Network architecture searching method and system for electroencephalogram signal processing

Abstract

The invention provides a network architecture searching method and system for electroencephalogram signal processing. Effective migration from high-channel EEG data to low-channel EEG data is achieved. Especially for children autism patients, since the children autism patients are difficult to cooperate with EEG signal acquisition for a long time, data incompleteness may be caused. According to the method, neural network architecture search is performed by using the multi-agent depth deterministic strategy gradient algorithm, and the optimal neural network architecture is searched, so that relatively high classification accuracy is kept in the process of migrating high-channel data to low-channel data.

Description

Technical Field

[0001] The present invention relates to the technical field of electroencephalogram (EEG) signal processing, and in particular to a network architecture search method and system for electroencephalogram (EEG) signal processing. Background Art

[0002] Autism Spectrum Disorder (ASD) is a broad neurodevelopmental disorder that includes core symptoms such as social communication disorders, language communication disorders, and repetitive stereotyped behaviors. Due to the large number of patients with autism spectrum disorder (hereinafter referred to as autism), early screening and accurate diagnosis are of great significance to the rehabilitation and intervention of patients.

[0003] However, the current assessment and diagnosis of autism mainly relies on expert interviews and behavioral scales. These methods are usually highly subjective and inefficient, and cannot meet the needs of large-scale screening. Electroencephalography (EEG), as a means of recording the electrophysiological activity of the cerebral cortex, has the advantages of being non-invasive, easy to collect, and having high temporal resolution. Through feature engineering and traditional machine learning techniques, automatic identification of people with autism can be achieved. Its technical path includes: (1) preprocessing. Filtering is used to remove artifacts and retain signals in specific frequency bands to obtain more effective features; (2) feature extraction. Information is extracted from the signal through analysis methods such as time, frequency, time-frequency, and entropy to form a feature set; (3) feature selection. Filtering or wrapping feature selection methods are used to select effective feature subsets; (4) autism identification. By training a classifier or a combination of classifiers, the classifier parameters are optimized using labeled data, and the optimized model is used to identify unlabeled samples. The advantage of this solution is that the training complexity is low and it does not rely on large-scale data and high-performance computing; but its disadvantage is that it requires high prior knowledge and professional skills.

[0004] With the improvement of big data and computing power, deep learning technologies such as Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) have made remarkable achievements in the fields of computer vision and natural language processing, and even surpassed human levels in some tasks. Therefore, applying deep learning technology to EEG signal processing to achieve high-precision discrimination of autism has become a research hotspot in the field of brain health. However, high-density EEG acquisition equipment is expensive and will increase the burden on subjects, especially children with autism, resulting in a significant increase in the cost and difficulty of EEG signal acquisition. At the same time, some EEG channels have no significant effect on the diagnosis of autism, but will reduce the efficiency and accuracy of model training.

[0005] Therefore, for the application of deep learning technology in the field of EEG signal processing and classification, it is necessary to propose a method with wider adaptability and higher accuracy. Summary of the invention

[0006] The present invention provides a network architecture search method and system for electroencephalogram (EEG) signal processing, which are used to solve the defects of the network architecture adopted in the prior art when deep learning technology is applied in the field of EEG signal processing classification, that is, the lack of universality and high precision.

[0007] In a first aspect, the present invention provides a network architecture search method for EEG signal processing, comprising: Collecting EEG signal data of the target subject; Preprocessing the EEG signal data to obtain preprocessed EEG signal data; Using the preprocessed EEG signal data, a multi-agent deep deterministic policy gradient algorithm is used to perform architecture search training on the neural network model to obtain an original EEG signal acquisition model; The original EEG signal acquisition model is evaluated and optimized to determine the optimal EEG signal acquisition model.

[0008] According to a network architecture search method for EEG signal processing provided by the present invention, EEG signal data of a target subject is collected, including: Collect high-density 62-channel 1000Hz resting-state EEG signal data.

[0009] According to a network architecture search method for EEG signal processing provided by the present invention, the EEG signal data is preprocessed to obtain preprocessed EEG signal data, including: Downsampling, filtering and removing artifacts on the EEG signal data in sequence to obtain filtered EEG signal data; Using a fixed time window to segment the screened EEG signal data into a plurality of EEG segments; Performing discrete Fourier transform and bandpass filtering on the multiple EEG segments respectively to obtain EEG frequency domain signals, and retaining EEG time domain signals; The EEG frequency domain signal and the EEG time domain signal are randomly mixed according to the subjects and divided into a training set, a validation set and a test set.

[0010] According to a network architecture search method for EEG signal processing provided by the present invention, the pre-processed EEG signal data is used to perform architecture search training on the neural network model using a multi-agent deep deterministic policy gradient algorithm to obtain an original EEG signal acquisition model, including: Determining that the neural network model includes an input layer, a plurality of hidden layers and an output layer, forming a multi-agent decision-making architecture; Searching the neural network model based on the multi-agent deep deterministic policy gradient algorithm, and searching for the optimal network architecture for high- and low-channel EEG data migration through collaboration and strategy optimization among agents; Initial training is performed on preset high-channel EEG data to obtain training data, and the training data is transferred to preset low-channel EEG data for learning to obtain the original EEG signal acquisition model.

[0011] According to a network architecture search method for EEG signal processing provided by the present invention, it is determined that the neural network model includes an input layer, multiple hidden layers and an output layer, forming a multi-agent decision architecture, including: The input layer receives EEG data features; The multiple hidden layers use convolutional layers and pooling layers of different sizes to extract time-frequency features from the EEG data features; The output layer generates classification results using the Softmax function.

[0012] According to a network architecture search method for EEG signal processing provided by the present invention, the neural network model is searched based on the multi-agent deep deterministic policy gradient algorithm, and the optimal network architecture for high- and low-channel EEG data migration is searched through collaboration and strategy optimization between agents, including: Determine that the intelligent agent includes a strategy network and an evaluation network, wherein the input of the strategy network is the intelligent agent observation data, and the output is an action vector adapted to the action space, and the input of the evaluation network is the observation and action of the intelligent agent, and the output is the action quality of the intelligent agent; The evaluation network evaluates the agent action quality Q value by minimizing the loss function and updates the evaluation network; The policy network generates a Q-value-maximizing action by calculating a policy gradient, and updates the policy network; The target network is added to the strategy network and the evaluation network respectively, and soft update is performed to obtain the optimal network architecture for high and low channel EEG data migration after the update.

[0013] According to a network architecture search method for EEG signal processing provided by the present invention, the original EEG signal acquisition model is evaluated and optimized to determine the optimal EEG signal acquisition model, including: The classification accuracy of the original EEG signal acquisition model on the preset low-channel EEG data after migration is calculated on the test set, and the EEG data under different channels are compared to verify the effectiveness of the multi-agent deep deterministic policy gradient algorithm; Multiple rounds of cross-validation and segmented experiments were used to detect the stability of the original EEG signal acquisition model under different input conditions; Perform hyperparameter optimization based on classification accuracy and model stability evaluation results; The selected optimal network architecture is transplanted to a low sampling rate EEG device, and the optimal network architecture is subjected to model lightweight processing to obtain the optimal EEG signal acquisition model.

[0014] In a second aspect, the present invention further provides a network architecture search system for EEG signal processing, comprising: An acquisition module, used to acquire EEG signal data of a target subject; A preprocessing module, used for preprocessing the EEG signal data to obtain preprocessed EEG signal data; A training module, for using the pre-processed EEG signal data to perform architecture search training on the neural network model using a multi-agent deep deterministic policy gradient algorithm to obtain an original EEG signal acquisition model; The optimization module is used to evaluate and optimize the original EEG signal acquisition model to determine the optimal EEG signal acquisition model.

[0015] In a third aspect, the present invention also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements a network architecture search method for electroencephalogram signal processing as described in any one of the above.

[0016] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a network architecture search method for EEG signal processing as described in any one of the above.

[0017] Compared with the prior art, the present invention has the following beneficial effects: The present invention effectively improves the efficiency of network architecture search through collaborative search and strategy optimization among intelligent agents, making the model more adaptable to the complex distribution of EEG data features. By optimizing model parameters through transfer learning, the model can still maintain high classification accuracy on low-channel data, improving the adaptability and versatility of the algorithm; After the model is transplanted to a low sampling rate device, the present invention can still maintain efficient classification performance and real-time response capability, ensuring that EEG signals can be quickly processed and classified in practical applications to meet the needs of real-time analysis; The present invention has high device compatibility and is suitable for EEG devices with different sampling rates and channel numbers. The technical support and deployment documents provided simplify the integration and application of the model, providing technical support for the large-scale promotion of low-cost EEG devices. Through lightweight and optimized design, the model's dependence on hardware is effectively reduced, making it suitable for low-cost EEG devices, providing a cost-effective solution for devices with limited resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 It is a flow chart of a network architecture search method for EEG signal processing provided by the present invention; Figure 2 It is the overall structural principle diagram provided by the present invention; Figure 3 It is a structural schematic diagram of a network architecture search system for EEG signal processing provided by the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] In view of the problems existing in the prior art, the present invention proposes an adaptive neural network architecture search method, which aims to reduce the cooperation requirements of patients for EEG acquisition through an effective method of migrating from high-channel EEG data to low-channel EEG data, and is particularly suitable for the collection and analysis of EEG data of children with autism. The present invention uses the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm to search for neural network architecture, screen out the optimal architecture, realize the migration between high and low channel data, and ensure the high accuracy and reliability of classification.

[0022] Figure 1 is a flow chart of a network architecture search method for EEG signal processing provided by an embodiment of the present invention, such as Figure 1As shown, including: Step 100: Collecting EEG signal data of a target subject; Step 200: preprocessing the EEG signal data to obtain preprocessed EEG signal data; Step 300: Using the pre-processed EEG signal data, and using a multi-agent deep deterministic policy gradient algorithm to perform architecture search training on a neural network model, to obtain an original EEG signal acquisition model; Step 400: Evaluate and optimize the original EEG signal acquisition model to determine the optimal EEG signal acquisition model.

[0023] Specifically, the embodiment of the present invention proposes an adaptive neural network architecture search method for migrating high-channel EEG data to low-channel EEG data. The core of this method is to automatically search for a neural network architecture through a multi-agent deep deterministic policy gradient algorithm to effectively migrate between different channel data and maintain a high classification accuracy during the migration process. The specific implementation steps include: Step 1: Data acquisition. This work uses high-density 62-channel 1000Hz resting-state EEG signal data as the data source for training the model.

[0024] Step 2: Data preprocessing. The data collected by the EEG acquisition device is a 3-5 minute raw signal. The processing required in this step is divided into the following parts: (1) Use downsampling, bandpass filtering and artifact removal algorithms to reduce the impact of noise such as electrooculography and electromyography on subsequent analysis; (2) Use a fixed time window to divide the original signal into multiple time segments as the minimum unit for autism discrimination, and then generate individual discrimination results through voting; (3) Perform discrete Fourier transform processing on each EEG segment to make the frequency domain features more prominent; (4) Divide the data into training set, validation set and test set by randomly selecting different people.

[0025] Step 3: Model construction and training. (1) Network architecture initialization: Based on the multi-layer structure of the neural network, the initial model contains an input layer, multiple hidden layers, and an output layer. (2) MADDPG algorithm training: Use the multi-agent deep deterministic policy gradient algorithm to search for the model architecture. The algorithm searches for the optimal network architecture suitable for the migration of high- and low-channel EEG data through collaboration and policy optimization between agents. (3) Transfer learning: Use high-channel data as input for transfer training to adapt to the feature distribution of low-channel data and optimize model parameters.

[0026] Step 4: Evaluate and optimize the classification accuracy of the model. The specific evaluation indicators include: Classification accuracy: Evaluate the classification accuracy after migration of high and low channel EEG data.

[0027] Model stability: Ensure the stability and adaptability of the model during the migration between different data channels.

[0028] Step 5: Migration and deployment. The optimal neural network architecture finally selected is suitable for the classification task of low-channel EEG data. This model can be deployed on EEG devices with low sampling rates to further reduce the equipment acquisition cost.

[0029] The present invention uses a multi-agent deep deterministic policy gradient algorithm to search for a neural network architecture and find the optimal neural network architecture, so that a high classification accuracy can be maintained during the process of migrating from high-channel data to low-channel data.

[0030] Based on the above embodiment, step 1 includes: 1.1 EEG signal acquisition process EEG signal collection is usually carried out in a professional EEG collection room. Sensors are attached to the surface of the subject's scalp to record the brain's electrophysiological activity data in resting or task states, and the real-time collected signals are transmitted to a computer for storage.

[0031] 1.2 Preliminary preparation and environment setup To ensure the quality of signal collection, the subjects are required to remove items that may affect signal collection, such as glasses, earrings, etc. At the same time, during the collection process, the subjects will be given a headgear made of special materials to ensure that the electrodes can fit the scalp tightly and reduce external interference.

[0032] 1.3 Electrode arrangement and conductive treatment The electrodes corresponding to the 62 channels, a reference electrode and a ground wire were passed through the headgear respectively, so that the electrodes were closely attached to the scalp surface of the subject. In order to further reduce the resistance and improve the stability of signal acquisition, conductive paste was applied to the contact position.

[0033] 1.4 Signal acquisition and recording The collection process requires the subject to maintain steady breathing, sit quietly, and keep their eyes open for 3-5 minutes. The collected EEG signals will be stored together with the subject's basic information for subsequent analysis.

[0034] Based on the above embodiment, step 2 includes: 2.1 Data downsampling and filtering First, the collected EEG data is downsampled to 256Hz to enhance its anti-noise ability and reduce storage and computing resource consumption. Next, a bandpass filter (0.5-45Hz) is used to remove interference signals in non-EEG frequency bands, effectively retaining the main components of EEG activity. In addition, the artifact removal algorithm is used to remove artifact signals such as electrooculography and electromyography in the data to ensure the purity of the data.

[0035] 2.2 Data fragmentation processing The complete EEG recording of each subject was divided into 4-second segments as the basic unit for subsequent processing. In the discrimination process, the voting results of multiple segments were used to generate the final classification judgment of the subject to improve the robustness of individual discrimination.

[0036] 2.3 Time domain and frequency domain feature extraction Discrete Fourier transform (DFT) was performed on each EEG segment to convert the data from the time domain to the frequency domain, making the frequency domain features more prominent while retaining the time domain features. After DFT processing, bandpass filtering was applied again to retain only the frequency band of 0.5-45Hz.

[0037] 2.4 Dataset Division All samples are randomly mixed according to people and divided into training set, validation set and test set (ratio is 3:1:1). The training set is used for model training, the validation set is used for parameter optimization, and the test set is used to evaluate the final model performance.

[0038] Based on the above embodiment, step 3 includes: 3.1 Network architecture initialization Based on the multi-layer structure of the neural network, the initial model design includes an input layer, multiple hidden layers and an output layer, forming a multi-agent decision-making architecture. This model architecture is suitable for high- and low-channel EEG data migration. The input layer receives EEG data features, the hidden layer uses convolutional layers and pooling layers of different sizes to extract time-frequency features in the data, and the output layer uses the Softmax function to generate classification results to complete multi-classification tasks.

[0039] 3.2 Multi-agent Deep Deterministic Policy Gradient Algorithm Training The MADDPG algorithm is used to search the model architecture and conduct collaborative training among multiple agents to achieve the optimal strategy for the EEG data migration task. The MADDPG algorithm specifically includes the Actor network and the Critic network, which are used for action generation and value evaluation respectively.

[0040] 3.2.1 Agent Design Policy network (Actor network): The MADDPG algorithm assigns an independent Actor network to each agent, which inputs from the observation space of the agent and outputs actions in its action space. The input of each Actor network is the observation data of the agent, and the output is the action vector adapted to the action space.

[0041] Evaluation network (Critic network): The Critic network is responsible for evaluating the quality of each agent's actions. In a multi-agent environment, the input of each Critic network includes the observations and actions of all agents, so that the strategies of other agents can be taken into account during the evaluation process, improving the effect of collaborative learning.

[0042] 3.2.2 Evaluation Network Update The evaluation network evaluates the quality of the agent's actions by minimizing the loss function. The loss function of the Critic network is:

[0043] in, Represents the Q value output by the Critic network, and the parameter is ; For TD target, For the current reward, is the discount factor, is the next state, and The actions of all agents in the next state generated by the Actor network. The Critic network optimizes the parameters by minimizing the loss function. .

[0044] 3.3 Policy Network Update The goal of the policy network is to generate the optimal policy so that the agent's actions can maximize the Q value. The policy gradient calculation formula of the Actor network is:

[0045] in, Represents the gradient of the Actor network, and the parameters are ,and is the Q value output by the Critic network, Represents the policy function of the Actor network, which calculates the action of the agent under the current observation. Through back propagation, the Actor network is updated to generate actions that maximize the Q value.

[0046] 3.4 Target Network Soft Update In order to maintain the stability of the network, the target network is added to the Actor and Critic networks, and soft updates are performed. The soft update formula is:

[0047]

[0048] Among them, τ is the soft update coefficient, which is usually small (such as 0.001) and is used to gradually update the parameters of the target network to keep the training stable.

[0049] 3.5 Transfer Learning Initial training is performed on high-channel EEG data, and the knowledge gained from training is applied to low-channel EEG data through transfer learning, which reduces training time. This step retains the data features of high-channel training, and retraining is performed on low-channel data to adjust model parameters so that the model can adapt to the feature distribution of low-channel data, thereby improving classification accuracy and stability.

[0050] Based on the above embodiment, step 4 includes: 4.1 Classification Accuracy Evaluation The model was evaluated by calculating the classification accuracy of the model for low-channel EEG data after transfer on the test set, and EEG data under different channels were used for comparison to verify the effectiveness of the MADDPG algorithm in transfer learning.

[0051] 4.2 Model stability evaluation Multiple rounds of cross-validation and segmented experiments are used to test the stability of the model under different input conditions, including the training convergence speed and performance differences under different numbers of channels, to ensure the adaptability of the migration process and high-performance output.

[0052] 4.3 Optimization and Adjustment Based on the classification accuracy and model stability evaluation results, hyperparameter optimization (such as learning rate, batch size, layer selection and dropout rate adjustment) is performed to improve the model's ability to handle complex features while ensuring model adaptability and stability.

[0053] Based on the above embodiment, step 5 includes: The optimal network architecture finally selected is applied to the classification task of low-channel EEG data. In order to further reduce the equipment acquisition cost, the model is transplanted to EEG equipment with low sampling rate, and the model is lightweight, such as reducing the number of convolution and pooling layers and reducing the number of model parameters. During the deployment process, the model is verified to maintain efficient and accurate classification performance on low-channel EEG data, meeting the requirements of real-time and equipment computing limitations, and providing technical support for the application of low-cost EEG equipment.

[0054] The embodiments of the present invention are described below with specific examples. Figure 2 As shown, including: The first step is data collection. The collection of EEG signals is usually carried out in a professional EEG collection room. To ensure the signal quality, the environment should be kept quiet and the power supply voltage should be stable. Try to avoid choosing subjects with thick or hard hair, and ask them to clean their heads and remove head ornaments before wearing the headgear. Accurately stick the electrodes to the corresponding positions of the subject's scalp, and apply special conductive paste to the contact points to reduce the resistivity and ensure that each resistance is less than 100kΩ (less than 30kΩ in the best case). During the collection process, the subject needs to keep his eyes open, still, and silent, try to keep steady breathing and relaxed, and use air conditioning and other means to avoid sweating and signal drift. Finally, the collected 3 to 5 minutes of data together with the subject's detailed information are saved.

[0055] In this embodiment, the default frequency of EEG data collected is 1000Hz, and this resolution far exceeds the basic requirement of deep neural network for data accuracy. To enhance the anti-noise ability, downsampling is used to reduce the frequency to 256Hz. Subsequently, a bandpass filter is used to remove noise interference above 0-0.5Hz and 45Hz. The EEG recording of each subject is divided into segments of 4 seconds each as the processing unit for subsequent analysis; the tail segment of less than 4 seconds is discarded. When classifying and distinguishing, the discrimination results of more than half of all the segments of the individual are used as the overall discrimination results. After manual screening, the segments with heavy noise or obvious signal drift are eliminated. Then, a discrete Fourier transform is performed on each EEG segment to convert the signal from the time domain to the frequency domain to make the features more significant, and the effective frequency band of 0.5-45Hz is retained, while retaining the original time domain features. For the convenience of subsequent analysis, each sample is adjusted to a three-dimensional structure, and the size is set to [1 × 62 × 196], where 62 represents the number of channels and 196 is the characteristic length of the 0.5-45Hz frequency band. Finally, samples were randomly mixed by individual (not fragment) and divided into training set, validation set and test set in a ratio of 3:1:1. The first two were used for model training and optimization, and the latter was used for model testing. The label format of each sample was [0, 1] or [1,0], corresponding to autistic individuals and normally developing individuals, respectively.

[0056] The model in this embodiment is designed based on a multi-layer neural network structure, which mainly includes an input layer, multiple hidden layers and an output layer. The input layer is responsible for receiving EEG data features and extracting the time and frequency features of EEG signals through different convolutional layers and pooling layers. The hidden layer applies batch normalization and activation functions to enhance the generalization ability of the model and avoid overfitting. At the same time, a dropout layer is set in the hidden layer to randomly discard neurons with a certain probability to further improve the stability of the model. Finally, the output layer uses a Softmax activation function to convert the output into a probability for high and low awareness states or other multi-classification tasks. The architecture of the model supports migration between EEG data with different numbers of channels, which helps to cope with different data input formats. In order to search for the optimal network structure suitable for EEG data migration, the multi-agent deep deterministic policy gradient (MADDPG) algorithm is used for collaborative training. The MADDPG algorithm includes multiple agents, each of which consists of an independent Actor network and a Critic network. (1) Actor network: An independent Actor network is assigned to each agent. The Actor network obtains input from the observation space of the agent and outputs an action vector that adapts to the action space. The input of the policy network is the observation data of each agent, and the output is the action, which indicates the strategy of each agent. (2) Critic network: used to evaluate the action quality of each agent. The input of each Critic network includes the observations and actions of all agents, and the strategies of other agents are taken into account during the evaluation to achieve collaborative optimization. The Critic network adjusts parameters to optimize the Q value of the action by minimizing the loss function. The update goal of the Critic network is to minimize the Q value deviation of the action, that is, to make the policy output in the current state close to the optimal. In order to ensure that the model obtains better feature representation on complex EEG data, a dynamic reward function is used to reward high-precision classification output and encourage the Actor and Critic networks to optimize adaptability. Through the Actor-Critic mechanism in the MADDPG algorithm, the model obtains better migration ability between EEG data with different numbers of channels.

[0057] In order to evaluate and optimize the transfer learning effect of the model on low-channel EEG data, the classification accuracy of the model was first calculated on the test set, and the classification results under different channel numbers were compared to verify the effectiveness of the MADDPG algorithm in transfer learning. Subsequently, multiple rounds of cross-validation and segmented experiments were used to detect the stability of the model under different input conditions, including the performance differences under the training convergence speed and the number of channels, so as to ensure the adaptability and output performance of the model during the migration process. According to the classification accuracy and stability evaluation results, the model's hyperparameters (such as learning rate, batch size, number of hidden layers, and drop rate, etc.) were optimized and adjusted, and finally its ability to handle complex features was improved while maintaining the adaptability and stability of the model.

[0058] The optimal neural network architecture finally selected is suitable for the classification task of low-channel EEG data. In order to reduce the equipment acquisition and computing costs, the model is lightweight and then ported to EEG devices with low sampling rates. The specific processing steps include reducing the number of convolution and pooling layers and optimizing the number of model parameters to reduce storage and computing requirements. In addition, based on the characteristics of low-sampling rate devices, the input data format of the model is readjusted to ensure that the model can adapt to the reduction in sampling rate while maintaining classification accuracy. After the model verifies its performance on low-channel EEG data, it can achieve real-time classification prediction to meet the actual application needs of low-cost EEG devices and provide technical support for deployment on mobile devices or in resource-constrained scenarios.

[0059] The network architecture search system for EEG signal processing provided by the present invention is described below. The network architecture search system for EEG signal processing described below and the network architecture search method for EEG signal processing described above can be referenced to each other.

[0060] Figure 3 is a schematic diagram of the structure of a network architecture search system for EEG signal processing provided by an embodiment of the present invention. Figure 3 As shown, it includes: a collection module 31, a preprocessing module 32, a training module 33 and an optimization module 34, wherein: The acquisition module 31 is used to collect the EEG signal data of the target subject; the preprocessing module 32 is used to preprocess the EEG signal data to obtain the preprocessed EEG signal data; the training module 33 is used to use the preprocessed EEG signal data to perform architecture search training on the neural network model using a multi-agent deep deterministic policy gradient algorithm to obtain the original EEG signal acquisition model; the optimization module 34 is used to evaluate and optimize the original EEG signal acquisition model to determine the optimal EEG signal acquisition model.

[0061] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A network architecture search method for electroencephalogram signal processing, characterized in that: include: Collecting EEG signal data of the target subject; Preprocessing the EEG signal data to obtain preprocessed EEG signal data; Using the preprocessed EEG signal data, a multi-agent deep deterministic policy gradient algorithm is used to perform architecture search training on the neural network model to obtain an original EEG signal acquisition model; The original EEG signal acquisition model is evaluated and optimized to determine the optimal EEG signal acquisition model.

2. The network architecture search method for EEG signal processing according to claim 1, characterized in that: Collect EEG signal data of the target subject, including: Collect high-density 62-channel 1000Hz resting-state EEG signal data.

3. The network architecture search method for EEG signal processing according to claim 1, characterized in that: Preprocessing the EEG signal data to obtain preprocessed EEG signal data includes: Downsampling, filtering and removing artifacts on the EEG signal data in sequence to obtain filtered EEG signal data; Using a fixed time window to segment the screened EEG signal data into a plurality of EEG segments; Performing discrete Fourier transform and bandpass filtering on the multiple EEG segments respectively to obtain EEG frequency domain signals, and retaining EEG time domain signals; The EEG frequency domain signal and the EEG time domain signal are randomly mixed according to the subjects and divided into a training set, a validation set and a test set.

4. The network architecture search method for EEG signal processing according to claim 1, characterized in that: The pre-processed EEG signal data is used to perform architecture search training on the neural network model using a multi-agent deep deterministic policy gradient algorithm to obtain an original EEG signal acquisition model, including: Determining that the neural network model includes an input layer, a plurality of hidden layers and an output layer, forming a multi-agent decision-making architecture; Searching the neural network model based on the multi-agent deep deterministic policy gradient algorithm, and searching for the optimal network architecture for high- and low-channel EEG data migration through collaboration and strategy optimization among agents; Initial training is performed on preset high-channel EEG data to obtain training data, and the training data is transferred to preset low-channel EEG data for learning to obtain the original EEG signal acquisition model.

5. The network architecture search method for EEG signal processing according to claim 4, characterized in that: Determining that the neural network model includes an input layer, multiple hidden layers and an output layer, forming a multi-agent decision-making architecture, including: The input layer receives EEG data features; The multiple hidden layers use convolutional layers and pooling layers of different sizes to extract time-frequency features from the EEG data features; The output layer generates classification results using the Softmax function.

6. The network architecture search method for EEG signal processing according to claim 4, characterized in that: The neural network model is searched based on the multi-agent deep deterministic policy gradient algorithm, and the optimal network architecture for high- and low-channel EEG data migration is searched through collaboration and strategy optimization among agents, including: Determine that the intelligent agent includes a strategy network and an evaluation network, wherein the input of the strategy network is the intelligent agent observation data, and the output is an action vector adapted to the action space, and the input of the evaluation network is the observation and action of the intelligent agent, and the output is the action quality of the intelligent agent; The evaluation network evaluates the agent action quality Q value by minimizing the loss function and updates the evaluation network; The policy network generates a Q-value-maximizing action by calculating a policy gradient, and updates the policy network; The target network is added to the strategy network and the evaluation network respectively, and soft update is performed to obtain the optimal network architecture for high and low channel EEG data migration after the update.

7. The network architecture search method for EEG signal processing according to claim 1, characterized in that: Evaluating and optimizing the original EEG signal acquisition model to determine the optimal EEG signal acquisition model includes: The classification accuracy of the original EEG signal acquisition model on the preset low-channel EEG data after migration is calculated on the test set, and the EEG data under different channels are compared to verify the effectiveness of the multi-agent deep deterministic policy gradient algorithm; Multiple rounds of cross-validation and segmented experiments were used to detect the stability of the original EEG signal acquisition model under different input conditions; Perform hyperparameter optimization based on classification accuracy and model stability evaluation results; The selected optimal network architecture is transplanted to a low sampling rate EEG device, and the optimal network architecture is subjected to model lightweight processing to obtain the optimal EEG signal acquisition model.

8. A network architecture search system for EEG signal processing, characterized in that: include: An acquisition module, used to acquire EEG signal data of a target subject; A preprocessing module, used for preprocessing the EEG signal data to obtain preprocessed EEG signal data; A training module, for using the pre-processed EEG signal data to perform architecture search training on the neural network model using a multi-agent deep deterministic policy gradient algorithm to obtain an original EEG signal acquisition model; The optimization module is used to evaluate and optimize the original EEG signal acquisition model to determine the optimal EEG signal acquisition model.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the network architecture search method for EEG signal processing as described in any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the network architecture search method for EEG signal processing as described in any one of claims 1 to 7 is implemented.

Images