Bioelectric signal acquisition method and device based on gel electrode and neural network

Abstract

The application discloses a biological electric signal collection method and device based on a gel electrode and a neural network, and the method comprises the following steps: selecting a suitable gel electrode according to the type of the collected biological electric signal; preparing a corresponding biological sample according to the type of the biological electric signal (such as a human body, an animal, etc.) to be collected; collecting the biological electric signal by using the gel electrode, converting the signal into a digital signal, and saving the digital signal on a local storage; compressing and representing the biological electric signal by using a feature compression coding network; transmitting the compressed biological electric signal to a cloud storage through a network; and managing the biological electric signals from different collection devices by using a cloud file management system, so that a user can select and download data according to permissions. The application combines various technical means such as biological electric signal collection and artificial intelligence, solves the inconvenience of traditional electrodes by designing the gel electrode, and solves the non-optimized storage scheme by designing the feature compression coding network.

Description

Technical Field

[0001] This invention belongs to the field of bioelectric signal acquisition and artificial intelligence, and in particular relates to a bioelectric signal acquisition method and device based on gel electrodes and neural networks. Background Technology

[0002] Bioelectric signals are electrical signals within living organisms, including electrocardiogram (ECG), electroencephalogram (EEG), and electromyogram (EMG). These signals can be collected and analyzed using various bioelectric signal acquisition devices, with applications primarily including clinical medicine, neuroscience, motor control, human-computer interaction, and cognitive science research.

[0003] The commonly used electrode types in the acquisition of bioelectrical signals mainly include dry electrodes and wet electrodes. Dry electrodes are typically made of conductive materials such as metal or carbon fiber and are directly attached to the scalp to collect EEG signals. The advantage of dry electrodes is their ease of operation. However, dry electrodes have a relatively low signal-to-noise ratio, which significantly affects signal quality. Wet electrodes are also typically made of conductive materials such as metal or carbon fiber, but require the use of conductive media such as electrode saline or gel to attach them to the scalp for EEG signal acquisition. The advantage of wet electrodes is a relatively high signal-to-noise ratio, enabling the acquisition of higher quality signals. However, compared to dry electrodes, their use and maintenance are slightly more complex, and the user's skin usually needs to be cleaned after signal acquisition. Furthermore, bioelectrical signals are mostly stored locally after acquisition, which poses significant challenges for their management and analysis. The sampling rate of bioelectrical signals is generally high, resulting in a large data volume that cannot be transmitted to the cloud using conventional network transmission methods.

[0004] In summary, the bioelectrical acquisition process suffers from the inconvenience of traditional electrodes and the lack of optimized storage solutions. Summary of the Invention

[0005] The purpose of this invention is to solve the problems of inconvenience of traditional electrodes and non-optimal storage schemes in the process of bioelectric signal acquisition, and to provide a bioelectric signal acquisition method and device based on gel electrodes and neural networks.

[0006] To achieve the above-mentioned objectives, the embodiments of the present invention provide the following technical solutions:

[0007] Bioelectric signal acquisition methods based on gel electrodes and neural networks include:

[0008] Select the appropriate gel electrode according to the type of bioelectrical signal collected. Multiple electrodes are needed to collect an electroencephalogram (EEG) and are placed in different positions on the scalp to obtain electrical signals from different areas of the brain.

[0009] Prepare the appropriate biological samples according to the type of bioelectrical signal to be collected (such as human, animal, etc.). When collecting human electroencephalograms, the subject should lie down and rest to relax the body and avoid activities that may interfere with signal collection.

[0010] Bioelectrical signals are collected using gel electrodes, converted into digital signals, and stored in local memory. During this process, it is necessary to maintain the stability of the biological sample and avoid interference.

[0011] Bioelectric signals are compressed and represented using a feature compression coding network;

[0012] The compressed bioelectric signals are transmitted to a cloud storage device via a network.

[0013] The cloud-based file management system manages bioelectrical signals from different acquisition devices, and users can select and download data according to their permissions.

[0014] In the above scheme, the gel electrode is composed of a gel solution and a conductive agent. The gel solution is a water-soluble polymer material, composed of polyacrylamide, polyimide gel, gelatin, agar, and water. The conductive agent is an electrolyte, composed of sodium chloride, potassium chloride, silver chloride, and ammonium chloride. The preparation process of the gel electrode is as follows:

[0015] (1) Gel solution preparation: Mix gel material and water in a certain proportion, heat and stir until completely dissolved to prepare gel solution. The proportions of polyacrylamide, polyimide gel, gelatin, agar and water are controlled at 18%, 12%, 10%, 10% and 50%, respectively. (2) Addition of conductive agent: Take out 1000 ml of the prepared gel solution, add 10 ml of sodium chloride solution (0.2 g sodium chloride dissolved in 10 ml deionized water) and heat and stir to mix. Add 10 ml of potassium chloride solution (0.3 g potassium chloride dissolved in 10 ml deionized water) and heat and stir to mix. Add 10 ml of silver chloride solution (0.3 g silver chloride dissolved in 10 ml deionized water) and heat and stir to mix. Add 10 ml of ammonium chloride solution (0.1 g ammonium chloride dissolved in 10 ml deionized water) and heat and stir to mix, so that it is fully and evenly dispersed, thereby forming a conductive gel material. (3) Electrode preparation: The prepared gel is applied to the electrode base to form an electrode-gel composite. Then it is dried or cured. After drying or curing, the gel electrode is taken out, and the gel electrode is prepared.

[0016] It is important to note that careful selection and proportion of materials, as well as precise preparation techniques, are crucial in the fabrication of gel electrodes to ensure their quality and stability. Furthermore, the prepared gel electrodes should be stored in a dry, light-protected, and low-temperature environment to extend their lifespan. During the preparation process, parameters such as the ratio of gel material and conductive agent, preparation temperature, and stirring speed must be controlled to ensure good conductivity and stability. Additionally, to further enhance the lifespan and stability of the gel electrodes, a protective coating, such as polypropylene or silicone, can be applied to their surface. Compared to traditional metal electrodes, the gel electrode proposed in this invention exhibits better stability and lower noise levels. Moreover, the gel electrode proposed in this invention is convenient, non-invasive, and reusable.

[0017] In the above scheme, the feature compression coding network consists of an encoder network, a decoder network, and an objective function.

[0018] More specifically, the structure of the encoder network is as follows:

[0019] (1) Input layer: The raw data of bioelectric signals are input into the neural network. (2) Hidden layer: It consists of multiple correlation convolutional layers and pooling layers. The number of neurons and the size of the convolutional kernel in each correlation convolutional layer and pooling layer can be adjusted according to the actual situation. The correlation convolutional layer is different from traditional two-dimensional convolution and three-dimensional convolution. The correlation convolutional layer first calculates typical correlation values ​​for the corresponding row elements and column elements of a small region in the input data and normalizes them to between 0 and 1. Then, the correlation results are added to obtain a scalar value. Then, the convolutional kernel is slid to the next region of the input data and the above steps are repeated. Finally, the above steps are repeated until the convolutional kernel covers the entire input data and a new feature map is obtained. (3) Residual coding layer: After feature extraction, a residual encoder is added. (4) Latent space: It consists of a fully connected layer with a small number of neurons. The encoded feature data is mapped to a low-dimensional space to achieve signal compression.

[0020] More specifically, the structure of the decoder network is as follows:

[0021] (1) Hidden layer, which consists of multiple deconvolutional layers and upsampling layers, maps the feature vectors in the latent space back to the original data space;

[0022] (2) Residual decoding layer: A residual decoder is added after the deconvolution layer;

[0023] (3) Output layer: The output of the decoder is compared with the input data to calculate the reconstruction error.

[0024] More specifically, the objective function consists of two parts: the encoder's loss function and the decoder's loss function.

[0025] The encoder's loss function primarily refers to how the encoder generates a compressed signal similar to the original signal. During training, the encoder learns to generate compressed signals by minimizing the distance between itself and the original signal. The encoder's loss function can be divided into two parts: one part is the difference between the compressed signal generated by the encoder and the original signal, which is determined using canonical correlation analysis; the other part is the probability that the generated compressed signal is identified as a true signal. The encoder's loss function is expressed in the form of a negative log-likelihood function.

[0026] The loss function of the decoder can also be divided into two parts: one part is the sum of the probability that the decoder identifies the generated signal as the original signal and the probability that the decoder identifies the original signal as the real signal; the other part is a nonlinear logarithmic constraint on the decoder. The loss function of the decoder is expressed in the form of a negative log-likelihood function.

[0027] With this structure, the encoder can extract the most important feature information from the raw bioelectrical signal data and compress it into a smaller potential space, thereby minimizing the data volume and facilitating storage and processing. At the same time, the encoder has a certain degree of robustness, maintaining good compression performance even if the data exhibits some noise and distortion.

[0028] More specifically, the bioelectrical signal compression method based on the aforementioned feature compression coding network is as follows:

[0029] (1) Data preprocessing: The collected bioelectric signals are preprocessed, such as noise removal, filtering, downsampling, etc.;

[0030] (2) Network Construction: An encoder is a neural network structure consisting of an encoder and a decoder. The encoder compresses the input signal into a low-dimensional representation, while the decoder restores the low-dimensional representation back to the original signal. The number of layers, nodes, and other parameters of the autoencoder are designed according to requirements.

[0031] (3) Network Training: Using pre-processed bioelectrical signal data, the network model is trained so that the encoder can compress the original signal into a low-dimensional representation, and the decoder can restore the low-dimensional representation back to the original signal. During training, the backpropagation algorithm is used to optimize the network weights and biases.

[0032] (4) Compressing bioelectric signals: Using a pre-trained network, the collected bioelectric signals are compressed to obtain a low-dimensional representation;

[0033] (5) Restore bioelectric signals: Use a decoder to restore the compressed signals to the original signals.

[0034] More specifically, bioelectrical signal acquisition devices based on gel electrodes and neural networks include:

[0035] (1) A gel electrode unit, which enhances contact with the organism by using gel electrodes, improves the accuracy and stability of signal acquisition, and increases the conductivity of the signal. The gel electrode unit typically consists of multiple gel electrodes, which are usually placed on the scalp or other surface of the organism of interest;

[0036] (2) Signal amplification unit: This unit amplifies the bioelectric signal at the front end to enhance its amplitude and clarity. The signal amplification unit consists of an amplifier and a filter, which amplifies and filters the bioelectric signal to improve its quality;

[0037] (3) Local storage unit, which stores the collected bioelectrical signals in a local storage medium for subsequent processing and analysis. The local storage unit typically includes storage media such as hard disks and flash memory;

[0038] (4) A data compression unit, which compresses the acquired bioelectrical signals to reduce the cost of data storage and transmission. The data compression unit is equipped with a feature compression coding network.

[0039] (5) Network transmission unit: This unit transmits the collected bioelectrical signals to a remote server via a network. The network transmission unit includes a network interface and communication protocol components, which enable reliable data transmission and reception;

[0040] (6) Cloud storage unit: This unit stores the collected bioelectrical signals in a cloud storage medium for convenient remote access and processing. The cloud storage unit consists of a cloud server and a database, which enable efficient storage and management of the data;

[0041] (7) A cloud-based file management system unit, which manages and analyzes the collected bioelectrical signals. The cloud-based file management system unit consists of file management software and data analysis tools. These components can manage and analyze data to meet the needs of different application scenarios.

[0042] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0043] The bioelectrical signal acquisition device based on gel electrodes and neural networks proposed in this invention can be divided into: a gel electrode unit, a signal amplification unit, a local storage unit, a data compression unit, a network transmission unit, a cloud storage unit, and a cloud file management system unit. The acquisition device utilizes gel electrodes, effectively reducing user discomfort during the acquisition process. The gel electrodes proposed in this invention differ from the dry motors and wet electrodes used in traditional bioelectrical signal acquisition devices, effectively solving the inconvenience problem of traditional electrodes. Simultaneously, this invention uses a feature compression coding network to convert the raw bioelectrical signal into a compressed signal, which is then stored in the cloud, effectively addressing the issue of suboptimal storage solutions. Attached Figure Description

[0044] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is a flowchart of the bioelectric signal acquisition method based on gel electrodes and neural networks of the present invention;

[0046] Figure 2 This is a flowchart illustrating the preparation process of the gel electrode of the present invention;

[0047] Figure 3 This is a flowchart of the bioelectrical signal compression method using the feature compression coding network of the present invention;

[0048] Figure 4 This is a structural diagram of the bioelectric signal acquisition device based on gel electrodes and neural networks of the present invention. Implementation

[0049] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments. Example 1

[0050] This invention is achieved through the following technical solution, please refer to [link / reference]. Figure 1 Bioelectric signal acquisition methods based on gel electrodes and neural networks include:

[0051] Select the appropriate gel electrode according to the type of bioelectrical signal collected. Multiple electrodes are needed to collect an electroencephalogram (EEG) and are placed in different positions on the scalp to obtain electrical signals from different areas of the brain.

[0052] Prepare the appropriate biological samples according to the type of bioelectrical signal to be collected (such as human, animal, etc.). When collecting human electroencephalograms, the subject should lie down and rest to relax the body and avoid activities that may interfere with signal collection.

[0053] Bioelectrical signals are collected using gel electrodes, converted into digital signals, and stored in local memory. During this process, it is necessary to maintain the stability of the biological sample and avoid interference.

[0054] Bioelectric signals are compressed and represented using a feature compression coding network;

[0055] The compressed bioelectric signals are transmitted to a cloud storage device via a network.

[0056] The cloud-based file management system manages bioelectrical signals from different acquisition devices, and users can select and download data according to their permissions.

[0057] Specifically, the gel electrode consists of a gel solution and a conductive agent. The gel solution is a water-soluble polymer material composed of polyacrylamide, polyimide gel, gelatin, agar, and water. The conductive agent is an electrolyte composed of sodium chloride, potassium chloride, silver chloride, and ammonium chloride. The preparation process of the gel electrode is as follows; please refer to [link to relevant documentation]. Figure 2 (1) Gel solution preparation: Mix gel material and water in a certain proportion, heat and stir until completely dissolved to prepare gel solution. The proportions of polyacrylamide, polyimide gel, gelatin, agar and water are controlled at 18%, 12%, 10%, 10% and 50%, respectively. (2) Addition of conductive agent: Take out 1000 ml of the prepared gel solution, add 10 ml of sodium chloride solution (0.2 g sodium chloride dissolved in 10 ml deionized water) and heat and stir to mix. Add 10 ml of potassium chloride solution (0.3 g potassium chloride dissolved in 10 ml deionized water) and heat and stir to mix. Add 10 ml of silver chloride solution (0.3 g silver chloride dissolved in 10 ml deionized water) and heat and stir to mix. Add 10 ml of ammonium chloride solution (0.1 g ammonium chloride dissolved in 10 ml deionized water) and heat and stir to mix, so that it is fully and evenly dispersed, thereby forming a conductive gel material. (3) Electrode preparation: The prepared gel is applied to the electrode base to form an electrode-gel composite. Then it is dried or cured. After drying or curing, the gel electrode is taken out, and the gel electrode is prepared.

[0058] It is important to note that careful selection and proportion of materials, as well as precise preparation techniques, are crucial in the fabrication of gel electrodes to ensure their quality and stability. Furthermore, the prepared gel electrodes should be stored in a dry, light-protected, and low-temperature environment to extend their lifespan. During the preparation process, parameters such as the ratio of gel material and conductive agent, preparation temperature, and stirring speed must be controlled to ensure good conductivity and stability. Additionally, to further enhance the lifespan and stability of the gel electrodes, a protective coating, such as polypropylene or silicone, can be applied to their surface. Compared to traditional metal electrodes, the gel electrode proposed in this invention exhibits better stability and lower noise levels. Moreover, the gel electrode proposed in this invention is convenient, non-invasive, and reusable.

[0059] Specifically, the feature compression coding network consists of an encoder network, a decoder network, and an objective function.

[0060] More specifically, the structure of the encoder network is as follows: (1) Input layer, which inputs the raw data of bioelectric signals into the neural network; (2) Hidden layer, which consists of multiple correlation convolutional layers and pooling layers. The number of neurons and the size of the convolutional kernel in each correlation convolutional layer and pooling layer can be adjusted according to the actual situation. The correlation convolutional layer is different from traditional two-dimensional convolution and three-dimensional convolution. The correlation convolutional layer first calculates typical correlation values ​​for the corresponding row elements and column elements of a small region in the input data, and normalizes them to between 0 and 1. Then, the correlation results are added to obtain a scalar value. After that, the convolutional kernel is slid to the next region of the input data and the above steps are repeated. Finally, the above steps are repeated until the convolutional kernel covers the entire input data and a new feature map is obtained. (3) Residual coding layer, which adds a residual encoder after feature extraction; (4) Latent space, which consists of a fully connected layer with a small number of neurons, and maps the encoded feature data to a low-dimensional space to achieve signal compression.

[0061] More specifically, the structure of the decoder network is as follows: (1) a hidden layer, consisting of multiple deconvolutional layers and upsampling layers, which maps the feature vectors in the latent space back to the original data space; (2) a residual decoding layer, which adds a residual decoder layer after the deconvolutional layer; and (3) an output layer, which compares the output of the decoder with the input data to calculate the reconstruction error.

[0062] More specifically, the objective function consists of two parts: the encoder's loss function and the decoder's loss function.

[0063] The encoder's loss function primarily refers to how the encoder generates a compressed signal similar to the original signal. During training, the encoder learns to generate compressed signals by minimizing the distance between itself and the original signal. The encoder's loss function can be divided into two parts: one part is the difference between the compressed signal generated by the encoder and the original signal, which is determined using canonical correlation analysis; the other part is the probability that the generated compressed signal is identified as a true signal. The encoder's loss function is expressed in the form of a negative log-likelihood function.

[0064] The loss function of the decoder can also be divided into two parts: one part is the sum of the probability that the decoder identifies the generated signal as the original signal and the probability that the decoder identifies the original signal as the real signal; the other part is a nonlinear logarithmic constraint on the decoder. The loss function of the decoder is expressed in the form of a negative log-likelihood function.

[0065] With this structure, the encoder can extract the most important feature information from the raw bioelectrical signal data and compress it into a smaller potential space, thereby minimizing the data volume and facilitating storage and processing. At the same time, the encoder has a certain degree of robustness, maintaining good compression performance even if the data exhibits some noise and distortion.

[0066] More specifically, the bioelectrical signal compression method based on the aforementioned feature compression coding network is as follows; please refer to [link / reference]. Figure 3 (1) Data preprocessing: The collected bioelectric signals are preprocessed, such as noise removal, filtering, downsampling, etc.; (2) Network construction: The encoder is a neural network structure consisting of an encoder and a decoder. The encoder compresses the input signal into a low-dimensional representation, and the decoder restores the low-dimensional representation back to the original signal. The number of layers, nodes, and other parameters of the autoencoder are designed as needed; (3) Network training: The network model is trained using the pre-processed bioelectric signal data so that the encoder can compress the original signal into a low-dimensional representation, and the decoder can restore the low-dimensional representation back to the original signal. The backpropagation algorithm is used during training to optimize the weights and biases of the network; (4) Bioelectric signal compression: The collected bioelectric signals are compressed using the pre-trained network to obtain a low-dimensional representation; (5) Bioelectric signal restoration: The compressed signal is restored back to the original signal using the decoder. Example 2

[0067] This invention also provides a bioelectrical signal acquisition device based on gel electrodes and neural networks, including, please refer to, Figure 4 :

[0068] (1) A gel electrode unit, which enhances contact with the organism by using gel electrodes, improves the accuracy and stability of signal acquisition, and increases the conductivity of the signal. The gel electrode unit typically consists of multiple gel electrodes, which are usually placed on the scalp or other surface of the organism of interest;

[0069] (2) Signal amplification unit: This unit amplifies the bioelectric signal at the front end to enhance its amplitude and clarity. The signal amplification unit consists of an amplifier and a filter, which amplifies and filters the bioelectric signal to improve its quality;

[0070] (3) Local storage unit, which stores the collected bioelectrical signals in a local storage medium for subsequent processing and analysis. The local storage unit typically includes storage media such as hard disks and flash memory;

[0071] (4) A data compression unit, which compresses the acquired bioelectrical signals to reduce the cost of data storage and transmission. The data compression unit is equipped with a feature compression coding network.

[0072] (5) Network transmission unit: This unit transmits the collected bioelectrical signals to a remote server via a network. The network transmission unit includes a network interface and communication protocol components, which enable reliable data transmission and reception;

[0073] (6) Cloud storage unit: This unit stores the collected bioelectrical signals in a cloud storage medium for convenient remote access and processing. The cloud storage unit consists of a cloud server and a database, which enable efficient storage and management of the data;

[0074] (7) A cloud-based file management system unit, which manages and analyzes the collected bioelectrical signals. The cloud-based file management system unit consists of file management software and data analysis tools. These components can manage and analyze data to meet the needs of different application scenarios. Example 3

[0075] To verify the effectiveness and implementation of the bioelectric signal acquisition method and device based on gel electrodes and neural networks proposed in this application, the following scientific experiments were conducted.

[0076] Experimental objective: To verify the effectiveness of the bioelectric signal acquisition method and device based on gel electrodes and neural networks in bioelectric signal acquisition.

[0077] Experimental equipment: The bioelectric signal acquisition device based on gel electrodes and neural networks proposed in this application, and the bioelectric signal acquisition device using conventional electrodes.

[0078] Recruitment and grouping of subjects: Ten healthy adults aged 20 to 40 years with no history of neurological or muscular diseases were selected. Subjects were randomly divided into two groups: an experimental group, where bioelectrical signals were acquired using a gel electrode and neural network-based bioelectrical signal acquisition device; and a control group, where bioelectrical signals were acquired using conventional electrodes.

[0079] Experimental procedure:

[0080] Subjects signed informed consent forms and completed necessary questionnaires before the experiment to confirm that they met the experimental conditions;

[0081] Subjects will sit upright in a quiet, non-stimulating environment with their hands placed on the table;

[0082] Researchers will place two types of electrodes on the back of the subject's hand: gel-based electrodes and conventional electrodes;

[0083] Researchers will collect bioelectric signals using a bioelectric signal acquisition device and simultaneously collect signals using a traditional bioelectric signal acquisition device for comparison.

[0084] During the experiment, the researchers will record the basic information of each subject and any factors that may affect the experimental results during the collection process, such as the subject's emotions, time, and environmental factors.

[0085] Data Analysis: Appropriate statistical analysis methods were used to analyze the experimental data to evaluate the differences between the proposed bioelectrical signal acquisition device based on gel electrodes and neural networks and bioelectrical signal acquisition devices using conventional electrodes. The main contents of the data analysis included: the signal-to-noise ratio (SNR) and amplitude differences of the bioelectrical signals; performance parameters of the bioelectrical signal acquisition devices, such as sampling rate and resolution; and a survey of subject subjective experiences, such as assessments of the comfort and ease of use of the two types of electrodes.

[0086] Experimental Results: Bioelectrical signals acquired using a gel electrode and neural network-based bioelectrical signal acquisition device exhibited a better signal-to-noise ratio and higher amplitude, demonstrating superior performance compared to devices using traditional electrodes. Furthermore, subjects rated the gel electrode acquisition device for greater comfort and ease of use. These results indicate that gel electrode and neural network-based bioelectrical signal acquisition methods and devices offer greater feasibility and reliability, as well as a better user experience.

[0087] The above description is merely a preferred embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A bioelectrical signal acquisition method based on gel electrodes and neural networks, characterized in that, Includes the following steps: The type of bioelectrical signal collected is selected based on the type of gel electrode. Multiple electrodes are required for collecting electroencephalograms and are placed in different positions on the scalp. Preparation of human subjects: When collecting human electroencephalograms, subjects should lie down and rest, relax their bodies, and avoid activities that may interfere with signal acquisition. Bioelectrical signals are collected using gel electrodes, converted into digital signals, and stored in local memory, while maintaining the stability of biological samples and avoiding interference. Bioelectric signals are compressed and represented using a feature compression coding network; The compressed bioelectric signals are transmitted to a cloud storage device via a network. The cloud-based file management system manages bioelectrical signals from different acquisition devices, and users can select and download data according to their permissions; The gel electrode consists of a gel solution and a conductive agent; the gel solution is a water-soluble polymer material composed of polyacrylamide, polyimide gel, gelatin, agar, and water; the conductive agent is an electrolyte composed of sodium chloride, potassium chloride, silver chloride, and ammonium chloride. The preparation process of the gel electrode is as follows: (1) Gel solution preparation: Mix the gel material and water according to the ratio, heat and stir until completely dissolved to prepare a gel solution, wherein the proportions of polyacrylamide, polyimide gel, gelatin, agar and water are 18%, 12%, 10%, 10% and 50%, respectively; (2) Addition of conductive agent: Take out 1000 ml of the prepared gel solution, add 10 ml of sodium chloride solution and heat and stir to mix, add 10 ml of potassium chloride solution and heat and stir to mix, add 10 ml of silver chloride solution and heat and stir to mix, add 10 ml of ammonium chloride solution and heat and stir to mix, so that it can be mixed. The gel is uniformly dispersed to form a conductive gel material; the sodium chloride solution is prepared by dissolving 0.2 g of sodium chloride in 10 ml of deionized water, the potassium chloride solution is prepared by dissolving 0.3 g of potassium chloride in 10 ml of deionized water, the silver chloride solution is prepared by dissolving 0.3 g of silver chloride in 10 ml of deionized water, and the ammonium chloride solution is prepared by dissolving 0.1 g of ammonium chloride in 10 ml of deionized water; (3) Electrode preparation: the prepared gel is coated on the electrode base to form an electrode-gel composite, and then dried or cured. After drying or curing, the gel electrode is taken out, and the gel electrode is prepared. The feature compression coding network consists of an encoder network, a decoder network, and an objective function; The structure of the encoder network is as follows: (1) Input layer, which inputs the raw data of bioelectric signals into the neural network; (2) Hidden layer, which consists of multiple correlation convolutional layers and pooling layers; the correlation convolutional layer is different from traditional two-dimensional convolution and three-dimensional convolution; the correlation convolutional layer first calculates typical correlation values ​​for the corresponding row elements and column elements of a small region in the input data and normalizes them to between 0 and 1, then adds the correlation results to obtain a scalar value, then slides the convolutional kernel to the next region of the input data and repeats the above steps, and finally repeats the above steps until the convolutional kernel covers the entire input data and obtains a new feature map; (3) Residual coding layer, which adds a residual encoder after feature extraction; (4) Latent space, which consists of a fully connected layer with a small number of neurons, which maps the encoded feature data to a low-dimensional space to achieve signal compression; The structure of the decoder network is as follows: (1) Hidden layer, which consists of multiple deconvolutional layers and upsampling layers, maps the feature vectors in the latent space back to the original data space; (2) Residual decoding layer: A residual decoder is added after the deconvolution layer; (3) Output layer: compares the output of the decoder with the input data and calculates the reconstruction error; The objective function consists of two parts: the encoder's loss function and the decoder's loss function; The encoder's loss function refers to the compressed signal that the encoder produces that is similar to the original signal. During training, the encoder learns to generate compressed signals by minimizing the distance between itself and the original signal. The encoder's loss function consists of two parts: one part is the difference between the compressed signal generated by the encoder and the original signal, which is determined using canonical correlation analysis; the other part is the probability that the generated compressed signal is identified as a real signal. The encoder's loss function is expressed in the form of a negative log-likelihood function. The loss function of the decoder consists of two parts: one part is the sum of the probability that the decoder identifies the generated signal as the original signal and the probability that the decoder identifies the original signal as the real signal, and the other part is the nonlinear logarithmic constraint on the decoder; the loss function of the decoder is expressed in the form of a negative log-likelihood function. The bioelectric signal compression method based on the feature compression coding network is as follows: (1) Data preprocessing: The collected bioelectric signals are preprocessed to remove noise, filter, and downsample; (2) Network construction: The encoder is a neural network structure consisting of an encoder and a decoder; the encoder compresses the input signal into a low-dimensional representation, and the decoder restores the low-dimensional representation to the original signal; the number of layers and nodes of the autoencoder are designed; (3) Network training: The network model is trained using the pre-processed bioelectric signal data, the encoder compresses the original signal into a low-dimensional representation, and the decoder restores the low-dimensional representation to the original signal; the backpropagation algorithm is used during training to optimize the weights and biases of the network; (4) Bioelectric signal compression: The collected bioelectric signals are compressed using the pre-trained network to obtain a low-dimensional representation; (5) Bioelectric signal restoration: The compressed signal is restored to the original signal using the decoder.

2. A bioelectrical signal acquisition device based on gel electrodes and neural networks, used to implement the method of claim 1, characterized in that, include: (1) Gel electrode unit, which enhances contact with the organism by using gel electrodes, improves the accuracy and stability of signal acquisition, and increases the conductivity of the signal; the gel electrode unit consists of multiple gel electrodes, which are placed on the scalp or other organisms of interest. (2) Signal amplification unit, which amplifies the bioelectric signal at the front end to enhance the amplitude and clarity of the signal; the signal amplification unit consists of an amplifier and a filter, which amplifies and filters the bioelectric signal to improve the signal quality; (3) Local storage unit, which stores the collected bioelectric signals in a local storage medium for subsequent processing and analysis; the local storage unit includes a hard disk and a flash memory storage medium; (4) Data compression unit, which compresses the collected bioelectric signals to reduce the cost of data storage and transmission; the data compression unit is equipped with a feature compression coding network. (5) Network transmission unit: This unit transmits the collected bioelectric signals to a remote server via a network; the network transmission unit includes a network interface and communication protocol components to reliably transmit and receive data; (6) Cloud storage unit, which stores the collected bioelectric signals in the cloud storage medium for easy remote access and processing; The cloud storage unit consists of a cloud server and a database, which efficiently stores and manages data. (7) Cloud file management system unit, which manages and analyzes the collected bioelectric signals; the cloud file management system unit consists of file management software and data analysis tools to manage and analyze the data.

Images