acquisition electrodes, an electrostatic protection unit, an anti-aliasing circuit, and a high-precision analog-to-digital converter (ADC) to be converted into a digital EEG signal. The signal processing system connects to the ADC, wireless communication module, and power management system via a microcontroller. The microcontroller connects to the ADC via an SPI bus to acquire the EEG signal, then performs preprocessing and assesses mental fatigue. Based on the assessment result, the microcontroller generates a control command, which is sent to the electrical stimulation module via a radio frequency transceiver for stimulation feedback.

[0068] The electrostatic discharge protection circuit and anti-aliasing circuit of the EEG acquisition and processing module are as follows: Figure 3 As shown, the electrostatic discharge (ESD) protection circuit uses ESD9B3.3ST5G electrostatic impedance devices. When an abnormal overvoltage occurs and reaches the breakdown voltage, these devices rapidly change from a high-resistance state to a low-resistance state, providing a low-impedance conduction path for the instantaneous current and clamping the abnormal high voltage within a safe level, thus protecting the circuit and components. Once the abnormal overvoltage disappears, the ESD impedance devices return to a high-resistance state, and the circuit operates normally. Since the subsequent analog-to-digital converter is an ESD (electrostatic discharge) sensitive device, this step is essential. The acquired signal passes through the ESD impedance devices and is then transmitted through an anti-aliasing circuit to... Figure 4 The analog-to-digital converter (ADS1299) in the system uses an anti-aliasing circuit consisting of a single-stage low-pass filter made up of resistors and capacitors. This circuit is used before the ADS1299 to prevent high-frequency noise signals from introducing spectral aliasing during the conversion process, which could cause errors in the ADS1299. After conversion by the ADS1299, the EEG signal is transmitted to the microcontroller via an SPI converter, where it undergoes EEG data preprocessing and fatigue condition assessment. Figure 5 As shown, the microcontroller model is STM32U575VGT6. The reason for choosing this microcontroller is that it provides a sufficiently large amount of RAM and ROM, and at the same time, it integrates a Cortex-M33 core with a floating-point unit, which makes the microcontroller process EEG data faster.

[0069] After acquiring the EEG signal from the STM32U575VGT6 microcontroller, preprocessing is performed, including baseline drift removal, high-frequency interference removal, and eye-movement artifact removal. This paper uses a 500th-order finite impulse response (FIR) filter (bandwidth: 0.5-45Hz) for baseline drift and high-frequency noise removal, combined with wavelet transform and Kalman filtering for eye-movement artifact removal. After obtaining the clean EEG signal, we obtain the power spectral density of the delta (0.5-4Hz), theta (4-8Hz), alpha (8-13Hz), beta (13-30Hz), and gamma (31-45Hz) bands using Fast Fourier Transform (FFT), and then obtain the power spectrum of each band using the periodogram method. After obtaining the power spectrum, it is input into a trained CNN model for mental state classification. Based on the classification results, control information is sent to the electrical stimulation module.

[0070] The CNN model is set as follows: Figure 6As shown, local feature extraction and abstraction of the EEG power spectrum are performed through four convolutional layers. First, a small convolutional kernel initially captures basic local patterns in the power spectrum. Then, the kernels in the subsequent three layers become larger, extracting higher-level abstract features layer by layer. To prevent overfitting and increase the model's generalization ability, a dropout layer is added after each convolutional layer. Average pooling layers are used instead of fully connected layers to reduce the model parameter size while retaining key features. Fully connected layer 1 flattens the pooled low-dimensional feature vectors and maps them to a higher-dimensional latent space, enhancing expressive power through a non-linear activation function. Fully connected layer 2 maps the abstract features from fully connected layer 1 to the final binary classification space, calculating the probability of each category using the Softmax function.

[0071] The trained CNN model is deployed to the microcontroller as follows: First, the optimal trained model is used to obtain model parameters through a converter provided by the TensorFlow Lite framework. During the parameter acquisition process, the trained model is lightweighted by setting the converter parameters to reduce the size of the model parameters. Second, the obtained model parameters are compiled in Arduino and burned into the ROM of the STM32U575VGT6. The model is then interpreted in the STM32U575VGT6 using the TensorFlow Lite interpreter, and then inference and classification are performed on the acquired EEG signals.

[0072] as follows Figure 7 As shown, the electrical stimulation module includes a programmable constant current source, a signal processing system, and a power management system. The power management system generates the various voltages required by the entire module. First, the Type-C port charges the battery via a charging circuit. Then, the power-on / off circuit controls the device's operation. Next, the power module generates the necessary voltages. The signal processing system connects to a digital-to-analog converter (DAC), an RF transceiver, and the power management system via a microcontroller. After receiving control commands from the EEG acquisition and processing module via the RF transceiver, the microcontroller generates corresponding DAC control commands, causing the DAC to generate a corresponding analog voltage value. Simultaneously, the microcontroller generates the PWM frequency and duty cycle based on the control commands from the EEG acquisition and processing module. The programmable constant current source generates a corresponding stimulation current based on the generated analog voltage value, PWM frequency, duty cycle, and stimulation direction. This current is introduced into the frontal lobe of the brain through electrode pads.

[0073] The voltage-controlled constant current source in the programmable constant current source module is as follows: Figure 8 , 9As shown, this voltage-controlled constant current source consists of a differential operational amplifier and a general-purpose operational amplifier. The voltage value generated by the digital-to-analog converter and the transcranial voltage value acquired by the general-purpose operational amplifier are superimposed and input to the differential operational amplifier. After being amplified 1:1 by the differential operational amplifier, the output is sent to a sampling resistor with an accuracy of 0.1% and a resistance of 1KΩ. That is, the current flowing through the transcranial cavity is the voltage value generated by the digital-to-analog converter divided by the sampling resistor, which is the current value of the transcranial current. After the transcranial constant current value is determined, the output direction, frequency, and duty cycle of the transcranial current are controlled by a MOS array.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A transcranial electrical stimulation control system for regulating mental fatigue, characterized in that, It includes an EEG acquisition and processing module and an electrical stimulation module. The EEG acquisition and processing module and the electrical stimulation module are wirelessly connected to form a closed-loop control. The EEG acquisition and processing module is used to acquire EEG signals from the prefrontal cortex leads. After preprocessing the EEG signals, the module uses an onboard lightweight convolutional neural network (CNN) algorithm to classify mental fatigue states and generates control commands based on the classification results, which are then sent to the electrical stimulation module. The electrical stimulation module is used to receive the control command, generate bipolar pulse width modulated transcranial direct current (tPWMDCs), and electrically stimulate the frontal lobe through electrodes to regulate mental fatigue.

2. The system according to claim 1, characterized in that, The EEG acquisition and processing module includes an analog front-end, a signal processing system, and a power management system. The analog front end includes a data acquisition electrode, an electrostatic impedance device, an anti-aliasing circuit, and an analog-to-digital converter (ADC) connected in sequence, used to convert the EEG signal into a digital EEG signal; The signal processing system includes a microcontroller, which is connected to the analog-to-digital converter and the wireless communication module, and is used to execute the preprocessing, lightweight CNN algorithm and control command generation. The power management system includes an isolation circuit, a charging circuit, a power-on / off circuit, and a power module, which provides a stable voltage for the EEG acquisition and processing module.

3. The system according to claim 2, characterized in that, The electrostatic impedance device is model ESD9B3.3ST5G, used for electrostatic protection; the analog-to-digital converter is model ADS1299, used to convert analog EEG signals into high-precision digital EEG signals.

4. The system according to claim 1, characterized in that, The electrical stimulation module includes a programmable constant current source, a signal processing system, and a power management system. The signal processing system includes a microcontroller, which is connected to a digital-to-analog converter (DAC) and a wireless communication module, respectively, and is used to receive the control commands and generate DAC control commands and pulse width modulation (PWM) parameters. The programmable constant current source is connected to the digital-to-analog converter and is used to generate the tPWMDCs current according to the DAC control command, PWM parameters and stimulation direction. The power management system is used to provide a stable voltage for the electrical stimulation module.

5. The system according to claim 4, characterized in that, The programmable constant current source includes a differential operational amplifier, a general-purpose operational amplifier, sampling resistors, and a MOS array; The analog voltage generated by the digital-to-analog converter is superimposed with the transcranial stimulation site pressure difference acquired by the general operational amplifier and then input to the differential operational amplifier. After being amplified 1:1, it is output to the sampling resistor. The sampling resistor has an accuracy of 0.1% and a resistance of 1KΩ; The MOS array is used to control the output direction, frequency, and duty cycle of the tPWMDCs current.

6. The system according to claim 2 or 4, characterized in that, The wireless communication module is an HC-42 Bluetooth module; the microcontroller in the EEG acquisition and processing module is an STM32U575VGT6; and the microcontroller in the electrical stimulation module is an STM32F1038T6.

7. A transcranial electrical stimulation control method for regulating mental fatigue, characterized in that, Includes the following steps: S1. Collect EEG signals from the frontal lobe leads of the subject; S2. Preprocess the acquired EEG signal to remove interference signals; S3. Input the preprocessed EEG signal into the onboard lightweight convolutional neural network (CNN) algorithm to classify mental fatigue state and obtain the classification result. S4. Generate an electrical stimulation control command based on the classification result, and wirelessly transmit the control command to the electrical stimulation execution unit; S5. The electrical stimulation execution unit generates a bipolar pulse width modulated transcranial direct current (tPWMDCs) according to the control command, and performs transcranial electrical stimulation on the subject to achieve closed-loop regulation of mental fatigue.

8. The method according to claim 7, characterized in that, The model structure of the lightweight CNN algorithm includes: Four convolutional layers are connected in sequence, with a dropout layer following each convolutional layer; The four convolutional layers are followed by an average pooling layer; The average pooling layer is followed by the first fully connected layer and the second fully connected layer in sequence. The output of the second fully connected layer is connected to the Softmax function, which is used to output the classification probability of mental fatigue state.

9. The method according to claim 8, characterized in that, The deployment process of the lightweight CNN algorithm includes: The trained CNN model is lightweighted using the TensorFlowLite framework to obtain the model parameters; The model parameters are compiled in the Arduino development environment and burned into the ROM of the microcontroller; The model parameters are loaded by the TensorFlowLite interpreter in the microcontroller, and the preprocessed EEG signal is inferred and classified.

10. The method according to claim 7, characterized in that, The preprocessing in step S2 includes: A finite impulse response (FIR) filter is used to remove baseline drift and high-frequency interference. The FIR filter has an order of 500 and a bandwidth of 0.5-45Hz. Wavelet transform combined with Kalman filtering is used to remove eye movement artifacts.

Images