Comprehensive regulation and control method and system based on multi-mode electromagnetic and ultrasonic stimulation

Abstract

The invention discloses a comprehensive regulation and control method and system based on multi-mode electromagnetic and ultrasonic stimulation, and the method comprises the steps: collecting brain wave signals in real time through a multi-electrode EEG system, synchronously collecting skin electroreaction and heart rate variability signals, preprocessing the collected signals, and carrying out the feature extraction; the extracted signal features are subjected to multi-modal feature fusion through a CNN, and the multi-modal features are subjected to time sequence dependency analysis through a long-short-term memory network to generate a real-time emotional state quantization vector and identify a current emotional state category. Through synchronous acquisition and feature fusion of multi-electrode EEG, EDA and HRV signals, a real-time brain function connection network is constructed, accurate recognition and stimulation strategy making are performed on emotion regulation and control nodes based on graph theory analysis, the accuracy and reliability of emotion recognition are improved, personalized regulation and control of individual emotion states are realized, and the emotion regulation and control efficiency is improved. The cognitive function and language expression ability of an individual can be improved, and movement coordination disorder caused by emotional fluctuation is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of emotion regulation, and in particular to a comprehensive regulation method and system based on multi-mode electromagnetic and ultrasonic stimulation. Background Art

[0002] In recent years, emotional state monitoring and regulation technology has received extensive attention and rapid development in the fields of brain-computer interface, neural regulation and psychotherapy. Traditional emotion regulation methods mainly rely on a single signal source to monitor and identify emotional states. Although these methods can reflect individual emotional changes to a certain extent, the accuracy and reliability of emotion recognition are significantly limited because a single signal source cannot fully reflect the complexity and multi-dimensional characteristics of emotional fluctuations. Most of the existing emotion regulation methods are based on simple neural stimulation strategies, and do not fully consider the dynamic change characteristics of brain functional networks under different emotional states. This regulation method based on a fixed stimulation strategy has limited effects in practical applications, especially in complex emotional states (such as anxiety, stress, depression, etc.), where a single stimulation often fails to achieve the expected emotional regulation effect. Therefore, the existing methods cannot dynamically adjust the neural stimulation strategy according to the real-time changes in the emotional state, which easily leads to insufficient generalization of the emotion regulation strategy, making it difficult to achieve accurate and personalized regulation of the individual's emotional state, and easily leading to individual cognitive confusion, thereby affecting the individual's motor coordination ability and language expression ability. Summary of the invention

[0003] In view of the above-mentioned problems existing in the existing comprehensive control method based on multi-mode electromagnetic and ultrasonic stimulation, the present invention is proposed.

[0004] Therefore, the problem to be solved by the present invention is that the existing methods are unable to dynamically adjust the neural stimulation strategy according to the real-time changes in the emotional state, which may easily lead to insufficient generalization ability of the emotion regulation strategy, making it difficult to achieve accurate and personalized regulation of the individual's emotional state, and easily lead to individual cognitive confusion and thus affect the individual's motor coordination ability and language expression ability.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a comprehensive control method based on multi-mode electromagnetic and ultrasonic stimulation, which includes collecting brain wave signals in real time and synchronously collecting skin electrical response and heart rate variability signals through a multi-electrode EEG system, and extracting features after preprocessing the collected signals; The extracted signal features are fused into multimodal features through CNN, and the multimodal features are analyzed for temporal dependency through long short-term memory network to generate real-time emotional state quantification vector and identify the current emotional state category; based on the results of emotional state recognition, the phase coupling characteristics and oscillation synchronization characteristics between brain regions under different emotional states in the EEG signal are extracted to construct a real-time brain functional connection network, and the active emotion regulation nodes in the functional network are identified through graph theory analysis technology. The emotion regulation nodes are divided into surface targets and deep targets, and the emotional activity of each node is quantified, and the stimulation strategy is formulated according to the emotional activity; each emotion regulation data is recorded and a personalized emotion profile is generated for each subject.

[0006] As a preferred solution of the comprehensive control method based on multi-mode electromagnetic and ultrasonic stimulation described in the present invention, wherein: the real-time collection of brain wave signals and synchronous collection of skin electrical response and heart rate variability signals by a multi-electrode EEG system, and feature extraction after preprocessing the collected signals refer to using a multi-electrode EEG system, arranging electrodes in the prefrontal cortex, parietal lobe and hippocampus of the subject according to the 10-20 electrode system standard, recording the EEG signal of each electrode, placing EDA sensors on the fingertips and palms of the subject to collect EDA signals, placing HRV sensors on the chest and wrists to collect HRV signals, setting the sampling frequency of the EEG system, EDA and HRV sensors to 250Hz, and synchronously recording and collecting data; The artifacts of EEG signals were removed by ICA algorithm, and the frequency domain filtering of the EEG signals after artifact removal was performed by bandpass filter. The power spectrum density of EEG signals was extracted by Welch method and time-frequency analysis was performed by wavelet transform. The sample entropy was calculated to measure the complexity of EEG signals. A 3-second window was used to perform sliding window mean filtering on the EDA signal and the HRV signal to remove instantaneous spike signals, extract the skin electrical response characteristics of the EDA signal, and perform frequency domain analysis on the HRV signal through fast Fourier transform to extract low-frequency power, high-frequency power, and LF / HF ratio.

[0007] As a preferred solution of the comprehensive control method based on multi-modal electromagnetic and ultrasonic stimulation of the present invention, wherein: the multi-modal feature fusion of the extracted signal features through CNN refers to normalizing the feature data of EEG, EDA and HRV signals, aligning the normalized modal features according to the time axis and splicing them to generate a splicing matrix X; Expand the concatenated multimodal feature matrix into a three-dimensional input tensor of CNN and construct the input data format of multimodal feature fusion; Constructing a CNN model includes input layer, convolution layer, pooling layer, fully connected layer and output layer; The three-dimensional tensor is convolved with the convolution kernel to obtain the multimodal feature map tensor [J T K], where K is the number of convolution kernels, T is the time step, and J is the number of types of multimodal features; Perform nonlinear activation on the convolutional feature map and use the ReLU activation function to return the negative values ​​in the feature values ​​to zero. Use 2 2 The maximum pooling operation reduces the dimension of the activated feature map to obtain the pooled feature map tensor, flattens the pooled feature map into a one-dimensional vector F, inputs it into the fully connected layer, and generates a multimodal emotion feature vector through linear transformation.

[0008] As a preferred solution of the comprehensive control method based on multi-modal electromagnetic and ultrasonic stimulation described in the present invention, wherein: the generation of a real-time emotional state quantization vector and identification of the current emotional state category by performing temporal dependency analysis on multi-modal features through a long short-term memory network refers to organizing the multi-modal emotional feature vector fused by the CNN model according to a time series to form a temporal feature matrix input to the LSTM model, and reformatting the temporal feature matrix into a three-dimensional input format to form a three-dimensional tensor input to the LSTM model; Initialize the LSTM unit and set the number of network layers and parameters. Input the three-dimensional input tensor into the LSTM model time-by-time to perform a time-series update of the hidden state and generate the hidden state sequence of the LSTM at each time step. The final hidden state of the LSTM network is used as input, and the hidden state is mapped into a quantized vector of emotional state through a fully connected layer; Set emotional state categories include anxiety, joy, calm, stress, and relaxation; The emotional state quantization vector is input into the Softmax classifier, the probability distribution of each emotional state category is calculated, and the emotional state category corresponding to the maximum probability is selected from the probability distribution of the Softmax classifier as the emotion recognition result at the current moment.

[0009] As a preferred solution of the comprehensive control method based on multi-mode electromagnetic and ultrasonic stimulation described in the present invention, wherein: the phase coupling characteristics and oscillation synchronization characteristics between brain regions under different emotional states in the EEG signal are extracted according to the results of emotional state recognition, and a real-time brain functional connection network is constructed, which means extracting the current emotional state category and recording the corresponding emotional index, using a label to identify the emotional state at each moment, identifying the time period when the current emotional state is continuously stable and marking it as the target emotional state time window, extracting the EEG signal segment within the emotional state stable period from the target emotional state time window, and synchronously extracting the EDA signal and HRV signal segments within the time period; According to the emotion state recognition results, n key brain region pairs related to the current emotion state are selected and marked as target brain region pairs; Use short-time Fourier transform to perform time-frequency decomposition on each time window, extract phase value sequences in different frequency bands for each target brain region pair, and calculate the PLV value in each frequency band based on the extracted phase sequence of the target brain region pair; The PLV value of the target brain region pair is calculated for each time domain window within the target EEG signal segment to form a PLV value sequence that changes over time; The target EEG signal is analyzed in the time and frequency domains using Hilbert transform to obtain the instantaneous amplitude and phase sequence of each target brain area in the target frequency band and calculate the oscillation synchronization value; The phase coupling values ​​and oscillation synchronization values ​​of all brain regions are matrixed to generate a functional connectivity matrix between brain regions. The element relationships in the functional connectivity matrix are used to construct a functional connectivity network of the brain under emotional states. The nodes of the network represent different brain regions, and the edges represent the intensity of neural interactions between nodes.

[0010] As a preferred solution of the comprehensive regulation method based on multi-modal electromagnetic and ultrasonic stimulation described in the present invention, wherein: the active emotion regulation nodes in the functional network are identified by graph analysis technology, the emotion regulation nodes are divided into surface targets and deep targets, and the emotional activity of each node is quantified, and the stimulation strategy is formulated according to the emotional activity, which means mapping the constructed brain functional connection network into a directed weighted graph, wherein the nodes represent emotion-related brain regions, the edges represent the functional connection strength between different brain regions, and the weight value of the edge is set by the phase coupling characteristics and oscillation synchronization characteristics between every two nodes; Calculate the degree centrality and betweenness centrality of each node and combine the degree centrality and betweenness centrality to get the activity score of each node; Set an activity threshold Q. If the node activity exceeds the threshold Q, the activity of the node is abnormal. According to the spatial location and depth of abnormally active nodes in each brain region in the EEG signal, the abnormally active nodes are divided into surface targets and deep targets; Nodes with abnormal activity were quantified as active emotion regulation nodes and the nodes were sorted in descending order according to the scores, and the nodes with the highest scores were prioritized as stimulation targets; If the abnormally active node is a surface target, an initial transcranial magnetic stimulation strategy is formulated. For areas with abnormally active gamma waves, the initial frequency of transcranial magnetic stimulation is set to 30-50 Hz to reduce the excitability of local neurons. The pulse intensity of transcranial magnetic stimulation is adjusted according to the node activity value. If the gamma wave power does not return to the normal range within 500ms, the stimulation time is automatically extended. After the emotional state of the surface target is restored, the EGG characteristics of the deep brain area are further analyzed. If the abnormally active node is a deep target, an initial transcranial focused ultrasound strategy is formulated. For areas with abnormally active deep theta waves, the frequency of transcranial focused ultrasound is set to 0.5-1.5 MHz, the initial sound pressure intensity and focal radius are set, and the theta wave recovery of the EEG signal in the deep brain area is monitored in real time. If the amygdala-hippocampus connection is enhanced, the amygdala is used as the main deep target, and the spatial position of the amygdala is located through MRI guidance, and the activity of the target area is determined through EEG signals. If the theta wave power spectral density gradually returns to normal levels, it means that the emotional state tends to be balanced, and the intensity and duration of transcranial focused ultrasound are gradually reduced; If the surface gamma waves recover after regulation but the deep theta waves are still abnormal, the intensity and frequency of the electromagnetic and ultrasonic waves are adjusted according to the real-time feedback of the EEG signal, and the brain areas that recover slowly are mainly intervened. Then the stimulation intensity of the surface targets is gradually adjusted. The overall emotional state is dynamically evaluated and the stimulation strategy is adaptively adjusted in each regulation. The same operation is performed if the deep theta waves recover but the surface gamma waves are still abnormal.

[0011] As a preferred solution of the comprehensive regulation method based on multimodal electromagnetic and ultrasonic stimulation described in the present invention, wherein: the recording of each emotion regulation data and generating a personalized emotion profile for each subject refers to real-time recording of EEG signals, multimodal physiological characteristics and stimulation parameters during each emotion regulation process, identifying the personalized emotion feature pattern of each subject to generate a personalized emotion profile.

[0012] Another object of the present invention is to provide a comprehensive control system based on multi-mode electromagnetic and ultrasonic stimulation, which comprises: The data acquisition module is used to collect the brain wave data and multimodal physiological data of the subject in real time and perform preprocessing and feature extraction on the collected data; The timing analysis module is used to perform multimodal feature fusion on the extracted signal features and generate a quantitative vector of the emotional state by performing timing dependency analysis of the emotional state through the CNN and LSTM models; The network construction module is used to extract EEG signal features under different emotional states based on the emotion state recognition results to build a real-time brain functional connection network and identify active emotion regulation nodes; The stimulation strategy module is used to formulate personalized electromagnetic and ultrasonic stimulation strategies according to the activity and location of the emotion regulation nodes, and to record the data of each regulation process in real time to generate a personalized emotion profile.

[0013] A computer device comprises: a memory and a processor; the memory stores a computer program, and the processor implements the steps of a comprehensive control method based on multi-mode electromagnetic and ultrasonic stimulation when executing the computer program.

[0014] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a comprehensive control method based on multi-mode electromagnetic and ultrasonic stimulation.

[0015] The beneficial effects of the present invention are as follows: the present invention constructs a real-time brain functional connection network through the synchronous acquisition and feature fusion of multi-electrode EEG, EDA and HRV signals, and accurately identifies and formulates stimulation strategies for emotion regulation nodes based on graph theory analysis, thereby improving the accuracy and reliability of emotion recognition, realizing personalized regulation of individual emotional states, improving individual cognitive functions and language expression abilities, and reducing motor coordination disorders caused by emotional fluctuations. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0017] Figure 1 A schematic diagram of the process of the integrated regulation method based on multi-modal electromagnetic and ultrasonic stimulation; Figure 2 Schematic diagram of the structure of the comprehensive control system based on multi-modal electromagnetic and ultrasonic stimulation. DETAILED DESCRIPTION

[0018] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0019] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0020] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0021] Example 1, reference Figure 1 , which is the first embodiment of the present invention, and which provides a comprehensive control method based on multi-mode electromagnetic and ultrasonic stimulation. The comprehensive control method based on multi-mode electromagnetic and ultrasonic stimulation includes: S1, collect brain wave signals in real time through a multi-electrode EEG system and synchronously collect skin electrical response and heart rate variability signals, and perform feature extraction after preprocessing the collected signals; Specifically, the real-time collection of brain wave signals and synchronous collection of skin electrical response and heart rate variability signals by a multi-electrode EEG system, and feature extraction after preprocessing the collected signals refer to using a multi-electrode EEG system, arranging electrodes in the prefrontal cortex, parietal lobe and hippocampus of the subject according to the 10-20 electrode system standard, recording the brain wave signals of each electrode, placing EDA sensors on the fingertips and palms of the subject to collect regional skin electrical activity, placing HRV sensors on the chest and wrists to collect real-time fluctuation data of heart rate, setting the sampling frequency of the EEG system, EDA and HRV sensors to 250 Hz, and synchronously recording the collected data; The ICA algorithm is used to remove artifacts from the EEG signal, and a bandpass filter is used to perform frequency domain filtering on the EEG signal after artifact removal, which is mainly used to eliminate noise caused by eye movements, muscle activities and other movements, remove low-frequency DC drift and high-frequency noise, and ensure that the main frequency bands in the signal are retained. After filtering, the EEG activity in the signal is further separated from the noise interference, and the signal quality is significantly improved. The power spectral density of the EEG signal is extracted by the Welch method and time-frequency analysis is performed by wavelet transform, and the sample entropy is calculated to measure the complexity of the EEG signal; A 3-second window was used to perform sliding window mean filtering on the EDA signal and HRV signal to remove instantaneous spike signals, extract the skin electrical response characteristics of EDA, and perform frequency domain analysis on the HRV signal through fast Fourier transform to extract low-frequency power, high-frequency power and LF / HF ratio.

[0022] Through the synchronous acquisition of multi-electrode EEG system, EDA sensor and HRV sensor, the EEG activity, skin galvanic response and heart rate variability signals of the subject during emotional fluctuations can be captured in real time. The acquisition of this multimodal signal can comprehensively reflect the emotional state of the subject from different dimensions such as nerves, skin galvanic response and heart rate. The introduction of multimodal signals not only makes up for the shortcomings of single modal signals in emotion recognition, but also improves the recognition accuracy through the complementarity between modalities. The ICA algorithm can effectively separate the independent components in EEG, distinguish the artifacts from the EEG activity signals, and thus achieve artifact removal. Bandpass filtering can further eliminate the low-frequency DC drift and high-frequency noise in the EEG signal, thereby improving the purity and stability of the signal. The Welch method uses overlapping windows to perform spectrum estimation, which can effectively improve the frequency resolution and reduce the variance in spectrum estimation. The sliding window mean filter can eliminate these instantaneous spikes and smooth the signal trend, thereby better characterizing the intensity and persistence of the skin galvanic response. After fast Fourier transform, the HRV signal can be decomposed into low-frequency power (LF) and high-frequency power (HF). The LF / HF ratio is an important indicator to measure the balance state of the sympathetic and parasympathetic nerves, and can effectively identify the autonomic nervous response patterns of individuals during emotional fluctuations (such as anxiety and relaxation).

[0023] S2, using CNN to perform multimodal feature fusion on the extracted signal features, and using long short-term memory network to perform temporal dependency analysis on the multimodal features to generate a real-time emotional state quantification vector and identify the current emotional state category; Specifically, the multimodal feature fusion of the extracted signal features through CNN refers to normalizing the feature data of EEG, EDA and HRV signals to eliminate the magnitude difference between different modal features, aligning the normalized modal features according to the time axis and splicing them to generate a splicing matrix X; The process of aligning the normalized modal features according to the time axis and splicing them to generate the splicing matrix X includes: the feature vector of the EEG in each time window is expressed as ,in The number of dimensions representing EEG features (such as power spectrum density dimension + sample entropy dimension + time-frequency domain feature dimension), the feature vector of the EDA signal is expressed as ,in is the number of dimensions of the EDA feature (peak response and sustained response), and the feature vector of the HRV signal is expressed as ,in is the number of dimensions of HRV features (low-frequency power (LF), high-frequency power (HF), and LF / HF ratio); The feature vectors of different modes are concatenated row by row to form a multimodal feature matrix. The matrix arranges the features of each modality along the time window so that all modal features are aligned in the time dimension. The specific construction method is as follows: , In the formula, is the EEG feature vector, is the EDA feature vector, is the HRV feature vector; Its shape is For example, assuming that the EEG feature has 20 dimensions, the EDA feature has 5 dimensions, and the HRV feature has 3 dimensions, the dimension of the multimodal feature matrix is ; Perform the same concatenation operation on all time windows, combine the feature matrices in each time window, and form a complete multimodal feature matrix; Expand the concatenated multimodal feature matrix into a three-dimensional input tensor of CNN and construct the input data format of multimodal feature fusion; With this input format, all modal features are aligned along the time axis, and different modal features are represented by the feature dimension M, ensuring that multimodal data can be recognized and processed by the CNN model; Constructing a CNN model includes input layer, convolution layer, pooling layer, fully connected layer and output layer; The three-dimensional tensor is convolved with the convolution kernel to obtain the multimodal feature map tensor [J T K], where K is the number of convolution kernels, T is the time step, and J is the number of types of multimodal features; Perform nonlinear activation on the convolutional feature map, use the ReLU activation function to return the negative values ​​in the feature values ​​to zero, enhance the sparsity of the feature and introduce nonlinearity; ReLU helps the convolutional features retain more nonlinear changes and enhances the performance of the network in complex emotional signals; Use 2 2 The maximum pooling operation reduces the dimension of the activated feature map to obtain the pooled feature map tensor, flattens the pooled feature map into a one-dimensional vector F, inputs it into the fully connected layer, and generates a multimodal emotion feature representation vector through linear transformation.

[0024] Through multimodal feature fusion, the complex associations between features of each modality can be effectively captured, the redundancy and deficiencies of single modal data can be avoided, and the changes in emotional states can be more comprehensively expressed. Traditional feature extraction methods (such as PCA or linear projection) can usually only capture the global patterns of low-dimensional feature space and cannot effectively extract local relationships in high-dimensional feature space. CNN can perform local sliding on the feature matrix through convolution operations, and extract local patterns and spatiotemporal dependencies of different modal features through the action of multiple convolution kernels, which can better capture the nonlinear relationship between cross-modal features, thereby improving the expression ability of the emotion recognition model. Traditional emotion recognition algorithms (such as static classification models based on SVM or KNN) are difficult to handle the temporal changes of emotions, resulting in a decrease in the accuracy and robustness of emotion recognition. Through LSTM, the temporal dependency analysis of multimodal emotion features can effectively model the dynamic evolution of emotional states, thereby solving the problem of insufficient temporal dependency modeling in emotional state recognition. Through the combination of ReLU and maximum pooling, the effectiveness of feature retention can be maximized in the process of nonlinear mapping and dimensionality reduction, thereby improving the performance of the model in complex emotional signals.

[0025] Furthermore, the method of performing temporal dependency analysis on multimodal features through a long short-term memory network to generate a real-time emotional state quantization vector and identify the current emotional state category refers to organizing the multimodal emotional feature vector after fusion by the CNN model according to a time series to form a temporal feature matrix input to the LSTM model, and reformatting the temporal feature matrix into a three-dimensional input format to form a three-dimensional tensor input to the LSTM model; Initialize the LSTM unit and set the number of network layers and parameters. Input the three-dimensional input tensor into the LSTM model time-by-time to perform a time-series update of the hidden state and generate the hidden state sequence of the LSTM at each time step. The final hidden state of the LSTM network is used as input, and the hidden state is mapped into a quantized vector of emotional state through a fully connected layer; Set emotional state categories include anxiety, joy, calm, stress, and relaxation; The emotional state quantization vector is input into the Softmax classifier, the probability distribution of each emotional state category is calculated, and the emotional state category corresponding to the maximum probability is selected from the probability distribution of the Softmax classifier as the emotion recognition result at the current moment.

[0026] The time-series feature matrix organizes multimodal emotional features in time series, allowing LSTM to use the dependencies of time series to learn the dynamic evolution of emotional states. In the change of emotional state, some signals may not show obvious changes immediately, but gradually accumulate effects over time. LSTM can remember the long-term dependencies of emotional features by updating the hidden state at each time step. Therefore, compared with the static model, the introduction of LSTM significantly improves the time sensitivity of emotional state recognition, especially for capturing dynamic emotional fluctuations. STM updates the hidden state at each time step, and the generated hidden state sequence reflects the accumulated information of emotional features over the entire time period. The final hidden state can be regarded as a comprehensive memory of the emotional state over a period of time, covering all information about the change of emotions over time. This method avoids the limitation of independent analysis of each moment in traditional methods and ensures the coherence of emotional features in the time dimension. Therefore, the hidden state sequence generated by LSTM can provide a more comprehensive representation of emotional states than the features of a single moment. By mapping the hidden state of LSTM to the emotional state quantization vector through the fully connected layer, complex time series features can be simplified into a numerical emotional quantization model. By calculating the probability distribution of different emotion categories, Softmax ensures that the model can give the most likely emotional state classification results in the form of probability. This probability-based classification method can provide the system with a more flexible and reliable emotion recognition mechanism, especially when the emotional states are ambiguous or overlapping, Softmax can effectively capture the subtle differences between various emotions, thereby improving the accuracy of classification.

[0027] S3. Extract the phase coupling characteristics and oscillation synchronization characteristics between brain regions under different emotional states in EEG signals based on the results of emotional state recognition, build a real-time brain functional connection network, identify active emotional regulation nodes in the functional network through graph analysis technology, divide emotional regulation nodes into surface targets and deep targets, quantify the emotional activity of each node, and formulate stimulation strategies based on emotional activity; Specifically, according to the results of emotional state recognition, the phase coupling characteristics and oscillation synchronization characteristics between brain regions under different emotional states in the EEG signal are extracted. The real-time brain functional connection network is constructed, which means extracting the current emotional state category and recording the corresponding emotional index, using labels to identify the emotional state at each moment (for example, 0-5 seconds is "anxiety state", 5-10 seconds is "pleasure state"), identifying the time period (≥3 seconds) when the current emotional state is continuously stable and marking it as the target emotional state time window, extracting the EEG signal fragments within the emotional state stability period from the target emotional state time window, and synchronously extracting the EDA and HRV data fragments within the time period to ensure the time consistency in subsequent analysis. Through the results of emotional state recognition, the signal fragments under the target emotional state are determined, and basic data is provided for the subsequent extraction of phase coupling characteristics and oscillation synchronization characteristics. According to the emotion state recognition results, select n key brain region pairs related to the current emotion state and mark them as target brain region pairs, (e.g., pleasure state: prefrontal lobe-parietal lobe (cognition and emotion), prefrontal lobe-cingulate gyrus (emotional reward and pleasure), anxiety state: prefrontal lobe-amygdala (emotional inhibition and regulation), hippocampus-amygdala (emotional memory and emotional response)); Use short-time Fourier transform to perform time-frequency decomposition on each time window, extract the phase value sequence in different frequency bands for each target brain region pair, and calculate the PLV value in each frequency band based on the extracted phase sequence of the target brain region pair: , Where PLV is the phase locking value, which quantifies the phase synchronization degree of the target brain region in a certain frequency band, and the value range is 0 to 1, where 1 indicates that the phases of the two brain regions in this frequency band are highly synchronized, and 0 indicates that they are not synchronized. N is the total number of time points, which indicates the number of time points for phase synchronization calculation within a given time window. and They represent the phase values ​​of the target brain region pair (such as brain region 1 and brain region 2) at time t respectively; The PLV value of the target brain region pair is calculated for each time domain window within the target EEG segment to form a PLV value sequence that changes over time; Use Hilbert transform to analyze the target EEG signal in the time domain and frequency domain, obtain the instantaneous amplitude and phase sequence of each target brain area in the target frequency band, and calculate the oscillation synchronization value S: , Where N is the total number of time points, and are the instantaneous amplitudes of brain area 1 and brain area 2 at time t, and are the instantaneous phases of brain area 1 and brain area 2 at time t, respectively; The phase coupling values ​​and oscillation synchronization values ​​of all brain regions are matrixed to generate a functional connectivity matrix between brain regions. The element relationships in the functional connectivity matrix are used to construct a functional connectivity network of the brain under emotional states. The nodes of the network represent different brain regions (such as the prefrontal cortex, hippocampus, and amygdala), and the edges represent the intensity of neural interactions between nodes (such as the enhanced connection between the amygdala and the prefrontal cortex represents an emotional stress state).

[0028] By identifying the emotional state category and recording the corresponding emotional index, the signals of different emotional states in the time domain are marked as independent time windows (for example, 0-5 seconds for anxiety state and 5-10 seconds for pleasure state). On this basis, the stable duration period of emotional state (≥3 seconds) is further identified as the target emotional state time window. This precise time window calibration method can significantly improve the accuracy of emotional state analysis, because emotional states often fluctuate in a short time, and identifying stable emotional state time periods can eliminate the interference of short-term artifacts on emotion recognition. In addition, synchronous extraction of EDA and HRV data within the target time window can ensure the temporal consistency of different modal data and provide basic data for subsequent multimodal feature fusion. Specific brain region pairs are selected based on the emotional state recognition results, and the phase coupling features between these brain regions (such as prefrontal lobe-amygdala, prefrontal lobe-parietal lobe) are extracted in the corresponding frequency band. The phase coupling features can reveal the coordinated activity patterns between different brain regions under emotional states. For example, in the anxious state, the phase coupling between the prefrontal lobe and the amygdala is often enhanced, indicating the strengthening of emotional inhibition and regulation functions. Therefore, the selection of target brain region pairs and the extraction of phase coupling features can more accurately identify the functional connection network under emotional state and effectively distinguish the neural activity patterns of different emotional states. The EEG signals of the target brain regions were analyzed in the time domain and frequency domain using Hilbert transform to extract the instantaneous amplitude and phase sequence. By calculating the oscillation synchronization value, the temporal synchronization pattern of neural activity between brain regions under different emotional states can be revealed. This method can capture the dynamic characteristics of emotional state in neural activity, rather than being limited to the analysis of average features. For example, the oscillation synchronization between the prefrontal cortex and the cingulate gyrus is higher in the state of pleasure, indicating the enhancement of positive emotion regulation function. By matrixing the phase coupling values ​​and oscillation synchronization values ​​of all target brain region pairs, a functional connection matrix between brain regions can be generated. Based on this matrix, the functional connection network of the brain is constructed using graph theory analysis technology to identify the active emotion regulation nodes under emotional state. The functional connection network can reflect the overall neural interaction pattern of the brain under emotional state and identify the surface targets (such as the prefrontal cortex) and deep targets (such as the amygdala and hippocampus) of emotion regulation. This networked node identification method can effectively improve the accuracy and effectiveness of emotion regulation strategies and enhance individuals' cognition of the outside world.

[0029] Furthermore, the method of identifying active emotion regulation nodes in the functional network by using graph analysis technology, dividing the emotion regulation nodes into surface targets and deep targets and quantifying the emotional activity of each node, and formulating stimulation strategies according to the emotional activity refers to mapping the constructed brain functional connection network into a directed weighted graph, wherein the nodes represent emotion-related brain regions, the edges represent the functional connection strength between different brain regions, and the weight values ​​of the edges are set by the phase coupling characteristics and oscillation synchronization characteristics between every two nodes; Calculate the degree centrality and betweenness centrality of each node and combine the degree centrality and betweenness centrality to get the activity score of each node; Set the activity threshold Q, Where W is the baseline activity, which refers to the average node activity under normal emotional state, obtained through long-term monitoring. is the standard deviation of node activity, k is the control coefficient (usually 2 or 3); if the node activity exceeds the threshold Q, the activity of the node is abnormal; According to the spatial location and depth of abnormally active nodes in each brain region in the EEG signal, the abnormally active nodes are divided into surface targets and deep targets; The surface targets are mainly concentrated in the cerebral cortex (such as the prefrontal lobe and parietal lobe) and are significantly correlated with surface EEG features (such as gamma wave activity); The deep targets are concentrated in subcortical areas (such as amygdala and hippocampus) and are significantly correlated with deep EEG features (such as abnormally active theta waves); Nodes with abnormal activity were quantified as active emotion regulation nodes and the nodes were sorted in descending order according to the scores, and the nodes with the highest scores were prioritized as stimulation targets; If the abnormally active node is a surface target, an initial transcranial magnetic stimulation strategy is formulated. For areas with abnormally active gamma waves, the initial frequency of transcranial magnetic stimulation is set to 30-50 Hz to reduce the excitability of local neurons. The pulse intensity of transcranial magnetic stimulation is adjusted according to the node activity value (initial intensity range 1-2T). If the gamma wave power does not return to the normal range within 500ms, the stimulation time is automatically extended. After the emotional state of the surface target is restored, the EGG characteristics of the deep brain area are further analyzed. If the abnormally active node is a deep target, an initial transcranial focused ultrasound strategy is formulated. For areas with abnormally active deep theta waves, the frequency of transcranial focused ultrasound is set to 0.5-1.5 MHz, the initial sound pressure intensity and focal radius are set, and the theta wave recovery of the EEG signal in the deep brain area is monitored in real time. If the amygdala-hippocampus connection is enhanced, the amygdala is used as the main deep target, and the spatial position of the amygdala is located through MRI guidance, and the activity of the target area is determined through EEG signals, so that the focused ultrasound can accurately act on the deep emotional nodes. If the theta wave power spectral density gradually returns to normal levels, it means that the emotional state is tending to be balanced, and the intensity and duration of transcranial focused ultrasound are gradually reduced; If the surface gamma waves recover after regulation but the deep theta waves are still abnormal, the intensity and frequency of the electromagnetic and ultrasonic waves are adjusted according to the real-time feedback of the EEG signal, and the brain areas that recover slowly are mainly intervened. Then the stimulation intensity of the surface targets is gradually adjusted. In each regulation, the overall emotional state is dynamically evaluated and the stimulation strategy is adaptively adjusted to prevent whole-brain emotional imbalance caused by excessive regulation of a single area. The same operation is performed if the deep theta waves recover but the surface gamma waves are still abnormal.

[0030] By mapping the functional connections of different emotion-related brain regions into directed weighted graphs, graph analysis techniques (such as degree centrality and betweenness centrality) are used to identify the key nodes of emotion regulation, and these nodes are divided into surface targets and deep targets. This concept can not only fully reflect the interactive relationship between different brain regions under emotional states, but also quantify the importance of nodes in the network, so as to more accurately determine the regulatory targets. Compared with traditional static feature analysis, the emotion regulation method based on network topology structure of the present invention can more effectively capture the dynamic interaction pattern between brain regions under emotional states. By calculating the degree centrality and betweenness centrality of each node, the two are combined to obtain the activity score of the node, and different emotion regulation strategies are formulated according to the score. The activity score can reflect the abnormal activity of the node under a specific emotional state, so as to effectively identify the key regulatory nodes during emotional fluctuations. Nodes whose activity exceeds the threshold are regarded as abnormal nodes, and they are divided into surface targets and deep targets according to the depth of the brain region where they are located, providing an accurate hierarchical basis for the subsequent formulation of stimulation strategies. According to the spatial location and depth of abnormally active nodes, they are divided into surface targets (such as the prefrontal lobe and parietal lobe) and deep targets (such as the amygdala and hippocampus), and transcranial magnetic stimulation (TMS) and transcranial focused ultrasound (TUS) are used for hierarchical regulation. Surface targets are mainly concentrated in the cortical area, and transcranial magnetic stimulation can effectively reduce the excitability of local neurons. Deep targets (such as the amygdala) are closely related to deep emotional regulation, and focused ultrasound can achieve precise regulation of deep brain areas. Through real-time feedback of EEG signals, the stimulation strategy is dynamically adjusted to ensure that the stimulation parameters (such as frequency, intensity, and duration) can be optimized in time when the emotional state has not fully recovered. If the emotional state of a certain target has not returned to normal levels, the stimulation time will be automatically extended or the stimulation intensity will be adjusted. When the surface targets and deep targets are linked during the regulation process, the real-time adjustment strategy can prevent the whole-brain emotional imbalance caused by excessive regulation of a single target. By formulating different stimulation strategies, it is possible to respond to the emotional out-of-control of the subjects in different situations, further improving the subjects' cognition of the outside world, effectively reducing the cognitive load of patients in anxiety and depression, and enhancing concentration, thereby improving their decision-making and problem-solving abilities in complex situations, reducing motor coordination disorders caused by emotional fluctuations, significantly improving patients' language expression ability, and reducing language pauses and repetitions.

[0031] S4, record each emotion regulation data and generate a personalized emotion profile for each subject; Specifically, recording each emotion regulation data and generating a personalized emotion profile for each subject means recording the EEG signals, multimodal physiological characteristics and stimulation parameters in real time during each emotion regulation process to identify the personalized emotion characteristic pattern of each subject and generate a personalized emotion profile.

[0032] The real-time data recording method can completely preserve the neural responses and physiological fluctuations of the subjects under different emotional states, avoiding the incompleteness of data fragmentation or later reconstruction in traditional emotional intervention technology. The emotional regulation data recorded in real time can provide high-precision data support for emotional state recognition, personalized feature extraction and dynamic adjustment of intervention strategies. Through the accumulation and analysis of multiple emotional regulation data, the present invention can identify the personalized emotional characteristic patterns of the subjects. The emotional response patterns of each individual may show significant differences under similar stimulation conditions, such as different individuals produce different neural responses to the same electromagnetic or ultrasonic stimulation intensity. By establishing personalized emotional characteristic patterns, the response characteristics of individuals to specific emotional stimuli can be quantified, thereby improving the accuracy of emotional intervention strategies. The present invention generates a personalized emotional profile for each subject through the accumulation and analysis of long-term emotional regulation data. The profile not only contains the characteristic patterns of emotional states under different stimulation conditions, but also records the long-term evolution trend of emotional states.

[0033] Example 2, reference Figure 2 , which is the second embodiment of the present invention, is different from the previous embodiment and provides a comprehensive control system based on multi-mode electromagnetic and ultrasonic stimulation, which includes: The data acquisition module is used to collect the brain wave data and multimodal physiological data of the subject in real time and perform preprocessing and feature extraction on the collected data; The timing analysis module is used to perform multimodal feature fusion on the extracted signal features and generate a quantitative vector of the emotional state by performing timing dependency analysis of the emotional state through the CNN and LSTM models; The network construction module is used to extract EEG signal features under different emotional states based on the emotion state recognition results to build a real-time brain functional connection network and identify active emotion regulation nodes; The stimulation strategy module is used to formulate personalized electromagnetic and ultrasonic stimulation strategies according to the activity and location of the emotion regulation nodes, and to record the data of each regulation process in real time to generate a personalized emotion profile.

[0034] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0035] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0036] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0037] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit with a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit with a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

Claims

1. A comprehensive control method based on multi-mode electromagnetic and ultrasonic stimulation, characterized in that: include, The multi-electrode EEG system collects brain wave signals in real time and synchronously collects skin electrical response and heart rate variability signals, and extracts features after preprocessing the collected signals; The extracted signal features are fused into multimodal features through CNN, and the multimodal features are analyzed for temporal dependency through long short-term memory network to generate real-time emotional state quantification vector and identify the current emotional state category; Based on the results of emotional state recognition, the phase coupling characteristics and oscillation synchronization characteristics between brain regions under different emotional states in EEG signals are extracted, and a real-time brain functional connection network is constructed. Active emotion regulation nodes in the functional network are identified through graph analysis technology, and the emotion regulation nodes are divided into surface targets and deep targets, and the emotional activity of each node is quantified, and stimulation strategies are formulated according to the emotional activity. Record each emotion regulation data and generate a personalized emotion profile for each subject.

2. The comprehensive control method based on multi-mode electromagnetic and ultrasonic stimulation according to claim 1, characterized in that: The method of collecting brain wave signals in real time and synchronously collecting skin electrical response and heart rate variability signals through a multi-electrode EEG system, and extracting features after preprocessing the collected signals refers to using a multi-electrode EEG system, arranging electrodes in the prefrontal cortex, parietal lobe and hippocampus of the subject according to the 10-20 electrode system standard, recording the EEG signal of each electrode, placing EDA sensors on the fingertips and palms of the subject to collect EDA signals, placing HRV sensors on the chest and wrists to collect HRV signals, setting the sampling frequency of the EEG system, EDA and HRV sensors to 250 Hz, and synchronously recording and collecting data; The artifacts of EEG signals were removed by ICA algorithm, and the frequency domain filtering of the EEG signals after artifact removal was performed by bandpass filter. The power spectrum density of EEG signals was extracted by Welch method and time-frequency analysis was performed by wavelet transform. The sample entropy was calculated to measure the complexity of EEG signals. A 3-second window was used to perform sliding window mean filtering on the EDA signal and the HRV signal to remove instantaneous spike signals, extract the skin electrical response characteristics of the EDA signal, and perform frequency domain analysis on the HRV signal through fast Fourier transform to extract low-frequency power, high-frequency power, and LF / HF ratio.

3. The comprehensive control method based on multi-mode electromagnetic and ultrasonic stimulation according to claim 2, characterized in that: The multi-modal feature fusion of the extracted signal features by CNN refers to normalizing the feature data of EEG, EDA and HRV signals, aligning the normalized features of each modality according to the time axis and splicing them to generate a splicing matrix X; Expand the concatenated multimodal feature matrix into a three-dimensional input tensor of CNN and construct the input data format of multimodal feature fusion; Constructing a CNN model includes input layer, convolution layer, pooling layer, fully connected layer and output layer; The three-dimensional tensor is convolved with the convolution kernel to obtain the multimodal feature map tensor [J T K], where K is the number of convolution kernels, T is the time step, and J is the number of types of multimodal features; Perform nonlinear activation on the convolutional feature map and use the ReLU activation function to return the negative values ​​in the feature values ​​to zero. Use 2 2 The maximum pooling operation reduces the dimension of the activated feature map to obtain the pooled feature map tensor, flattens the pooled feature map into a one-dimensional vector F, inputs it into the fully connected layer, and generates a multimodal emotion feature vector through linear transformation.

4. The comprehensive control method based on multi-mode electromagnetic and ultrasonic stimulation according to claim 3, characterized in that: The method of performing temporal dependency analysis on multimodal features through a long short-term memory network to generate a real-time emotional state quantization vector and identify the current emotional state category refers to organizing the multimodal emotional feature vector after fusion by the CNN model according to a time series to form a temporal feature matrix input to the LSTM model, and reformatting the temporal feature matrix into a three-dimensional input format to form a three-dimensional tensor input to the LSTM model; Initialize the LSTM unit and set the number of network layers and parameters. Input the three-dimensional input tensor into the LSTM model time-by-time to perform a time-series update of the hidden state and generate the hidden state sequence of the LSTM at each time step. The final hidden state of the LSTM network is used as input, and the hidden state is mapped into a quantized vector of emotional state through a fully connected layer; Set emotional state categories include anxiety, joy, calm, stress, and relaxation; The emotional state quantization vector is input into the Softmax classifier, the probability distribution of each emotional state category is calculated, and the emotional state category corresponding to the maximum probability is selected from the probability distribution of the Softmax classifier as the emotion recognition result at the current moment.

5. The comprehensive control method based on multi-mode electromagnetic and ultrasonic stimulation according to claim 4, characterized in that: The extracting of phase coupling features and oscillation synchronization features between brain regions under different emotional states in the EEG signal according to the result of emotional state recognition, and constructing a real-time brain functional connection network means extracting the current emotional state category and recording the corresponding emotional index, using a label to identify the emotional state at each moment, identifying the time period when the current emotional state is continuously stable and marking it as the target emotional state time window, extracting the EEG signal segment within the emotional state stable period from the target emotional state time window, and synchronously extracting the EDA signal and HRV signal segments within the time period; According to the emotion state recognition results, n key brain region pairs related to the current emotion state are selected and marked as target brain region pairs; Use short-time Fourier transform to perform time-frequency decomposition on each time window, extract phase value sequences in different frequency bands for each target brain region pair, and calculate the PLV value in each frequency band based on the extracted phase sequence of the target brain region pair; The PLV value of the target brain region pair is calculated for each time domain window within the target EEG signal segment to form a PLV value sequence that changes over time; The target EEG signal is analyzed in the time and frequency domains using Hilbert transform to obtain the instantaneous amplitude and phase sequence of each target brain area in the target frequency band and calculate the oscillation synchronization value; The phase coupling values ​​and oscillation synchronization values ​​of all brain regions are matrixed to generate a functional connectivity matrix between brain regions. The element relationships in the functional connectivity matrix are used to construct a functional connectivity network of the brain under emotional states. The nodes of the network represent different brain regions, and the edges represent the intensity of neural interactions between nodes.

6. The comprehensive control method based on multi-mode electromagnetic and ultrasonic stimulation according to claim 5, characterized in that: The method of identifying active emotion regulation nodes in the functional network by using graph theory analysis technology, dividing the emotion regulation nodes into surface targets and deep targets and quantifying the emotional activity of each node, and formulating stimulation strategies according to the emotional activity refers to mapping the constructed brain functional connection network into a directed weighted graph, wherein the nodes represent emotion-related brain regions, the edges represent the functional connection strength between different brain regions, and the weight value of the edges is set by the phase coupling characteristics and oscillation synchronization characteristics between every two nodes; Calculate the degree centrality and betweenness centrality of each node and combine the degree centrality and betweenness centrality to get the activity score of each node; Set an activity threshold Q. If the node activity exceeds the threshold Q, the activity of the node is abnormal. According to the spatial location and depth of abnormally active nodes in each brain region in the EEG signal, the abnormally active nodes are divided into surface targets and deep targets; Nodes with abnormal activity were quantified as active emotion regulation nodes and the nodes were sorted in descending order according to the scores, and the nodes with the highest scores were prioritized as stimulation targets; If the abnormally active node is a surface target, an initial transcranial magnetic stimulation strategy is formulated. For areas with abnormally active gamma waves, the initial frequency of transcranial magnetic stimulation is set to 30-50 Hz to reduce the excitability of local neurons. The pulse intensity of transcranial magnetic stimulation is adjusted according to the node activity value. If the gamma wave power does not return to the normal range within 500ms, the stimulation time is automatically extended. After the emotional state of the surface target is restored, the EGG characteristics of the deep brain area are further analyzed. If the abnormally active node is a deep target, an initial transcranial focused ultrasound strategy is formulated. For areas with abnormally active deep theta waves, the frequency of transcranial focused ultrasound is set to 0.5-1.5 MHz, the initial sound pressure intensity and focal radius are set, and the theta wave recovery of the EEG signal in the deep brain area is monitored in real time. If the amygdala-hippocampus connection is enhanced, the amygdala is used as the main deep target, and the spatial position of the amygdala is located through MRI guidance, and the activity of the target area is determined through EEG signals. If the theta wave power spectral density gradually returns to normal levels, it means that the emotional state tends to be balanced, and the intensity and duration of transcranial focused ultrasound are gradually reduced; If the surface gamma waves recover after regulation but the deep theta waves are still abnormal, the intensity and frequency of the electromagnetic and ultrasonic waves are adjusted according to the real-time feedback of the EEG signal, and the brain areas that recover slowly are mainly intervened. Then the stimulation intensity of the surface targets is gradually adjusted. The overall emotional state is dynamically evaluated and the stimulation strategy is adaptively adjusted in each regulation. The same operation is performed if the deep theta waves recover but the surface gamma waves are still abnormal.

7. The comprehensive control method based on multi-mode electromagnetic and ultrasonic stimulation according to claim 6, characterized in that: The recording of each emotion regulation data and generating a personalized emotion profile for each subject refers to real-time recording of EEG signals, multimodal physiological characteristics and stimulation parameters during each emotion regulation process, identifying the personalized emotion feature pattern of each subject and generating a personalized emotion profile.

8. A comprehensive control system based on multi-mode electromagnetic and ultrasonic stimulation based on the comprehensive control method based on multi-mode electromagnetic and ultrasonic stimulation according to any one of claims 1 to 7, characterized in that: include, The data acquisition module is used to collect the brain wave data and multimodal physiological data of the subject in real time and perform preprocessing and feature extraction on the collected data; The timing analysis module is used to perform multimodal feature fusion on the extracted signal features and generate a quantitative vector of the emotional state by performing timing dependency analysis of the emotional state through the CNN and LSTM models; The network construction module is used to extract EEG signal features under different emotional states based on the emotion state recognition results to build a real-time brain functional connection network and identify active emotion regulation nodes; The stimulation strategy module is used to formulate personalized electromagnetic and ultrasonic stimulation strategies according to the activity and location of the emotion regulation nodes, and to record the data of each regulation process in real time to generate a personalized emotion profile.

9. A computer device comprising: A memory and a processor; the memory stores a computer program, characterized in that: when the processor executes the computer program, the steps of the comprehensive control method based on multi-mode electromagnetic and ultrasonic stimulation described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the comprehensive control method based on multi-mode electromagnetic and ultrasonic stimulation described in any one of claims 1 to 7 are implemented.

Images