An algorithm for complexity analysis of EEG signals based on multi-domain fusion

Abstract

This invention relates to the field of computer technology and provides an algorithm for analyzing the complexity of EEG signals based on multi-domain fusion. The method includes: data preprocessing to acquire multi-band, multi-channel EEG signals; single-scale single-channel analysis at the single-scale level, performing single-scale single-channel analysis on the multi-band, multi-channel EEG signals and outputting multi-band fuzzy entropy; multi-scale single-channel analysis at the multi-scale level, performing multi-scale single-channel analysis on the multi-band, multi-channel EEG signals and outputting a multi-scale fuzzy entropy vector; and multi-scale multi-channel analysis at the multi-scale level, performing multi-scale multi-channel analysis on the multi-band, multi-channel EEG signals and outputting multi-channel multi-scale fuzzy entropy. This invention, through multi-band decomposition combined with multi-scale multi-channel analysis, can comprehensively capture the dynamic variation characteristics of EEG signals across different frequency bands, scales, and channels, effectively solving the problem of incomplete information extraction in traditional methods and improving the ability to characterize the dynamic characteristics and complexity of EEG signals.

Description

Technical Field

[0001] This invention belongs to the field of computer technology, and in particular relates to an algorithm for analyzing the complexity of electroencephalogram (EEG) signals based on multi-domain fusion. Background Technology

[0002] Electroencephalography (EEG), as an important physiological electrical signal, is widely used in neuroscience, diagnosis of mental illnesses, brain-computer interfaces (BCI), and cognitive state assessment. EEG signals are characterized by non-stationarity, nonlinearity, high dimensionality, and multi-channel properties, and their complexity analysis is crucial for understanding brain functional states and identifying abnormal brain activity. Currently, various methods exist for analyzing the complexity of EEG signals, such as sample entropy, permutation entropy, and multi-scale entropy.

[0003] However, most existing methods are limited to single-scale or single-channel analysis, making it difficult to comprehensively capture the dynamic characteristics of EEG signals in time and space. Especially in multi-channel collaborative processing and multi-band coupling analysis, existing technologies still have the following shortcomings: traditional methods usually only analyze a single frequency band or single channel, ignoring the synergistic effects between different frequency bands and different channels, resulting in incomplete information extraction; although existing entropy analysis methods introduce multi-scale processing, they often use a single method in the coarse-grained process, failing to fully consider the dynamic characteristics of signals at different scales, making it difficult to achieve refined multi-scale entropy assessment. Summary of the Invention

[0004] The purpose of this invention is to provide an EEG signal complexity analysis algorithm based on multi-domain fusion, which aims to solve the problems existing in the background technology.

[0005] This invention is implemented as follows: an EEG signal complexity analysis algorithm based on multi-domain fusion, including data preprocessing, single-scale single-channel analysis, multi-scale single-channel analysis, and multi-scale multi-channel analysis.

[0006] The data preprocessing is used to acquire multi-band, multi-channel EEG signals. Specifically, the raw EEG data is first loaded, and its key parameters, including sampling frequency, total number of channels, and information at each time point, are extracted. At the same time, the relevant parameters of the four standard frequency bands δ / θ / α / β are set. Then, the EEG signals of three specified channels, C3, Cz, and C4, are extracted from the raw data. Subsequently, bandpass filtering of the above four frequency bands is performed on each channel, and finally, four sets of three-channel frequency band EEG signals are obtained.

[0007] The single-scale single-channel analysis is used to perform single-scale single-channel analysis on the multi-band multi-channel EEG signals and output multi-band fuzzy entropy. Its specific implementation includes: inputting the four sets of three-channel frequency band EEG signals into a time-series delay embedding module for delay processing to obtain a phase space reconstruction matrix; then inputting this reconstruction matrix into a Chebyshev distance calculation module for vector pairs within the frequency band to calculate the pairwise distances in the phase space, obtaining rhythm distance matrices for each frequency band; next, sending each rhythm distance matrix into a multi-band distance sorting and symbolization module to obtain a sorting index and symbolization matrix through symbolization processing; then inputting the sorting index and symbolization matrix into a statistical permutation pattern occurrence frequency module, obtaining matching permutation patterns and their probability distributions after permutation pattern comparison operations; finally, inputting the above permutation patterns and probability distributions into a Shannon entropy and maximum entropy calculation module for entropy value calculation and normalization processing, outputting multi-band fuzzy entropy.

[0008] The multi-scale single-channel analysis is used to perform multi-scale single-channel analysis on the multi-band, multi-channel EEG signals, outputting a multi-scale fuzzy entropy vector. Its specific implementation includes: inputting the four sets of three-channel EEG signals into a multi-scale coarse-grained data processing module, which performs averaging through a non-overlapping window to output a coarse-grained time series; subsequently, inputting the coarse-grained time series into a multi-band fuzzy entropy module at different scales, which calculates the entropy value at each scale, and finally outputs a multi-scale fuzzy entropy vector.

[0009] The multi-scale multi-channel analysis is used to perform multi-scale multi-channel analysis on the multi-band multi-channel EEG signals and output multi-channel multi-scale fuzzy entropy. Its specific implementation includes: inputting the four sets of three-channel frequency band EEG signals into a multivariate time series coarse-grained processing module, and outputting a coarse-grained data matrix through multi-channel non-overlapping window averaging; then inputting the data matrix into a multivariate time delay embedding module, and obtaining a multivariate embedding matrix after time delay processing; next, inputting the multivariate embedding matrix into a vector Chebyshev distance module, and generating distance vectors for each frequency band by calculating pairwise distances between vectors; then sending each frequency band distance vector into a distance sorting and symbolization module at different scales, and obtaining a sorted symbol matrix after symbolization processing; this symbol matrix is ​​then input into a statistical permutation module at different scales, and obtaining the probability distribution of each permutation pattern through comparison and matching operations; finally, inputting the probability distribution of each permutation into a module for calculating the Shannon entropy and normalized entropy value of each permutation distribution, and outputting multi-channel multi-scale fuzzy entropy after entropy value calculation and normalization processing.

[0010] The present invention provides an EEG signal complexity analysis algorithm based on multi-domain fusion, which has the following beneficial effects:

[0011] This invention, by combining multi-band decomposition with key channel EEG signals, comprehensively captures the characteristics of brain electrical activity under different rhythms, providing a rich information foundation for subsequent complexity analysis. By introducing symbolic permutation entropy at the single-scale, single-channel level, it achieves precise measurement of the static complexity of each channel and frequency band, effectively characterizing the local dynamic characteristics of EEG signals. Through coarse-grained processing at the multi-scale, single-channel level and calculating multi-band fuzzy entropy at different scales, it reveals the structural changes of EEG signals at different time scales, overcoming the limitations of single-scale analysis. By performing multivariate coarse-grained processing, multivariate time-delay embedding, and joint symbolic processing at the multi-scale, multi-channel level, it calculates the joint probability distribution of four frequency bands and three channels, effectively assessing changes in brain region coordinative complexity. This invention integrates the advantages of multi-scale, multi-channel, and multi-band approaches, enabling a more comprehensive and refined characterization of the dynamic complexity and nonlinear features of EEG signals, providing reliable technical support for neuroscience, clinical diagnosis, and brain-computer interfaces. Attached Figure Description

[0012] Figure 1 A flowchart of an EEG signal complexity analysis algorithm based on multi-domain fusion is provided for this invention.

[0013] Figure 2 This is a simulation diagram of four groups of three-channel EEG signals after data preprocessing according to the present invention;

[0014] Figure 3 This is a 2D phase space diagram of the four-band three-channel EEG signal of the present invention;

[0015] Figure 4 This is a comparison chart of multi-scale coarse-grained EEG signals in the delta band of this invention;

[0016] Figure 5 This is a comparison diagram of the multi-scale coarse-grained EEG signals of the theta band three channels of this invention;

[0017] Figure 6 This is a comparison diagram of the multi-scale coarse-grained EEG signals of the three channels in the alpha band of this invention;

[0018] Figure 7 This is a multi-scale coarse-grained comparison diagram of the three-channel EEG signal in the beta band of this invention;

[0019] Figure 8 This is a bar chart showing the frequency band entropy of each channel at a single scale according to the present invention;

[0020] Figure 9 This is a bar chart of the multi-scale C3 channel bandwidth entropy of the present invention;

[0021] Figure 10 This is a bar chart of the multi-scale C4 channel bandwidth entropy of the present invention; Figure 11This is a bar chart of the multi-scale Cz channel bandwidth entropy of the present invention;

[0022] Figure 12 This is a comprehensive curve of multi-scale, multi-channel frequency band entropy for this invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0024] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0025] like Figure 1 As shown, this invention provides a multi-domain fusion-based EEG signal complexity analysis algorithm. Taking EEG signals as the research object, it performs multi-band, multi-scale, and multi-channel three-level entropy analysis on C3, C4, and Cz channel signals, including EEG data loading and parameter initialization, signal extraction and frequency band filtering, a three-level entropy analysis module, and a visualization output module.

[0026] It should be noted that the EEG data loading and parameter initialization specifically involves loading the EEG data to be analyzed and extracting basic parameters such as sampling frequency, number of time points, number of channels, and channel names, which are used for subsequent filtering, scale division, entropy calculation, and other operations. Four standard frequency bands are set for the EEG signal: δ band (0.5-4Hz), θ band (4-8Hz), α band (8-13Hz), and β band (13-30Hz). The signal extraction and band filtering process specifically involves extracting the time-domain signals of three channels (C3, C4, and Cz) from the EEG data; performing bandpass filtering on each channel signal for four frequency bands, i.e., each channel × four frequency bands, resulting in 3 × 4 sets of frequency band signals, providing input data for subsequent three-level entropy analysis.

[0027] The three-level entropy analysis module includes single-scale single-channel analysis, multi-scale single-channel analysis, and multi-scale multi-channel joint analysis.

[0028] The single-scale single-channel analysis specifically involves analyzing the four frequency band signals of each channel at a single scale. , and Taking the frequency band as an example, let the length be... The time series of the three frequency bands As shown in the following formula:

[0029]

[0030] in, For the first Time points Frequency band time information, For the first Time points Frequency band time information, For the first Time points Frequency band timing information.

[0031] Utilizing predefined embedding dimensions and time delay Construct a time-frequency space containing multiple frequency bands: .

[0032]

[0033] in, , for The time series matrix after frequency band delay embedding.

[0034] The maximum absolute difference between corresponding frequency bands in paired time-frequency spaces is calculated as shown in the following formula:

[0035]

[0036] in, In the distance matrix above, the distance attributes of each frequency band are arranged in ascending order according to their size.

[0037] Each frequency band signal is symbolized to construct a four-band joint embedding vector, and the joint probability distribution and symbolization entropy value are calculated, as shown in the following formula:

[0038]

[0039] in, , Let be the probability of occurrence of various permutations and combinations of relative frequencies.

[0040]

[0041] in, Let be the probability of occurrence of various permutations and combinations of relative frequencies. This represents the total number of symbolic states.

[0042] The multi-scale single-channel analysis specifically involves analyzing the signal of each channel at different scale factors. Under the condition of segmented mean, a coarse-grained sequence is obtained, as shown in the following formula:

[0043]

[0044] in, As a scale factor, For the first A coarse-grained data point.

[0045] The four-band signals at each scale are symbolized separately, and the output channel scale entropy matrix is ​​finally obtained.

[0046] Multi-scale single-channel analysis specifically involves analyzing the signal of each channel at different scale factors. Under the condition of segmented mean, a coarse-grained sequence is obtained, as shown in the following formula:

[0047]

[0048] in, As a scale factor, For the first A coarse-grained data point.

[0049] The four-band signals at each scale are symbolized separately, and the output channel scale entropy matrix is ​​finally obtained.

[0050] The multi-scale, multi-channel analysis specifically involves performing joint coarse-grained processing on multiple channel signals to form a multivariate time series, and constructing a multi-channel joint embedding vector. Joint symbol mapping is performed, and the joint symbol probability distribution of four-band three-channel is calculated at multiple scales. Finally, a unified multi-scale entropy sequence across channels is output to characterize the changes in brain region coordination complexity.

[0051] The visualization output module specifically includes a single-scale channel frequency band entropy bar chart, which displays the entropy value comparison of different frequency bands in the three channels; a multi-scale channel frequency band entropy bar chart, which displays the entropy of each frequency band in the three channels as the scale changes; and a multi-scale multi-channel frequency band entropy composite curve, which shows the trend of joint entropy as the scale changes.

[0052] like Figure 2 As shown, this embodiment simulates and processes three-channel EEG signals, and respectively... Signals were extracted from four typical EEG bands. Three typical channels from the international 10-20 system were selected: C3 channel (left sensorimotor cortex), C4 channel (right sensorimotor cortex), and Cz channel (midline sensorimotor cortex). These channels are representative in studies of motor imagery, consciousness state analysis, and neural modulation. (See figure) As shown in the frequency bands, the three-channel signal exhibits a smooth waveform and a slow, gradual change at low frequencies, with energy mainly concentrated in the initial stage of the signal. This characteristic is consistent with... The physiological characteristics of waves: low frequency and high amplitude; The frequency band signal exhibits low-to-mid frequency oscillations with relatively large amplitudes. The wave decreases, and the signal morphology conforms to EEG characteristics related to cognition and memory. The frequency band signal exhibits stable mid-frequency oscillations with moderate amplitude. Each channel shows amplitude enhancement in local time periods. This characteristic is used to simulate brain electrical activity in resting or attention-regulated states. The frequency band signal exhibits significant high-frequency oscillations and dense waveforms, making it suitable for simulating EEG characteristics in motion-related or highly alert states.

[0053] like Figure 3 As shown, this embodiment provides a method for constructing a two-dimensional phase space of EEG signals based on time delay embedding theory. According to Takens' embedding theorem, for a one-dimensional time series... Embedding vectors can be constructed: ,in, For delay parameters, embedding dimension For any frequency band and any channel signal First, construct the time delay sequence. , forming two-dimensional coordinate points Then, all sampling points are traversed to obtain a set of two-dimensional phase space trajectory points. Finally, a gradient is drawn using time points as color mapping variables. (See figure.) As shown in the frequency band diagram, the three channels The frequency band phase space trajectory exhibits an approximately linear structure, with highly concentrated point distribution and the trajectory almost entirely along a diagonal, indicating that the low-frequency signal changes slowly and has low dynamic complexity. This characteristic suggests... The band exhibits strong periodicity and stability. The phase space trajectory in the lower frequency band still exhibits a diagonal structure and the point cloud distribution begins to slightly diffuse, indicating that the system dynamics are beginning to show slight nonlinear characteristics. The frequency band phase space distribution range has expanded, the point cloud exhibits a banded diffusion structure with obvious oscillating components, and dynamic changes have intensified. Waves play a key role in the regulation of brain function, and their phase space structure reflects moderately complex dynamic behavior. The frequency band phase space point cloud is significantly divergent and broadband, exhibiting enhanced randomness and a markedly increased dynamic complexity, indicating stronger nonlinearity and instability in high-frequency signals. Comparison of channels C3, C4, and Cz reveals that C3 and C4 exhibit relatively symmetrical phase space structures, while the Cz signal amplitude range is slightly smaller. Experimental results show that as the frequency band... arrive As the frequency and complexity of EEG signals increased, the phase space trajectory of the EEG signal gradually changed from a regular set to a complex diffusion, verifying the positive correlation between the frequency and complexity of EEG signals.

[0054] like Figure 4-7 As shown, this embodiment uses three channels: C3, C4, and Cz. Four frequency band signals were used as the research objects. Coarse-grained processing was performed at different scale factors, and the processing results were compared and analyzed. For the original time series:

[0055]

[0056] At a scale factor of At that time, construct coarse-grained sequences:

[0057]

[0058] in, As a scale factor, For the first A coarse-grained data point, The length of the original sequence.

[0059] As shown in the figure As shown in the frequency band, when the scaling factor is 1 or 50, the signals are all smooth monotonic curves with a slow changing trend, and the signal amplitude converges significantly. This result indicates that low-frequency EEG signals have high stability over long time scales. The frequency band exhibits periodic oscillations with relatively stable amplitude at scale=1, while the oscillation frequency decreases and amplitude fluctuations are compressed at scale=50, indicating that... The frequency band contains mid-to-low frequency components. After coarsening, short-time oscillation components are suppressed, and only long-time structures are retained. The frequency band exhibits a regular oscillating structure with local energy enhancement regions at scale=1. At scale=50, the oscillations significantly decrease, the energy concentration region is smoothed, and the signal tends to fluctuate towards lower frequencies, indicating... Waves lose their short-term rhythmicity on a large scale, retaining only macroscopic trends of change. The frequency band behavior is most pronounced. At scale=1, high-frequency oscillations are dense, with violent fluctuations and strong randomness. At scale=50, the high-frequency components disappear significantly, and the amplitude decreases significantly, indicating that... The frequency band contains a large number of short-term random disturbances, which are effectively filtered out after coarsening. Observation of the three channels C3, C4, and Cz reveals that the overall trends of the three channels are similar, with C3 and C4 having slightly larger amplitudes, while the overall amplitude of the Cz signal is smaller. The differences between channels gradually decrease at different scales. This result indicates that coarsening can enhance cross-channel consistency, which is beneficial for subsequent multi-channel joint analysis.

[0060] like Figure 8 As shown, this embodiment is based on four standard frequency bands ( Entropy values ​​were calculated and compared for the signal and its three channels (C3, C4, Cz). After amplitude normalization of the frequency band signal, a symbolic mapping method was used to convert the continuous time series into a discrete symbol sequence, and the probability of each state in the symbol sequence was statistically analyzed.

[0061]

[0062] in, For the first The number of times each symbol appears. This represents the total number of samples.

[0063] The frequency band entropy value is calculated using the Shannon entropy formula:

[0064]

[0065] in, For the number of symbol types, This is the frequency band entropy value. For the first The probability of each symbol appearing.

[0066] As can be observed from the figure The frequency band entropy value is the highest (approximately 0.82-0.87). The second highest bandwidth entropy value (approximately 0.73-0.83) The frequency band is in the middle (approximately 0.72-0.76). The lowest bandwidth (approximately 0.26-0.27) indicates that... Bandwidth signals have the highest complexity at a single scale. The frequency band signal is the most regular and has the lowest randomness. Comparing the three channels, we can see that C3, C4, and Cz have a consistent overall trend, with channel C4 showing the most regular trend. The frequency band is slightly lower than C3, and the Cz channel is in The frequency band is slightly higher. The frequency bands showed almost identical entropy values ​​across the three channels, indicating slight differences in complexity among different brain regions.

[0067] like Figure 9-11 As shown, this embodiment, based on three channels C3, C4, and Cz, performs [the following] under scale factors S=1, S=25, and S=50. The entropy values ​​of each of the four frequency bands were calculated and compared. The attached figures show bar charts of the single-band entropy values ​​of channels C3, C4, and Cz at different scales, with the horizontal axis representing the scale and the vertical axis representing the normalized entropy value. From the C3 channel graph, it can be observed that... The bandwidth entropy value is always the highest. The bandwidth increases significantly with increasing scale. The bandwidth reaches its peak at S=25. The frequency band entropy value is the lowest and basically stable; observation results of channel C4 show... The frequency band reaches its highest entropy value when S=50. The bandwidth increases significantly with increasing scale. The frequency band initially increased and then slightly decreased. The frequency band remains essentially unchanged; the entropy variation trend of the Cz channel at different scales is similar to that of the C3 and C4 channels. Experimental results show that under multi-scale conditions, Band complexity increases significantly with scale, while low frequency... The wave is basically stable, indicating that multi-scale entropy analysis can effectively characterize the temporal structure and frequency characteristics of EEG signals.

[0068] like Figure 12 As shown, this embodiment decomposes the EEG signals of three channels (C3, C4, and Cz) into frequency bands and calculates the entropy values ​​of each frequency band at multiple scales, thereby constructing a multi-scale, multi-channel frequency band entropy composite curve to characterize the variation characteristics of EEG signal complexity at different time scales. To obtain the overall brain region complexity characteristics, this invention comprehensively processes the entropy values ​​of the three channels and four frequency bands. Let:

[0069]

[0070] Indicated in scale Below: Passage ,frequency band The corresponding entropy value.

[0071] The overall entropy can be obtained through average fusion:

[0072]

[0073] This yields the multi-scale, multi-channel frequency band integrated entropy values ​​at each scale. As shown in the attached figures, the integrated entropy values ​​remain relatively stable at small scales (Scale=1-20), ranging from approximately 0.54 to 0.56, indicating that the complexity of the EEG signal changes little at short time scales, and the system's dynamic structure is relatively stable. At medium scales (Scale=20-30), the integrated entropy values ​​begin to gradually decrease, indicating that as the time scale increases, short-term random fluctuations are gradually smoothed out, and the system complexity begins to decrease. At large scales (Scale=30-50), the integrated entropy values ​​show a significant decreasing trend, gradually decreasing from approximately 0.48 to approximately 0.33. The experimental results demonstrate that at larger time scales, the macroscopic structure of the EEG signal gradually emerges, high-frequency random components are weakened, and the system complexity is significantly reduced.

[0074] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A complexity analysis algorithm for electroencephalogram (EEG) signals based on multi-domain fusion, characterized in that, The EEG signal complexity analysis algorithm based on multi-domain fusion includes: Data preprocessing to acquire multi-band, multi-channel EEG signals; Single-scale single-channel analysis is performed on the multi-band multi-channel EEG signals to output multi-band fuzzy entropy; Multi-scale single-channel analysis is performed on the multi-band, multi-channel EEG signals to output a multi-scale fuzzy entropy vector. Multi-scale multi-channel analysis is performed on the multi-band multi-channel EEG signals to output multi-channel multi-scale fuzzy entropy.

2. The EEG signal complexity analysis algorithm based on multi-domain fusion according to claim 1, characterized in that, The data preprocessing specifically includes: Load raw EEG data and extract its key parameters, including sampling frequency, total number of channels and information at each time point, and complete the setting of relevant parameters for the four standard frequency bands δ / θ / α / β; Extract the EEG signals from the three specified channels C3, Cz, and C4 from the raw data; For each channel, bandpass filtering of the four frequency bands mentioned above is performed separately, resulting in four sets of three-channel frequency band EEG signals, each set containing data from three channels.

3. The EEG signal complexity analysis algorithm based on multi-domain fusion according to claim 2, characterized in that, The single-scale, single-channel analysis specifically refers to: The four sets of three-channel frequency band EEG signals are input into the time series delay embedding module for delay processing to obtain the phase space reconstruction matrix. The reconstructed matrix is ​​input into the Chebyshev distance module that calculates the vector pair distances within the frequency band, and the vector pair distances in the phase space are calculated to obtain the rhythm distance matrix for each frequency band. The rhythm distance matrix of each frequency band is sent to the multi-frequency band distance sorting and symbolization module, and the sorting index and symbolization matrix are obtained through symbolization processing; The sorting index and symbolic matrix are input into the statistical permutation pattern occurrence frequency module. After the permutation pattern comparison operation, the matching permutation patterns and their probability distributions are obtained. The above arrangement pattern and probability distribution are input into the Shannon entropy and maximum entropy calculation module. The entropy value is calculated and normalized by the Shannon entropy and maximum entropy calculation module, and finally the multi-band fuzzy entropy is output.

4. The EEG signal complexity analysis algorithm based on multi-domain fusion according to claim 3, characterized in that, The entropy calculation and normalization process using the Shannon entropy and maximum entropy modules is performed as follows: in, Let be the probability of occurrence of various permutations and combinations of relative frequencies. Let be the probability of occurrence of various permutations and combinations of relative frequencies. The total number of symbolic states. .

5. The EEG signal complexity analysis algorithm based on multi-domain fusion according to claim 2, characterized in that, The multi-scale single-channel analysis specifically refers to: The four sets of three-channel frequency band EEG signals are input into the data multi-scale coarse-grained processing module, which then performs non-overlapping window averaging to output the coarse-grained time series. The coarse-grained time series is input into the multi-band fuzzy entropy module at different scales. The multi-band fuzzy entropy module calculates the entropy value at each scale and finally outputs the multi-scale fuzzy entropy vector.

6. The EEG signal complexity analysis algorithm based on multi-domain fusion according to claim 5, characterized in that, The non-overlapping window averaging process is specifically implemented as follows: in, As a scale factor, For the first A coarse-grained data point.

7. The EEG signal complexity analysis algorithm based on multi-domain fusion according to claim 2, characterized in that, The multi-scale, multi-channel analysis specifically refers to: The four sets of three-channel frequency band EEG signals are input into the multivariate time series coarse-grained processing module, and the coarse-grained data matrix is ​​output through multi-channel non-overlapping window averaging operation. The data matrix is ​​input into the multivariate time delay embedding module, and after time delay processing, a multivariate embedding matrix is ​​obtained. Then, the multivariate embedding matrix is ​​input into the inter-vector Chebyshev distance module, and the distance vectors of each frequency band are generated by calculating the pairwise distances between vectors. The distance vectors of each frequency band are input into the distance sorting and symbolization module at different scales, and the sorted symbol matrix is ​​obtained after symbolization processing. Input the symbol matrix into the statistical permutation module at different scales, and obtain the probability distribution of each permutation pattern through comparison and matching operations; Input the probability distribution of each permutation into the module for calculating the Shannon entropy and normalized entropy value of each permutation distribution. Perform entropy value calculation and normalization processing through the module for calculating the Shannon entropy and normalized entropy value of each permutation distribution, and output multi-channel multi-scale fuzzy entropy.

8. The EEG signal complexity analysis algorithm based on multi-domain fusion according to claim 7, characterized in that, The entropy calculation and normalization process, which involves calculating the Shannon entropy and normalized entropy values ​​for each permutation, is performed in the following manner: in, As a scale, For the channel, In scale Below, passage ,frequency band The corresponding entropy value, This is the comprehensive entropy value.

Images