EEG-driven behavior recognition method based on dynamic evolution of coherent network features

Abstract

The present invention belongs to the field of signal processing and analysis technology, and specifically relates to a method for identifying EEG-driven behaviors based on the dynamic evolution of coherent network features. The method comprises the following steps: Step 1: Standard preprocessing is performed on the original EEG signal, the beta wave frequency band is selected as the analysis frequency band, and the coherence value within the analysis frequency band is calculated to measure the synchronization strength of the two EEG signals. This value is used as the network connection weight to construct a weighted coherent network; Step 2: At each time point, the degree and clustering coefficient of each node in the weighted coherent network are extracted. For each pair of nodes, the instantaneous phase difference is calculated, and the spatial geometric distance between the two electrodes is introduced to obtain the behavioral characteristic response index. Step 3: The power spectrum parameters of the target behavior signals in all brain regions within the analysis frequency band are obtained, and the target behavior recognition value is compared with the behavior threshold to determine whether the target behavior is activated. The present invention significantly improves the accuracy, real-time nature, and physiological interpretability of behavior recognition.

Description

Technical Field

[0001] The present invention belongs to the technical field of signal analysis and processing, and specifically relates to a method for recognizing EEG-driven behaviors based on the dynamic evolution of coherent network features. Background Art

[0002] EEG signals, as electrophysiological manifestations of neural activity, are widely used in fields such as brain-computer interfaces, neural engineering, clinical auxiliary diagnosis, and motor function monitoring. With the increasing demand for applications such as neural regulation, neurofeedback, motor intention recognition, and rehabilitation control, the precise extraction of behaviorally relevant neural activity patterns from EEG signals and the development of high-precision recognition methods for specific motor or cognitive behaviors have become a research focus in recent years.

[0003] Traditional EEG-driven behavior recognition methods are typically based on the following technical approaches: First, ERD / ERS (Electroresonance Resonance Response) analyzes power changes in specific frequency bands (such as α, β, and μ waves) before and after the target behavior is performed. For example, by comparing the power suppression of the β band before and after movement, it can identify movement behaviors or locate regional activation. However, these methods typically require a large number of trials and averages for stable recognition, have poor interference tolerance, and can only reflect the localized response of a single brain region, making it difficult to characterize the dynamic coordination between multiple brain regions. Second, independent component analysis (ICA) and source separation methods attempt to isolate independent neural source signals associated with specific behaviors from mixed EEG signals and perform feature extraction and classification based on these source signals. While these methods address signal aliasing to some extent, they are computationally expensive, lack robustness, and lack behavioral semantic interpretation of the components they actually isolate, making their results less generalizable.

[0004] The third is coherence analysis and functional connectivity map modeling. In recent years, EEG signals have received increasing attention as an expression carrier of functional connectivity networks. Related studies have constructed functional connectivity networks based on the coherence value, phase locking value, mutual information, etc. of the signals between two channels, and then combined them with graph theory indicators such as degree, clustering coefficient, efficiency, etc. for identification analysis. For example, some studies have constructed whole-brain coherence networks in the β band and used node centrality or motif structure analysis for classification. However, such methods have significant limitations: (1) Most networks are static networks, and the connectivity values ​​are only averaged over the entire behavioral cycle, ignoring the real-time evolution characteristics of the network under behavioral induction; (2) The connectivity value only considers the strength, such as the coherence value, but ignores the dynamic changes of the connectivity structure over time, such as the connectivity growth rate, the trend of local structural changes, etc.; (3) The connectivity index is not fused with the frequency domain power parameter, and there is a problem of decoupling the frequency domain information from the network structure; (4) The behavioral recognition value is usually not uniformly modeled, and there is a lack of comprehensive recognition function. Summary of the Invention

[0005] In view of this, the main purpose of the present invention is to provide an EEG-driven behavior recognition method based on the dynamic evolution of coherent network features. This method can simultaneously capture the dynamic reconstruction process of the neural network structure and the behavior-induced changes of frequency domain characteristics, overcoming the problems of insufficient dynamics, structural spectrum fragmentation, and weak feature fusion of traditional methods.

[0006] The technical solution adopted in the present invention is as follows:

[0007] A method for recognizing EEG-driven behaviors based on the dynamic evolution of coherent network features, the method comprising:

[0008] Step 1: Standard preprocessing is performed on the raw EEG signals. The beta wave band is selected as the analysis band, and the instantaneous frequency components of the analysis band are extracted using short-time wavelet transform. For each pair of electrodes, the coherence value in the analysis band is calculated within a specified time window to measure the synchronization strength of the two EEG signals. This value is used as the network connection weight to construct a weighted coherence network.

[0009] Step 2: At each time point, extract the degree and clustering coefficient of each node in the weighted coherence network and calculate the dynamic evolution index; combine the network connection weights with the dynamic evolution index to construct a unified EEG-driven behavioral feature map; at each time point, for each pair of nodes, calculate their instantaneous phase difference, and simultaneously introduce the spatial geometric distance between the two electrodes to obtain the behavioral feature response index;

[0010] Step 3: Obtain the power spectrum parameters of all brain region signals of the target behavior in the analysis frequency band, combine them with the characteristic response index, calculate the target behavior recognition value, set the behavior threshold of the target behavior in the analysis frequency band, compare the target behavior recognition value with the behavior threshold, and determine whether the target behavior is activated.

[0011] Furthermore, after selecting the beta wave frequency band as the analysis frequency band, the original EEG signal is processed using a bandpass filter, and the frequency range of the bandpass filter is 13Hz to 30Hz; the target behavior movement-related behavior includes at least: eye movement behavior, manual movement behavior and foot movement behavior.

[0012] Furthermore, the power spectrum parameters include: an average value of the power spectrum in a resting state, a standard deviation of the power spectrum in a resting state, an entropy value of the power spectrum, and an average value of the entropy value of the power spectrum.

[0013] Furthermore, the network connection weight is:

[0014] ;

[0015] in, Indicates electrode EEG signals and electrodes The EEG signal at time The network connection weight at time ; is the time window; is the time-integrated variable; Electrodes in the beta wave band EEG signals and electrodes The EEG signal at time The coherence value at time ; Electrodes in the beta wave band The spectral entropy of the EEG signal; Electrodes in the beta wave band The spectral entropy of the EEG signal; For electrodes The voltage amplitude of the EEG signal; For electrodes The voltage amplitude of the EEG signal; For time Electrodes in the beta wave band The center frequency of the EEG signal in the beta wave band The power spectral density at ; For time Electrodes in the beta wave band The center frequency of the EEG signal in the beta wave band The power spectral density at ; For electrodes The phase of the EEG signal; For electrodes The phase of the EEG signal; It is an indicator function, which takes a value of 1 when the condition is met, otherwise it takes a value of 0; For electrodes The signal-to-noise ratio of the EEG signal; is the set signal-to-noise ratio threshold; For electrodes The signal-to-noise ratio of the EEG signal.

[0016] Further, time Dynamic evolution index for:

[0017] ;

[0018] in, For time Time The clustering coefficient of the nodes corresponding to the electrodes; For time Time The degree of the node corresponding to each electrode.

[0019] Further, time The characteristic response index is :

[0020] ;

[0021] in, For time The electrode The phase and time of the EEG signal The electrode The phase difference of the EEG signal; For electrodes With electrodes The spatial Euclidean distance of is the spatial attenuation scale coefficient, when When the unit is centimeters, the value is 2 to 5; is the total number of electrodes.

[0022] Further, time Time The clustering coefficient of the nodes corresponding to the electrodes for:

[0023] ;

[0024] in, For the The actual number of edges between the neighbors of the node corresponding to the electrode.

[0025] Further, time Target behavior recognition value for:

[0026] ;

[0027] in, For time Time Electrode exist The integrated value of the power spectrum density under the wave frequency band, , is the frequency integration variable; is the average power spectrum of electrode 𝑖 in the resting state; For electrodes Standard deviation of the power spectrum in the resting state; For time Electrodes in the beta wave band The spectral entropy of the EEG signal; For time Time Electrode The entropy value of the power spectrum; is the average entropy value of the power spectrum of all electrodes in the resting state.

[0028] Furthermore, when the target behavior is eye movement behavior, the recognition threshold The value range is 0.20 to 0.30; when the target behavior is manual behavior, the recognition threshold The value range is 0.35 to 0.45; when the target behavior is foot behavior, the recognition threshold The value range of is 0.50 to 0.60; if Greater than the recognition threshold , then the target behavior is judged to be in the activated state.

[0029] By adopting the above technical solution, the present invention has the following beneficial effects: significantly improving the recognition accuracy and dynamic discrimination ability of EEG signals in complex motor behavior states. Traditional EEG recognition methods generally have technical bottlenecks such as being able to only capture local signal features, lacking the ability to model multi-region collaboratively, and being insensitive to temporal behavioral changes. To address these problems, the present invention proposes to extract multi-dimensional signal features such as synchronization between neural activities, connection strength, phase relationship, and spectral reconstruction from time-continuous, multi-channel EEG signals, and construct a dynamic evolution network based on these features to depict the topological evolution process of the EEG network structure as the behavioral state changes in real time. This network model not only takes into account the coherent connection weights between brain regions, but also introduces the connection growth rate, local clustering changes, spatial structure relationships, and network fluctuation trends, thereby achieving high-precision modeling of the brain region collaborative response mechanism. On this basis, the present invention further introduces spectral power changes, spectral entropy reconstruction, and baseline statistical features into the recognition index function to construct a unified recognition value expression form, so that the recognition model not only has a high dynamic response capability, but also has cross-subject adaptability and behavioral specificity. In addition, the present invention introduces a signal-to-noise ratio constraint mechanism in terms of neural signal quality control, which effectively suppresses the interference of low-quality signals on the recognition results, thereby improving the stability and reliability of the overall recognition system. Compared with traditional recognition methods based on static thresholds or single frequency features, the present invention can achieve continuous discrimination of multiple types of behavioral states such as eye movements, hand movements, and foot movements within a very short time window. It is particularly suitable for application scenarios such as real-time brain-computer interface control, motor intention recognition, and neural rehabilitation feedback that have high requirements for response speed and recognition accuracy. By integrating structural features and frequency domain indicators in the same modeling framework, the present invention achieves complexity modeling and dynamic analysis of neural behavioral signals while maintaining computational controllability, and has a high degree of engineering adaptability and promotion value. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 A schematic diagram of a method flow for recognizing EEG-driven behaviors based on the dynamic evolution of coherent network features provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0031] All features disclosed in this specification, or all steps in the disclosed methods or processes, except mutually exclusive features and / or steps, can be combined in any manner.

[0032] Any feature disclosed in this specification (including any appended claims and abstract), unless otherwise stated, may be replaced by other equivalent or similar features. In other words, unless otherwise stated, each feature is only an example of a series of equivalent or similar features.

[0033] refer to Figure 1 : A method for recognizing EEG-driven behaviors based on the dynamic evolution of coherent network features, the method comprising:

[0034] Step 1: Standard preprocessing is performed on the raw EEG signals. The beta wave band is selected as the analysis band, and the instantaneous frequency components of the analysis band are extracted using short-time wavelet transform. For each pair of electrodes, the coherence value in the analysis band is calculated within a specified time window to measure the synchronization strength of the two EEG signals. This value is used as the network connection weight to construct a weighted coherence network.

[0035] As a high-noise, low-amplitude, non-stationary bioelectric signal, EEG signals are often mixed with various interference components in their original form, including eye movement artifacts, myoelectric noise, and power frequency interference. Therefore, a multi-channel synchronous cleaning strategy is adopted in the preprocessing stage. A high-pass filter is used to remove DC drift, a low-pass filter is used to suppress high-frequency myoelectric interference, and a 50Hz or 60Hz notch processing is implemented to remove power frequency noise. On this basis, the inter-channel reference settings are standardized according to the experimental equipment configuration to ensure that all channels have a unified reference benchmark. In addition, it is necessary to eliminate possible outliers, use statistical distribution analysis methods to identify outlier data, and use interpolation or noise reduction reconstruction technology to repair missing or abnormal signal segments to ensure signal continuity and physical consistency.

[0036] After initial cleaning, this method specifically selects the beta band as the core region for analysis. This band, which generally corresponds to a frequency range of 13 Hz to 30 Hz, is widely considered to be closely related to brain activity in areas such as motor preparation, attention regulation, and behavioral activation. To accurately extract the dynamic frequency characteristics of the beta band, the short-time wavelet transform (SWT) technique is introduced. This technique possesses excellent time-frequency localization capabilities, achieving an optimal balance between temporal accuracy and frequency resolution, making it particularly suitable for processing non-stationary EEG signals. During the SWT process, the original signal is decomposed into multiple localized oscillation components of varying scales, effectively extracting information such as the instantaneous frequency, amplitude, and phase within the target frequency band. Applying this transform to each channel signal yields a representation of the beta band energy distributed in the two-dimensional time-frequency domain, which is used for subsequent coherence calculations and network weight assignment.

[0037] When constructing a coherent network, the core idea is to quantitatively measure the degree of synchronized activity between different brain regions in the beta band, and then use this as the connection weight between each node in the network. To this end, this method performs coherence analysis on the beta-band components of any two channel signals within a sliding time window. By calculating the coherence value of the signal, a metric reflecting the degree of resonance between the two signals in a specified frequency band can be obtained. Unlike traditional static connectivity analysis, this process has a clear time dependence, that is, each time window corresponds to a new set of coherence value matrices, reflecting the dynamic evolution of the brain network. This construction method can not only reflect the short-term coordinated relationship between different brain regions, but also preserve information about the changes in coherence patterns before and after the initiation of behavior.

[0038] The calculated results of the coherence value are used to construct a weighted network. In this network model, each EEG channel is regarded as a network node, and the edge weights between nodes are determined by the coherence value between them. The higher the edge weight, the stronger the functional coupling between the two brain regions in the beta band. This weighted network not only contains spatial distribution information, but also incorporates dynamic changes in the time dimension. It is an extension of the traditional static functional connection network. By advancing the time axis in a sliding window manner, the system can continuously construct multiple coherent network frames, forming a network evolution process with a time series structure, thereby providing continuous input for the extraction of dynamic features and behavioral discrimination.

[0039] In addition, in the process of building a coherent network, it is also necessary to combine the topological arrangement structure between EEG channels to ensure consistency between network connections and actual electrode layouts. In practical applications, EEG channels are arranged according to the international standard 10-20 system, and the geometric distribution between different electrodes affects the physical interpretability of their coherence values. Therefore, when calculating the connection strength, not only the frequency domain coherence is considered, but also the physical distance between channels is processed to avoid interference of non-physiological strong connections on the overall structure of the network. At the same time, in order to maintain the sparsity and robustness of the network structure, the connection edges below a certain threshold can be weakened or even deleted, thereby constructing a more recognizable EEG coherence network model.

[0040] Finally, after calculating the coherence values ​​between all channel pairs, a multi-frame time series weighted network cluster can be generated, providing basic data for subsequent analysis. These network frames serve as the core input for behavioral feature extraction and recognition in subsequent steps. The temporal evolution trend of their connectivity patterns is the key feature that this method aims to capture. Overall, this step not only achieves noise suppression and frequency band focusing of the original EEG data, but also effectively converts multi-channel biological signals into EEG networks with physical meaning and temporal structure through a combination of time-frequency analysis and network modeling, providing rich and structured input features for the behavior recognition algorithm.

[0041] Step 2: At each time point, extract the degree and clustering coefficient of each node in the weighted coherence network and calculate the dynamic evolution index; combine the network connection weights with the dynamic evolution index to construct a unified EEG-driven behavioral feature map; at each time point, for each pair of nodes, calculate their instantaneous phase difference, and simultaneously introduce the spatial geometric distance between the two electrodes to obtain the behavioral feature response index;

[0042] Specifically, this step first performs a structural analysis of the coherent network at each time point, and extracts the network degree and clustering coefficient of each node as basic topological indicators. The degree of a node reflects the activity level of the electrode in the functional connection network, that is, the number of significant functional connections between it and other brain regions in the current time window. Changes in degree can reflect the shift of EEG activity centers and the dynamic changes in brain region involvement. In particular, during the process of motor preparation or execution, the functional connectivity of certain specific brain regions may increase rapidly. The clustering coefficient is used to measure the tightness of the local sub-network around a node, reflecting whether the brain region where the node is located is in a highly integrated state at the current moment. For the initiation of cognitive behavior, changes in the aggregation of local networks are of great significance, and often produce obvious structural adjustments with the occurrence of specific behaviors.

[0043] After acquiring time series of degree and clustering coefficient, the method further introduces a dynamic evolution analysis strategy, performing temporal differencing on these indicators to extract first-order changes and second-order acceleration information along the temporal dimension. This analysis not only captures the absolute changes in indicator values ​​but also emphasizes their rate of change and trend, reflecting the dynamic transformation capabilities of EEG network structures during behavioral initiation. This dynamic evolution indicator exhibits strong burst responsiveness and high time sensitivity, effectively characterizing changes in the functional activity of behavior-related brain regions. In particular, during the transient phase when behavior transitions from preparation to execution, this indicator typically exhibits dramatic fluctuations, providing clear boundary signals for behavior recognition.

[0044] On this basis, to enhance the discriminative power of features, the method also combines the connection weights of the coherent network with the aforementioned dynamic evolution indicators to construct a unified feature map expression. This fusion not only retains the connection strength information in the coherent network, but also introduces the temporal evolution properties of the network topology, thereby forming a multi-dimensional, multi-scale feature representation system. The essence of the feature map is a tensor-like structure that combines the network weight matrix with the topological time derivative matrix, which is used to simultaneously encode signal synchronization, connection stability, and structural change trends. This map can be directly input into the subsequent behavior recognition module and has strong generalization and temporal reasoning capabilities.

[0045] In addition, this step further examines the distribution characteristics of the instantaneous phase difference between electrode pairs at each time point. As the basic component of the wavelet-transformed signal, the instantaneous phase can reveal the relative synchronization delay of signals between brain regions without changing the amplitude energy. In a functionally coupled network, the stability of the phase difference between two brain regions is often closely related to their degree of coordination. By calculating the instantaneous phase difference between all electrode pairs in each time window, this method not only captures the network coupling state but also provides a basis for identifying phase reconstruction patterns associated with specific behaviors.

[0046] To enhance the explanatory power of this feature in the spatial domain, this method also introduces the spatial geometric distance between electrodes as a modulation factor, incorporating distance into the response calculation of behavioral features. Physiologically, the strength of functional connectivity between different brain regions is significantly affected by their spatial distance. Strong synchronization signals are more likely to occur between closely spaced electrodes, but strong connections between distant brain regions are more behaviorally significant. Therefore, when calculating behavioral responses, the combination of phase difference and network weights is modulated by spatial distance to enhance the behavioral relevance expressed by long-range synchronous activation.

[0047] Step 3: Obtain the power spectrum parameters of all brain region signals of the target behavior in the analysis frequency band, combine them with the characteristic response index, calculate the target behavior recognition value, set the behavior threshold of the target behavior in the analysis frequency band, compare the target behavior recognition value with the behavior threshold, and determine whether the target behavior is activated;

[0048] During power spectrum extraction, this method uses the results of a short-time wavelet transform to integrate the energy of the signal within the beta band within each time window to obtain the total power of the frequency band at each electrode at that moment. Furthermore, by calculating the center frequency of the power spectrum, the center of gravity of the energy distribution in the current frequency band can be determined. Modeling the statistical mean and standard deviation of the center frequency over consecutive time periods helps assess the stability of the frequency distribution and the behavior-guided spectral reconstruction pattern. Furthermore, this method introduces the power spectral entropy metric to measure the flatness and information complexity of the spectrum. A higher power spectral entropy indicates a more uniform signal frequency distribution and more discrete brain activity states. A lower power spectral entropy typically indicates a concentrated burst of energy release within a specific frequency band, often directly related to specific motor behaviors or cognitive events. These power spectral parameters form a frequency-domain representation of behavioral characteristics, providing a stable and explanatory input for subsequent recognition.

[0049] At the same time, to achieve the fusion of frequency domain features and coherent network evolution characteristics, this method combines the frequency domain power spectrum features with the behavioral feature response index obtained in the previous stage to construct the target behavior recognition value. The behavioral feature response index comprehensively considers the connection strength, phase difference, spatial topology, and dynamic evolution trend between brain regions. It is a quantitative expression of the comprehensive response ability of the entire brain network during the behavioral triggering process. By combining this response index with the change in spectral energy, the accuracy and robustness of behavior recognition can be significantly improved. In particular, during the behavioral activation process with the overall synchronization of the neural network, the local aggregation and global synchronization of the coherent network structure will bring about a significant increase in spectral energy. This change is accurately captured by the recognition value function, manifested as a rapid increase in the recognition value.

[0050] To translate recognition values ​​into clear behavioral activation criteria, this method constructs a behavioral recognition threshold model for the target behavior in the beta frequency band. The threshold setting process is based on statistical analysis of extensive experimental data, taking into account the differences in recognition value distribution across different behavioral categories to ensure that the recognition value intervals for each behavior category are independent and non-overlapping. During the threshold determination process, the system compares the current recognition value with the threshold corresponding to the target behavior type in real time to determine whether the behavior is active. For example, if the recognition value exceeds the activation threshold for eye movement behavior but does not reach the judgment interval for manual behavior, the system determines that the current state is eye movement active. Similarly, when the recognition value significantly increases and crosses the high threshold interval corresponding to manual or foot movement behavior, the system can promptly identify higher-level or larger-scale behavioral execution states. The recognition threshold not only serves as a classification function but also provides a quantitative measure of behavioral intensity, allowing the system to not only identify whether a behavior has occurred but also preliminarily determine the degree of behavioral response and the extent of brain activity involved.

[0051] Furthermore, after selecting the beta wave frequency band as the analysis frequency band, the original EEG signal is processed using a bandpass filter, and the frequency range of the bandpass filter is 13Hz to 30Hz; the target behavior movement-related behavior includes at least: eye movement behavior, manual movement behavior and foot movement behavior.

[0052] The primary purpose of using a bandpass filter is to suppress interference components in other frequency bands unrelated to beta waves while retaining valid signal components within the beta band. Signal components above 30 Hz, such as gamma waves and myoelectric interference, are effectively attenuated, and components below 13 Hz, such as delta, theta, or alpha waves, are also filtered out. This ensures the frequency domain purity of the input signal and prevents information from non-target frequency bands from being mixed into behavioral characteristics, causing misjudgment. Furthermore, the filter design must also meet phase linearity requirements to prevent phase distortion caused by filtering from affecting subsequent instantaneous phase difference calculations and coherence analysis. Finite impulse response (FIR) filters or zero-phase IIR filters are typically used to maintain a balance between the signal's time domain structure and frequency domain accuracy.

[0053] After completing the frequency band limitation, the present invention limits the target behavior to the category of typical movement-related behaviors, which include at least three basic behavior types: eye movement behavior, manual behavior, and foot movement behavior. These three types of behaviors have highly specific spatial activation patterns and spectral response laws in neuroelectrophysiological activities, especially in the beta wave band, showing significant frequency energy enhancement, network synchronization enhancement, and coherent structure reconstruction. For example, eye movement behavior often activates low-intensity connections between the frontal lobe and the occipital lobe, showing relatively localized beta wave enhancement; manual behavior is more involved in the motor cortex area near the central sulcus, and its beta wave activity shows a stronger power increase and enhanced network coupling; foot movement behavior shows a more obvious cross-hemispheric long-distance synchronization feature. These behaviors have quantifiable differences in EEG network topology, power spectrum characteristics, and phase response patterns, so accurate identification of multiple types of behaviors can be achieved through a unified frequency band analysis framework.

[0054] Furthermore, the power spectrum parameters include: an average value of the power spectrum in a resting state, a standard deviation of the power spectrum in a resting state, an entropy value of the power spectrum, and an average value of the entropy value of the power spectrum.

[0055] Specifically, the average power spectrum in the resting state is obtained by time-domain averaging the power spectrum density of each electrode in the beta wave frequency band during the non-task period, that is, when the subject is not performing any target behavior. This parameter represents the energy baseline of the brain area of ​​an individual in a natural relaxed state and is an important reference for determining whether a certain brain area has behaviorally induced activation. When a behavior occurs, if the power spectrum density is significantly higher than the baseline value, it can usually be regarded as that the brain area is involved in a certain functional processing process. The standard deviation of the power spectrum reflects the degree of discreteness of power changes in the resting state and reveals the amplitude of fluctuations in spectral energy during the resting period. The smaller the standard deviation, the more stable the power spectrum in the resting period, and the easier it is to identify changes during behavioral activation as a significant response; if the standard deviation is large, a stronger power change is required to distinguish it from the resting state.

[0056] In order to further capture the complexity of the power spectrum in frequency distribution, this method introduces power spectrum entropy as a measurement indicator. Power spectrum entropy is an entropy value calculated based on the normalized probability density of the power distribution in the β frequency band, which represents the dispersion and complexity of the signal's frequency domain structure. The higher the entropy value, the more evenly the energy is distributed in the frequency band, and the signal exhibits the coordinated participation of multiple frequency bands; the lower the entropy value, the more concentrated the energy is in a narrower frequency band, which may correspond to the mandatory neural activation induced by a specific behavior. In behavior recognition, power spectrum entropy can provide more detailed discrimination capabilities than total energy, and is particularly suitable for identifying behavioral processes in which the frequency domain pattern is reconstructed but the total energy does not change significantly.

[0057] In addition, in order to normalize and standardize the changes in power spectral entropy, this method also introduces the average value of power spectral entropy in the resting state as a reference benchmark. When the subject is not performing a task, the system records the power spectral entropy of each electrode for a long time, and averages the entropy values ​​of all sampling points during this period to obtain a baseline level of entropy. In actual recognition, the power spectral entropy of the behavioral period will be compared with this baseline value to assess whether the frequency complexity of the current signal exceeds the normal physiological fluctuation range. This comparison mechanism helps to avoid recognition errors caused by differences in neural regulation between individuals and improves the stability of behavioral recognition across subjects.

[0058] Furthermore, the network connection weight is:

[0059] ;

[0060] in, Indicates electrode EEG signals and electrodes The EEG signal at time The network connection weight at time ; is the time window; is the time-integrated variable; Electrodes in the beta wave band EEG signals and electrodes The EEG signal at time The coherence value at time ; Electrodes in the beta wave band The spectral entropy of the EEG signal; Electrodes in the beta wave band The spectral entropy of the EEG signal; For electrodes The voltage amplitude of the EEG signal; For electrodes The voltage amplitude of the EEG signal; For time Electrodes in the beta wave band The center frequency of the EEG signal in the beta wave band The power spectral density at ; For time Electrodes in the beta wave band The center frequency of the EEG signal in the beta wave band The power spectral density at ; For electrodes The phase of the EEG signal; For electrodes The phase of the EEG signal; It is an indicator function, which takes a value of 1 when the condition is met, otherwise it takes a value of 0; For electrodes The signal-to-noise ratio of the EEG signal; is the set signal-to-noise ratio threshold; For electrodes The signal-to-noise ratio of the EEG signal.

[0061] The entire weight is defined as a sliding time window The core idea of ​​the integral average of the function terms at each moment is to analyze the time The coherence relationship and multi-dimensional neural electrical characteristics are collected in a period of time before the moment, and the functional connection strength between brain regions is estimated in this local time window. This method is manifested in engineering implementation as a weighted average of the distribution of samples in the sliding window, so as to obtain a connection value that is stable and has the ability to respond to dynamic trends, avoiding the recognition instability caused by the sharp fluctuation of instantaneous values. At the same time, the integral form reflects the time domain filtering characteristics, has a high response sensitivity to the enhancement of connections related to slow-changing behaviors, and shows strong robustness to sudden artifact disturbances. In the integral kernel, the first factor that appears is This is the electrode With electrodes In time The coherence value of the β wave frequency band. This coherence value is the most basic calculation element in the functional connection network, reflecting whether there is a stable phase relationship and frequency matching between two neural signals in a specific frequency band, that is, the degree of frequency domain synergy. The coherence enhancement of the β wave frequency band is usually closely related to the activation of the movement-related functional network. Therefore, in the process of identifying eye movement, hand movement and foot movement in the present invention, Coherence is a key indicator of brain region coordination. Unlike simple correlation calculations, coherence is modeled in the frequency domain, emphasizing cross-temporal consistency and functional network characteristics, resulting in greater behavioral specificity and physiological interpretability.

[0062] However, the coherence value itself cannot fully reflect the effectiveness of the connection between brain regions and the correlation with behavior. To this end, the present invention introduces multiple modulation factors to adjust the coherence value based on the multidimensional weight of the physiological signal attributes. Among them, the second product term is ,in and Represents electrodes and electrodes Spectral entropy in the β band. The larger the spectral entropy, the more uniform the energy distribution, the more complex the spectrum, and the lack of obvious dominant frequency components, which may mean that the current brain area is in a non-specific activation state or a random fluctuation state. On the contrary, the smaller the spectral entropy, the more concentrated the energy, which usually corresponds to the behavioral correlation neural activation at a specific frequency. The modulation is achieved in the form of average entropy. The smaller the entropy value, the greater the connection value, and the larger the entropy value, the inhibition of the connection, reflecting the enhancement effect of brain area "concentration" on the quality of connection. The third product term involves the joint ratio of voltage amplitude and power spectrum density. The formula is ,in is the instantaneous voltage amplitude, which is an intuitive physical quantity reflecting the intensity of neural activity, and the power spectral density in the denominator is The power spectral density (PSD) is the intensity of the energy distribution at the center frequency of the beta wave in that channel. This ratio combines time-domain amplitude with frequency-domain energy structure to emphasize the responsiveness of high-amplitude, low-spectral-density signals to behavior. Large voltage amplitudes indicate strong neural activity, while high power spectral density indicates overly concentrated signals, potentially lacking frequency modulation flexibility. This allows the system to more readily identify connection states that exhibit amplitude activity but are not overly concentrated in the spectrum, making them highly representative of motor behavior.

[0063] Later, the This is the electrode With electrodes The cosine term of the instantaneous phase difference measures the instantaneous synchronization of neural signals on a microscopic time scale. In neural activity, phase synchronization is a more sensitive connection indicator than power, and is particularly suitable for describing the phenomenon of long-distance, low-latency co-activation across brain regions in neural networks. When the phases of the two signals are exactly the same, the cosine value is 1, indicating the maximum connection strength; when the phase difference is 90 degrees, the term is 0; and when the phases are opposite, it is -1. This term therefore provides a direction-independent instantaneous synchronization enhancement, which helps the system identify short-term strong connection events induced by behavior. The last two factors are and , is an important part of the signal quality control mechanism. SNR, or signal-to-noise ratio, is a key indicator for measuring the proportion of effective neural activity in the signal. is a preset constant used to determine whether the current signal is valuable for analysis. If the signal-to-noise ratio of a channel falls below this threshold, the connections it constructs are deemed unreliable and are removed or assigned a value of 0. In practical applications, this control method can effectively eliminate spurious connection enhancement caused by electromyographic artifacts, eye movement interference, and environmental noise, thereby ensuring the physiological credibility and behavioral explanatory power of the connection network structure.

[0064] The integral expression constructed This is a dynamic network connection weight capable of multi-source information fusion. It not only exhibits smoothness and continuity on the timeline but also integrates multiple dimensions, including frequency domain coupling, phase consistency, amplitude intensity, spectral concentration, and signal quality control, at the physical level. The design of this connection weight follows the multi-scale coupling mechanism of neural network connections. Instead of relying on a single feature for EEG recognition, it employs a dynamic feature combination approach to model behavioral responses. By using this weight definition, the weighted coherent network constructed by the present invention accurately reflects the functional connectivity patterns of the brain under the current behavioral state and provides a high-resolution structural foundation for subsequent network topology index extraction, dynamic evolution trend modeling, and behavioral recognition feature generation. Considering the entire recognition process, this weight formula performs connection modeling on the cleaned beta wave signal after frequency band screening and signal filtering in the preprocessing phase. During the dynamic network construction phase, the connection value sequence is continuously updated using a sliding window approach, forming a time-varying coherent network structure sequence. This structure sequence becomes an important input source for behavioral feature mapping and is integrated with modules such as power spectrum parameters and evolution indicators to form a unified recognition feature vector. Ultimately, with the support of the recognition threshold mechanism, stable recognition of eye movements, hand movements, foot movements and other motor behaviors can be achieved in a complex EEG background.

[0065] Further, time Dynamic evolution index for:

[0066] ;

[0067] in, For time Time The clustering coefficient of the nodes corresponding to the electrodes; For time Time The degree of the node corresponding to each electrode.

[0068] This expression uses all nodes (i.e. EEG channels) in the network as indexes and summarizes the The relationship between the degree of topological change at this moment and its signal quality forms an estimate of the macro-dynamics of the entire network. Specifically, this indicator consists of two parts: the first half measures the rate at which the network structure changes over time, and the second half regulates the uncertainty caused by signal noise. First, focus on the expression , which means that at time Time Electrode The degree of the corresponding node. Node degree is a basic quantity in network topology analysis, indicating how many other nodes the node has functional connections with at the current time. In EEG coherence networks, a higher degree usually indicates that the brain region is in a more central position and participates in multiple functional pathways. Find the time first-order derivative and take the absolute value, that is , represents the temporal rate of change in the number of connections at that node. A large derivative indicates a dramatic shift in the connectivity structure of the brain region, potentially indicating activation or deactivation of behaviorally relevant brain regions, and thus a high degree of behavioral relevance. Absolute value processing is used to unify the magnitude of changes when the degree increases and decreases, thereby capturing sudden structural evolution rather than linear trends in a single direction.

[0069] Watch next , which represents the node The clustering coefficient is used to describe the compactness of the local subgraph in which the node is located in the current network structure. A high clustering coefficient indicates that its adjacent nodes are connected to each other, forming a compact local network cluster. Unlike the node degree, the clustering coefficient reflects more local integration and stability. During behavioral tasks, some brain regions not only increase the number of their connections, but also form a highly organized modular structure through local reconstruction. Find the second derivative of time and take the absolute value, that is , used to quantify the "acceleration" of the change in the local organizational structure of the node. Compared with the first-order derivative, the second-order derivative can better reflect the degree of sudden change in the structure, especially at the moment of behavior initiation, behavior conversion period and cognitive load adjustment stage. This indicator shows a significant response, which helps to distinguish short-term behavior induction from long-term trend changes. After adding the above two items, it constitutes the acceleration of each node in time. A comprehensive assessment of topological change trends over time: This includes both the rate of change in the number of connections and the degree of fluctuation in the local network structure. In theory, a larger weighted term indicates a dynamically active node, with its connected network undergoing significant structural changes and a high likelihood of being involved in a chain of behaviorally related activities.

[0070] The structural dynamic response of each node mentioned above needs to be combined with its signal quality in order to comprehensively evaluate its effectiveness. To this end, the method introduces the signal-to-noise ratio correction term , used to punish nodes with low signal-to-noise ratio. The function makes the correction value smaller when the signal-to-noise ratio is high, and the contribution weight to the overall dynamic index is larger; while the value of this item is larger for nodes with low signal-to-noise ratio, resulting in a decrease in the weight of their topological change. This mechanism avoids misjudging the structural changes of nodes with unreliable sources as real behavior-induced responses, thereby improving the accuracy and robustness of the overall recognition. Introducing a logarithmic function instead of nonlinear scaling also helps to control the degree of suppression of the maximum SNR value and enhance numerical stability. Finally, by adding the structural change rate of the nodes corresponding to all electrodes and their contribution after signal-to-noise ratio correction, the time Dynamic evolution index of the entire coherent network at any moment The larger the value of this indicator, the more likely the network is in a period of intense restructuring, which is likely the time point for the initiation or transformation of behavior. Conversely, if the value is continuously low, it indicates that the current network structure is stable, the behavior has not occurred, or it is in an inactive period.

[0071] Further, time The characteristic response index is :

[0072] ;

[0073] in, For time The electrode The phase and time of the EEG signal The electrode The phase difference of the EEG signal; For electrodes With electrodes The spatial Euclidean distance of is the spatial attenuation scale coefficient, when When the unit is centimeters, the value is 2 to 5; is the total number of electrodes.

[0074] External normalization term This is to ensure that the index does not lose comparability with changes in the number of electrodes. Electrode channels can be formed undirected connection pairs, all connection pairs are summed and then normalized and averaged to ensure The values ​​in different subjects and different EEG systems have the same physical meaning. The innermost summation term is the sum of all electrode pairs. Perform combined calculations, where , indicating that it does not include its own connections. This summation process fully incorporates the connection structure of the entire EEG coherence network into the response modeling field of view, thereby constructing a holistic whole-brain response index.

[0075] In the summation, the first part is , indicating that at time Moment Electrode and The network connection weight between them is derived from the aforementioned multi-factor integral expression. This weight itself has integrated the coherence value, spectral entropy, amplitude, power density, phase consistency and signal-to-noise ratio control factors, reflecting the strength and reliability of the connection, and is the prerequisite for whether the connection pair can actually have an effect on the behavioral response. Introducing this term in the characteristic response index can ensure that the corresponding electrode pair contributes weight to the network behavioral response only when there is an actual neural coupling relationship. The second term is , indicating the electrode and In time The instantaneous phase difference reflects the relative timing differences in neural activity between brain regions. In neuroscience, phase difference is an important indicator to measure whether a neuronal cluster is in a synchronous or asynchronous state. When the phase difference between two brain regions is small, it means that their signal oscillations tend to be synchronized, indicating that they may be in a state of co-activation. When the phase difference increases significantly, their synchronization decreases and the synergy may weaken. Therefore, this term introduces the degree of network synergy through the absolute value of the phase difference, enhancing the sensitivity to synchronous responses induced by behavior. It is worth noting that the cosine term is not used, but the absolute value is retained, emphasizing the amplitude rather than the direction of the phase difference, which is suitable for cooperative modeling under nonlinear conditions.

[0076] The third term is the spatial modulation factor ,in Indicates electrode and The Euclidean distance between the is the spatial attenuation scale coefficient, which ranges from 2 to 5. This term reflects the spatial propagation delay and coupling attenuation mechanism between neural activities. In the cerebral cortex, it is easier to establish synchronous activation patterns between close areas, and even if there is coherence between distant areas, the synchronization ability is often weakened due to structural limitations. Therefore, an exponential decay function is introduced to reasonably suppress the response contribution of long-range connections. This term ensures that the model has a priority response when identifying local synchronous clusters related to behavior, while retaining the ability to process certain long-distance coupling patterns, but making exponential adjustments to the contribution strength. The last multiplication factor is , that is, time The dynamic evolution index at the moment, which has been defined in detail in the previous section, reflects the rate of structural change and the degree of instability of the entire EEG network at that moment. This term is introduced in this formula to emphasize that local connectivity and phase response have practical behavioral significance only when there is a drastic evolution trend in the macroscopic structure of the brain network. If the network as a whole is in a static state, even if some local connectivity indicators are high, they may just be background noise or stable fluctuations, so their contribution must be adjusted according to the global dynamics. By introducing this term, It has become a composite response indicator that integrates network structure information, phase coordination relationship, spatial topology and macro dynamic trends. From a physical point of view, The value of reflects whether the brain is currently in a high-response state: when the connection strength is high, the phase difference is significant, the network evolves violently, and the spatial coupling structure is active, the value will increase significantly. Conversely, when the connection network is stable, the signal phase change is weak, or the network structure is stable, the value tends to be low. In particular, for behavior recognition tasks, It can be used to determine whether a behavior is in an active state, or further used for behavior recognition value calculation and classification boundary judgment.

[0077] Further, time Time The clustering coefficient of the nodes corresponding to the electrodes for:

[0078] ;

[0079] in, For the The actual number of edges between the neighbors of the node corresponding to the electrode.

[0080] In this expression, Indicates time Moment The degree of the node corresponding to the electrode, that is, how many other electrodes in the EEG coherence network have non-zero connection weights with the electrode; Indicates the number of edges that actually exist between these adjacent nodes. All possible node pairs in the neighbor set , if the node With node If there is a valid connection between them (i.e., the weight is non-zero), it is counted .therefore, is an indicator of the closeness of the connection between neighbors, and the denominator is the theoretical maximum number of edges that can exist between neighbors, so the formula gives It is a normalized ratio, ranging from 0 to 1.

[0081] when When , it means that all neighboring nodes of the node are connected to each other, forming a complete clique structure, that is, each node The connected nodes are also connected to each other, forming a closed local functional cluster. This usually means that the brain area is in a highly integrated state at that moment and may be the central area for behavioral-related neural collaborative processing. When the node Neighbors are not connected to each other, and the local structure is loosely connected radially, which may be just an information path rather than an actual processing center.

[0082] In the present invention, the introduction of clustering coefficient has multiple functional significances. First, in the dynamic evolution index In the definition of , the second-order derivative of the node clustering coefficient It is used as a measure of local structural mutations in a network. Specifically, when the connectivity between a node and its neighbors changes dramatically within a short period of time, particularly from a loosely distributed to a highly clustered state or vice versa, this term will rise rapidly, significantly affecting the overall network dynamics index. Such localized mutations are often characteristic of the transition from preparation to execution of neural activity, or when external stimuli are introduced, and have clear physiological and behavioral relevance.

[0083] Secondly, the clustering coefficient also participates in the characteristic response index of the entire network The spatial topological regulation of brain regions. The coherence between brain regions depends not only on the weight and phase relationship between them, but also on the local clustering pattern of the network structure. For example, eye movement behaviors often activate short-path clusters between the frontal lobe and the occipital lobe, manual behavior activates high-density cluster structures around the central sulcus, and foot movement behaviors involve long-distance connections across the hemispheres but with a lower degree of clustering. The clustering coefficient, as a local modular connection feature, is an important geometric topological quantity that characterizes the differences in the specific connection maps of these behaviors.

[0084] Furthermore, in network visualization and graph modeling, the clustering coefficient can serve as an important basis for selecting connection sparsification thresholds and adjusting feature boundaries. In practical applications, since EEG networks are naturally sparse graphs, connections between nodes are not evenly distributed, but rather have a small number of densely connected centers and a large number of edge nodes. By analyzing the changing trends of the node clustering coefficient, key module regions under behavioral activation can be identified, which can be used for neural engineering tasks such as electrode selection, spatial channel reconstruction, or regional functional assignment.

[0085] It is worth noting that in order to avoid the situation where the denominator of the formula is zero in actual calculation, the system will Nodes with a degree of 0 or 1 are skipped from clustering coefficient calculation or assigned a value of zero. This is because when a node has a degree of 0 or 1, its neighbors cannot be connected, and the theoretical maximum number of edges is 0, making the formula meaningless. Therefore, the clustering coefficient is primarily applicable to nodes with a degree greater than or equal to 2.

[0086] In engineering implementation, the calculation of clustering coefficient depends on the binarization or sparse processing of network connection weights. is a continuous value, but to achieve To accurately count edges, it's often necessary to set a connection strength threshold, deeming all connections above a certain threshold "connected" and the rest as zero. This threshold can be adaptively set using statistical methods, for example, retaining the top 10 percent of the maximum values ​​in the connection matrix. It can also be set as a tuning parameter during the recognition model training phase. In practice, combining connection strength with neighbor topology density can significantly improve the expressiveness of recognition features.

[0087] Further, time Target behavior recognition value for:

[0088] ;

[0089] in, For time Time Electrode exist The integrated value of the power spectrum density under the wave frequency band, , is the frequency integration variable; is the average power spectrum of electrode 𝑖 in the resting state; For electrodes Standard deviation of the power spectrum in the resting state; For time Electrodes in the beta wave band The spectral entropy of the EEG signal; For time Time Electrode The entropy value of the power spectrum; is the average entropy value of the power spectrum of all electrodes in the resting state.

[0090] In the EEG-driven behavior recognition method based on the dynamic evolution of coherent network features, the target behavior recognition value It is the core judgment indicator throughout the entire recognition system. Its purpose is to use the coupling relationship between the structural network response parameters extracted from the EEG signal and the frequency domain dynamic characteristics to determine at a certain point in time. The design logic of this indicator integrates multiple dimensions, such as neural network connection strength, phase coherence, spectrum energy variation, and spectrum complexity reorganization, to achieve accurate modeling and quantitative determination of behavioral states in a complex physiological signal environment.

[0091] The whole formula consists of two parts: the prefactor The posterior factor represents the characteristic response strength of the brain's functional connectivity network at the current moment, a measure of global brain activity from a structural and dynamic perspective. The posterior factor is a weighted average of the spectral responses of a multi-channel electrode, reflecting the activation state of the brain region in the frequency domain. This expression not only integrates information from the time, frequency, and spatial domains with network topology, but also provides a clear physiological interpretation path and engineering implementation logic.

[0092] Prefactor It has been derived and defined in detail in the previous section. It is an integrated network connection strength between multiple electrodes. , phase difference , spatial distance weight , and the network evolution index The high-order coupling index, whose value reflects the global responsiveness level of the brain network at the current time point. If the value is close to 0, it means that the current brain network as a whole is in a resting or low-activity state. Even if the power of some electrodes is locally enhanced, it does not have sufficient behavioral explanatory power. On the contrary, when the value rises, it means that the brain as a whole is in a state of linkage, high synchronization, and high evolution speed, which is suitable for judging that the neural activity related to the target behavior is ongoing. Therefore, in the entire recognition value, It plays the role of dynamic gating and response regulation.

[0093] The post-factor part is the frequency domain composite response intensity averaged across electrodes, and the internal factor is the electrode The product of the three spectrum characteristic factors at the current moment. The first factor is:

[0094] ;

[0095] This expression is the current time point electrode exist Total power in the frequency band The average power spectrum of the resting state Subtract and use the standard deviation in the resting state The result of normalization. Defined as:

[0096] ;

[0097] in It is an electrode In time Time, frequency The power spectral density under the β wave frequency band (13–30 Hz) is integrated. The physical significance of this normalization term is to measure whether the power of the current electrode deviates significantly from its resting baseline state. If the current power spectral density is significantly higher than the resting mean, the term is positive, indicating that the electrode may be in a state of behaviorally induced activation; if it is lower than the baseline, the term is negative, indicating that the brain region has no significant signs of activity. The standard deviation is used to eliminate individual EEG signal intensity differences and achieve consistent assessment across subjects and electrode channels.

[0098] The second factor is:

[0099] ;

[0100] in, Indicates electrode In time time, The spectral entropy under the wave frequency band is a measure of the complexity of the signal's energy distribution in the frequency domain. The larger the spectral entropy, the more uniform the power distribution and the more complex the frequency components, indicating that the signal may not have a frequency-focused behavioral characteristic; conversely, the smaller the entropy value, the more concentrated the frequency distribution, which may correspond to enhanced behavioral-related neural discharges in a certain frequency band. This factor uses a nonlinear transformation method to enhance medium and low entropy values ​​and suppress high entropy values, thereby strengthening the contribution of electrodes with obvious frequency-focusing characteristics. Add This is to avoid logarithmic divergence or division by zero in the denominator when the entropy value approaches 0. It is a typical numerical stability processing method.

[0101] The third factor is:

[0102] ;

[0103] in, Indicates electrode The entropy value of the power spectrum at the current moment, and It is the average value of the power spectrum entropy of all electrodes in the resting state. This factor uses the Sigmoid function to map the change of power spectrum entropy to a continuous value between 0 and 1. When the spectral entropy is higher than the resting state average, this factor approaches 1, indicating a significant change in the spectral structure, which may be related to behavioral execution. When it is lower than the resting entropy level, the Sigmoid value approaches 0, indicating that the current spectral entropy has not changed significantly and should not be considered a behavioral activation feature. The Sigmoid function provides flexible boundaries and nonlinear discrimination capabilities, and has better stability and differentiability than hard thresholding, which facilitates the subsequent application of behavioral recognition values ​​in regression or classification models.

[0104] Multiply the above three factors together, and we get:

[0105] ;

[0106] This product constitutes the electrode The comprehensive frequency-domain activation feature value of the target behavior response. Its physical meaning is: when the power of an electrode increases significantly at the current moment, the spectrum shows structural concentration (low entropy), or entropy reorganization occurs (higher than the baseline), then the electrode has a high behavioral feature contribution value at that moment.

[0107] The characteristic contribution values ​​of all electrodes are averaged, that is:

[0108] ;

[0109] This operation reflects that behavioral activation is a systemic phenomenon involving the coordinated participation of multiple regions, rather than the isolated activation of individual brain regions. By averaging across electrodes, the misleading effects of localized artifacts on overall recognition can be reduced, while also preserving the widespread response characteristics of behaviorally relevant brain regions.

[0110] Finally, this average result is compared with the structural response index Multiply together to form a complete behavior recognition value This value has the following characteristics: the larger the value, the stronger the brain structure connectivity, the higher the phase coupling, the more active the network evolution, and the presence of power enhancement, spectrum focusing, or spectrum entropy reorganization of multiple electrodes in the frequency domain, indicating that the target behavior is very likely to be activated. Conversely, if the structural response is low or the spectral characteristics are not significant, even if a certain parameter is large, it will not significantly increase the recognition value, thereby maintaining the stability and noise resistance of the recognition system.

[0111] Identification value Can be used with behavioral activation thresholds Real-time comparison enables online behavior judgment. Its continuous output feature also allows it to be used as a regression variable to measure the intensity of behavioral response or to build a behavioral boundary detection model based on the time series of recognition values.

[0112] Furthermore, when the target behavior is eye movement behavior, the recognition threshold The value range is 0.20 to 0.30; when the target behavior is manual behavior, the recognition threshold The value range is 0.35 to 0.45; when the target behavior is foot behavior, the recognition threshold The value range of is 0.50 to 0.60; if Greater than the recognition threshold , then the target behavior is judged to be in the activated state.

[0113] When the target behavior is eye movement behavior, the system sets its recognition threshold The value range is 0.30 to 0.45. This is based on the analysis of the EEG signal characteristics caused by eye movement behavior during execution. Eye movement usually causes a relatively localized network response in the prefrontal to occipital region. The characteristic response index Although there is an improvement in the process, the spectral energy change and spectral entropy reconstruction degree are lower than those of manual or foot movements, so the overall recognition value is The distribution center of gravity under eye movement behavior is usually lower than that of other behaviors. According to experimental statistics, recognition values ​​below 0.30 are mostly non-eye movement states, and between 0.30 and 0.45 have a higher probability of being eye movement activation areas, so this range is set as the recognition threshold interval for eye movement behavior. , the system determines that the brain is in an eye movement behavior activation state at this time.

[0114] When the target behavior is foot movement, its recognition threshold The value range is 0.50 to 0.60. Foot movement behavior is usually accompanied by higher characteristic response index and power spectrum enhancement characteristics because it involves the subcortical area of ​​motor function and a wider range of cross-hemispheric functional area coordination. Especially in the central area and parietal lobe area, the spectral energy and spectral entropy reconstruction degree are significantly enhanced. The mean and variance of are higher than those of eye movement or hand movement. According to the interval analysis of the statistical learning model, the distribution of recognition values ​​corresponding to foot movement is usually concentrated between 0.52 and 0.58. Based on this, the system sets 0.50 as the minimum judgment threshold and 0.60 as the upper limit of the stable interval to ensure recognition accuracy and stability. Therefore, during the recognition process, if the current recognition value , that is, the foot movement behavior is considered to be in an activated state.

[0115] The recognition logic can be summarized as follows:

[0116] If the target behavior is eye movement behavior, set the threshold ,when When , the eye movement behavior is judged to be in an activated state;

[0117] If the target behavior is foot movement, set the threshold ,when , the foot movement behavior is judged to be in the active state.

[0118] It should be noted that the recognition threshold corresponding to each behavior It is not a fixed single value, but is selected from a reasonable range verified by physiological experiments. It can be further optimized according to the individual parameters of the subjects, signal sampling conditions, or the results of recognition model training. In the system implementation, the threshold selection can also be dynamically and adaptively adjusted through cross-validation to meet the specific discrimination requirements under different subjects or task types. Ultimately, this behavior activation judgment mechanism based on discriminative behavior recognition values ​​enables this method to have a clear and continuous behavior recognition logic, and can form non-overlapping recognition boundaries between multiple behavior types, thereby ensuring high accuracy and mutual exclusivity of behavior judgment, and providing a stable threshold control strategy for multi-category motor EEG recognition systems.

[0119] The following is a complete implementation example of an EEG-driven behavior recognition method based on the dynamic evolution of coherent network features, including the original EEG data parameters, the processing process of each stage, the application of formulas, and the final behavior recognition results. This example uses eye movement behavior as the recognition target and selects The spectral range (13–30 Hz) was analyzed.

[0120] Number of electrodes: , taken from the standard 10-20 system Fp1, Fp2, C3, C4, P3, P4, O1, O2. Sampling rate: 256 Hz; filter bandpass range: 13 Hz to 30 Hz (beta wave); time window , sliding step: 0.25,s; current analysis time point: ; Signal-to-noise ratio threshold: .

[0121] Electrode , For example, calculate Coherence value under wave frequency band: ;

[0122] Spectral entropy of electrodes C3 and C4 (normalized): ;

[0123] Instantaneous voltage amplitude (unit: ): ;

[0124] Power spectral density at the center frequency: ;

[0125] Phase difference: ;

[0126] Signal-to-noise ratio: ;

[0127] Spatial distance (scalp electrodes): ,Pick ;

[0128] ;

[0129] right electrodes to build a network, a total of indivual Yes, take part of the calculation: set the average connection weight ; Average phase difference ; Spatial distance , take the average value of 5; dynamic evolution index ;

[0130] ;

[0131] ;

[0132] Actual measurement Total power in the band:

[0133] ;

[0134] ;

[0135] ;

[0136] The product of the three: ;

[0137] Assume that after averaging the 8 electrodes: ;

[0138] ;

[0139] The current recognition target is eye movement behavior;

[0140] Threshold setting range: ;

[0141] current , meeting the recognition activation conditions.

[0142] Therefore, the system judges that eye movement behavior has been activated.

[0143] Although specific embodiments of the present invention have been described above, those skilled in the art will appreciate that these specific embodiments are merely illustrative, and that those skilled in the art may omit, substitute, and modify the details of the methods and systems described above without departing from the principles and spirit of the present invention. For example, combining the above method steps to perform substantially the same functions in substantially the same manner to achieve substantially the same results falls within the scope of the present invention. Accordingly, the scope of the present invention is limited solely by the appended claims.

Claims

1. A method for recognizing EEG-driven behaviors based on the dynamic evolution of coherent network features, characterized by: The method comprises: Step 1: Standard preprocessing is performed on the raw EEG signals. The beta wave band is selected as the analysis band, and the instantaneous frequency components of the analysis band are extracted using short-time wavelet transform. For each pair of electrodes, the coherence value in the analysis band is calculated within a specified time window to measure the synchronization strength of the two EEG signals. This value is used as the network connection weight to construct a weighted coherence network. Step 2: At each time point, extract the degree and clustering coefficient of each node in the weighted coherence network and calculate the dynamic evolution index. Combine the network connection weights with the dynamic evolution index to construct a unified EEG-driven behavioral feature map: At each time point, for each pair of nodes, calculate their instantaneous phase difference and introduce the spatial geometric distance between the two electrodes to obtain the behavioral feature response index. Step 3: Obtain the power spectrum parameters of all brain region signals of the target behavior in the analysis frequency band, combine them with the characteristic response index, calculate the target behavior recognition value, set the behavior threshold of the target behavior in the analysis frequency band, compare the target behavior recognition value with the behavior threshold, and determine whether the target behavior is activated; time Dynamic evolution index for: ； in, For time Time The clustering coefficient of the nodes corresponding to the electrodes; For time Time The degree of the node corresponding to each electrode; time The characteristic response index is : ； in, For time The electrode The phase and time of the EEG signal The electrode The phase difference of the EEG signal; For electrodes With electrodes The spatial Euclidean distance of is the spatial attenuation scale coefficient, when When the unit is centimeters, the value is 2 to 5; is the total number of electrodes; Indicates electrode EEG signals and electrodes The EEG signal at time The network connection weight at time ; For time Time Electrode The signal-to-noise ratio of the EEG signal.

2. The method for recognizing EEG-driven behaviors based on dynamic evolution of coherent network features according to claim 1, characterized in that: After selecting the beta wave frequency band as the analysis frequency band, the original EEG signal is processed using a bandpass filter, wherein the frequency range of the bandpass filter is 13 Hz to 30 Hz; the target behavior movement-related behavior includes at least: eye movement behavior, manual movement behavior and foot movement behavior.

3. The method for recognizing EEG-driven behaviors based on the dynamic evolution of coherent network features according to claim 2, characterized in that: The power spectrum parameters include: the mean value of the power spectrum in the resting state, the standard deviation of the power spectrum in the resting state, the entropy value of the power spectrum, and the mean value of the entropy value of the power spectrum.

4. The method for recognizing EEG-driven behaviors based on the dynamic evolution of coherent network features according to claim 3, characterized in that: The network connection weight is: ； in, is the time window; is the time-integrated variable; Electrodes in the beta wave band EEG signals and electrodes The EEG signal at time The coherence value at time ; Electrodes in the beta wave band The spectral entropy of the EEG signal; Electrodes in the beta wave band The spectral entropy of the EEG signal; For electrodes The voltage amplitude of the EEG signal; For electrodes The voltage amplitude of the EEG signal; For time Electrodes in the beta wave band The center frequency of the EEG signal in the beta wave band The power spectral density at ; For time Electrodes in the beta wave band The center frequency of the EEG signal in the beta wave band The power spectral density at ; For electrodes The phase of the EEG signal; For electrodes The phase of the EEG signal; It is an indicator function, which takes a value of 1 when the condition is met, otherwise it takes a value of 0; For electrodes The signal-to-noise ratio of the EEG signal; is the set signal-to-noise ratio threshold; For electrodes The signal-to-noise ratio of the EEG signal.

5. The method for recognizing EEG-driven behaviors based on dynamic evolution of coherent network features according to claim 4, characterized in that: time Time The clustering coefficient of the nodes corresponding to the electrodes for: ； in, For the The actual number of edges between the neighbors of the node corresponding to the electrode.

6. The method for recognizing EEG-driven behaviors based on dynamic evolution of coherent network features according to claim 5, characterized in that: time Target behavior recognition value for: ； in, For time Time Electrode exist The integrated value of the power spectrum density under the wave frequency band, , is the frequency integration variable; is the average power spectrum of electrode 𝑖 in the resting state; For electrodes Standard deviation of the power spectrum in the resting state; For time Electrodes in the beta wave band The spectral entropy of the EEG signal; For time Time Electrode The entropy value of the power spectrum; is the average entropy value of the power spectrum of all electrodes in the resting state.

7. The method for recognizing EEG-driven behaviors based on dynamic evolution of coherent network features according to claim 6, characterized in that: When the target behavior is eye movement behavior, the recognition threshold The value range is 0.20 to 0.30; when the target behavior is manual behavior, the recognition threshold The value range is 0.35 to 0.45; when the target behavior is foot behavior, the recognition threshold The value range of is 0.50 to 0.60; if Greater than the recognition threshold , then the target behavior is judged to be in the activated state.

Images