A multi-dimensional EEG signal recognition method

Abstract

The present invention relates to the technical field of electroencephalogram signal recognition, and discloses a multi-dimensional electroencephalogram signal recognition method, the method comprising: collecting electroencephalogram signals at different positions, and adding resonance information to the electroencephalogram signals for random resonance processing; constructing an electroencephalogram signal resonance elimination model and using a training data set to train model parameters; collecting original electroencephalogram signals to be identified, and using the trained electroencephalogram signal resonance elimination model to perform resonance elimination on the original electroencephalogram signals; and constructing a multi-channel electroencephalogram signal recognition model to recognize the original electroencephalogram signals after resonance elimination. The present invention adds resonance information to the collected electroencephalogram signals to obtain a training data set of a non-stable electroencephalogram signal with enhanced randomness, enhances the robustness of the trained model in performing resonance elimination on the electroencephalogram signals, combines the frequency domain power spectrum information and the time domain information of the signals for signal filtering, and combines the feature decomposition information and the neighboring signal value information extraction method for signal recognition.

Description

Technical Field

[0001] The present invention relates to the technical field of electroencephalogram (EEG) signal recognition, and in particular to a multi-dimensional EEG signal recognition method. Background Art

[0002] With the continuous development of neuroscience, artificial intelligence and biomedical engineering, the recognition and application of EEG signals has become one of the research hotspots. EEG signals are weak electrical signals recorded on the surface of human scalp, which can reflect the activity of brain neurons. Through the analysis and recognition of EEG signals, various applications such as brain-computer interface, brain disease diagnosis, and psychological state monitoring can be realized. However, EEG signals themselves have the characteristics of low signal-to-noise ratio and instability, which brings certain challenges to their recognition and application. At the same time, the recognition method of a single EEG signal has certain limitations. There is a random resonance phenomenon in the process of simultaneous EEG signal collection by multiple electrodes, which increases the difficulty of accurate EEG signal recognition. In response to this problem, the present invention proposes a multidimensional EEG signal recognition method, which improves the accuracy and stability of recognition through random resonance elimination and multidimensional signal analysis, and expands the application scope of brain-computer interface technology. Summary of the invention

[0003] In view of this, the present invention provides a multi-dimensional EEG signal recognition method, the purpose of which is to: 1) collect EEG signals and add resonance information to the collected EEG signals to obtain unstable EEG signals with enhanced randomness, and form a training data set to train an EEG signal resonance elimination model, so that the model can effectively eliminate the resonance of EEG signals collected in various scenarios, and enhance the robustness of the model in eliminating the resonance of EEG signals. In the training process, a whale position iteration strategy combined with whale predation behavior is used to perform random search and global search model parameter iteration, and the Cauchy perturbation method is used to jump out of the local optimal solution, quickly solve the optimal solution, and construct a usable The EEG signal resonance elimination model is used; 2) The EEG signal resonance elimination model is used to perform time convolution processing combined with the power spectrum on the original EEG signal to be identified, thereby combining the frequency domain power spectrum information and the time domain information of the signal to filter the noise signal and the interference signal to obtain a pure original EEG signal, and combining the feature decomposition and the neighboring signal value information extraction method to construct the spatial and temporal characteristics of the original EEG signal, and obtain the probability of the original EEG signal under different labels as the label information of the original EEG signal, that is, the recognition result of the original EEG signal, to achieve the removal of interference information in the EEG signal and the recognition of EEG signal labels combined with the time and space dimension information.

[0004] To achieve the above object, the present invention provides a multi-dimensional EEG signal recognition method, comprising the following steps:

[0005] S1: Collecting EEG signals at different positions, and adding resonance information to the EEG signals for random resonance processing to obtain EEG signals after resonance enhancement, and the collected EEG signals and the EEG signals after resonance enhancement constitute a training data set, wherein the generalized Gaussian signal enhancement method based on signal value is the main implementation method of the resonance processing;

[0006] S2: Construct an EEG signal resonance elimination model and use the training data set to train the model parameters. The EEG signal resonance elimination model takes the original EEG signal as input and the original EEG signal after resonance elimination as output. The time convolutional neural network integrating the instantaneous prior power spectrum is the main implementation method of the resonance elimination model, and the whale optimization algorithm based on Cauchy variation perturbation is the main implementation method of the model parameter training.

[0007] S3: collecting the original EEG signal to be identified, and using the trained EEG signal resonance elimination model to eliminate the resonance of the original EEG signal to obtain the original EEG signal after the resonance is eliminated;

[0008] S4: Construct a multi-channel EEG signal recognition model to recognize the original EEG signal after the resonance is eliminated, and obtain the recognition result of the original EEG signal.

[0009] As a further improvement method of the present invention:

[0010] Optionally, in step S1, collecting original EEG signals at different positions and adding resonance information to the original EEG signals for stochastic resonance processing include:

[0011] The electrodes are placed on different areas of the head of the EEG signal collector, and the acquisition board connected to the electrodes is used to record the potential changes in the brain in different areas as EEG signals. A total of N groups of original EEG signals are collected, where the collection of the original EEG signals is expressed as:

[0012] ;

[0013] in:

[0014] represents the nth group of EEG signals collected, and t represents the time sequence information;

[0015] Represents the original EEG signal The signal values ​​corresponding to the L signal sampling moments in , Represents the original EEG signal The signal value corresponding to the i-th signal sampling moment in , , the time interval corresponding to adjacent signal sampling times is ;

[0016] Resonance information is added to the original EEG signal for random resonance processing to obtain the original EEG signal after resonance enhancement. The stochastic resonance processing flow is:

[0017] S11: Combining raw EEG signals , construct the generalized Gaussian signal probability density function based on the signal value:

[0018] ;

[0019] ;

[0020] ;

[0021] in:

[0022] represents the constructed generalized Gaussian signal probability density function, and x represents the signal value of the input generalized Gaussian signal probability density function;

[0023] represents the preset standard deviation, represents the adjustment parameter, ;

[0024] Represents the original EEG signal The mean of the signal values ​​in ;

[0025] represents the gamma function;

[0026] represents an exponential function with a natural constant as base;

[0027] S12: Calculate the original EEG signal The resonance probability of any signal value in , where the original EEG signal The signal value corresponding to the i-th signal sampling time in The resonance probability is:

[0028] ;

[0029] ;

[0030] in:

[0031] Represents the original EEG signal The signal value corresponding to the i-th signal sampling time in The resonance probability of

[0032] represents the probability density function of the generalized Gaussian signal The corresponding generalized Gaussian signal probability distribution value;

[0033] S13: Calculate the original EEG signal The resonance range of any signal value in , where the original EEG signal The signal value corresponding to the i-th signal sampling time The resonance range is:

[0034] ;

[0035] in:

[0036] Represents the original EEG signal The signal value corresponding to the i-th signal sampling time The resonance range of

[0037] Represents the original EEG signal Any signal value and mean The difference sequence of

[0038] Represents a difference sequence The mean of

[0039] Represents a difference sequence The standard deviation of

[0040] S14: The original EEG signal after resonance enhancement :

[0041] ;

[0042] in:

[0043] Represents the original EEG signal The resonance enhancement results;

[0044] Indicates the signal value The stochastic resonance processing result, where the stochastic resonance processing rule is: with the resonance probability For signal value Perform stochastic resonance processing to The probability of the signal value No stochastic resonance processing is performed; if the signal value Perform random resonance processing, then within the resonance range Randomly select any value and superimpose the selected value on the signal value The signal value The results of stochastic resonance processing Otherwise, the signal value is directly As a result of stochastic resonance .

[0045] Optionally, the step of forming a training data set from the collected original EEG signals and the original EEG signals after resonance enhancement includes:

[0046] The collected original EEG signals and the original EEG signals after resonance enhancement constitute a training data set, where the training data set is represented as follows:

[0047] ;

[0048] in:

[0049] Represents the constructed training data set;

[0050] Represents the nth group of training data in the training data set data, represents the nth group of EEG signals collected, Represents the original EEG signal The resonance enhancement result.

[0051] Optionally, constructing the EEG signal resonance elimination model in step S2 includes:

[0052] Constructing an EEG signal resonance elimination model, wherein the EEG signal resonance elimination model takes the original EEG signal as input and takes the original EEG signal after resonance elimination as output, wherein a temporal convolutional neural network integrating an instantaneous prior power spectrum is the main implementation method of the resonance elimination model, and a whale optimization algorithm based on Cauchy variation perturbation is the main implementation method of the model parameter training;

[0053] The EEG signal resonance elimination model includes an input layer, a power spectrum information extraction layer, a time convolution layer and an output layer. The input layer is used to receive the original EEG signal, the power spectrum information extraction layer is used to extract the power spectrum information of the original EEG signal, the time convolution layer is used to perform time series convolution processing on the original EEG signal based on the power spectrum information to obtain the original EEG signal after the resonance is eliminated, and the output layer is used to output the original EEG signal after the resonance is eliminated;

[0054] Using the training dataset The model parameters of the EEG signal resonance elimination model are trained.

[0055] Optionally, the use of training data sets The model parameters of the EEG signal resonance elimination model are trained, including:

[0056] The model parameter training process based on the training data set data is:

[0057] S21: Construct the model parameter training objective function based on the training data set data:

[0058] ;

[0059] in:

[0060] represents the L2 norm;

[0061] represents the model parameter training objective function, Represents the model parameters to be trained;

[0062] Represents model parameters based on The constructed EEG signal resonance elimination model;

[0063] Indicates that the signal Input into the EEG signal resonance elimination model, and the model outputs the signal after the resonance is eliminated;

[0064] S22: Initialize and generate U groups of whale positions. Each group of whale positions corresponds to a model parameter. The uth group of whale positions generated by initialization is ;

[0065] S23: Set the current iteration number of the whale position to c, and the maximum iteration number to C. Then the result of the cth iteration of the uth group of whale positions is , the initial value of c is 0;

[0066] S24: The U group of whale positions obtained in the cth iteration are used as the model parameters in the model parameter training objective function, and the output result of the model parameter training objective function is used as the fitness function value of the whale position, then the whale position The fitness function value is ; Select the whale position with the lowest fitness function value among the U group of whale positions obtained in the cth iteration as the optimal whale position for the cth iteration ;

[0067] S25: Set the current predation behavior coefficient of the whale position, where the whale position The current predation behavior coefficient of is:

[0068] ;

[0069] in:

[0070] Indicates the location of the whale The current predation behavior coefficient;

[0071] Represents a random number between 0 and 1;

[0072] S26: Iterate the whale position based on the current predation behavior coefficient of the whale position, where the whale position The iteration formula is:

[0073] ;

[0074] in:

[0075] Represents a random number between -1 and 1;

[0076] represents any whale position in the U group of whale positions obtained in the cth iteration;

[0077] S27: Cauchy mutation perturbation is performed on the iterative result of the whale position, where the whale position The Cauchy variation perturbation formula is:

[0078]

[0079] ;

[0080] in:

[0081] It means element-by-element addition;

[0082] represents a random perturbation sequence that conforms to the standard Cauchy distribution;

[0083] S28: Let c=c+1, return to step S24, until the maximum number of iterations C is reached, and use the U group of whale positions obtained in the Cth iteration as the model parameters in the model parameter training objective function, and select the whale position with the lowest fitness function value as the model parameter training result of the EEG signal resonance elimination model .

[0084] Optionally, in step S3, the original EEG signal to be identified is collected, and the EEG signal resonance elimination model obtained by training is used to eliminate the resonance of the original EEG signal, including:

[0085] Collect the original EEG signal to be identified ,in Represents the original EEG signal The signal value corresponding to the i-th signal sampling moment in the training is used to eliminate the original EEG signal using the EEG signal resonance elimination model. Resonance elimination is performed, where the original EEG signal The resonance elimination process is:

[0086] S31: The input layer receives the original EEG signal ;

[0087] S32: Power spectrum information extraction layer extracts raw EEG signals Power spectrum information:

[0088] ;

[0089] in:

[0090] Represents the original EEG signal The power spectrum information of

[0091] j represents the imaginary unit, ;

[0092] Represents the original EEG signal frequency;

[0093] S33: Temporal convolution layer based on power spectrum information Raw EEG signals Perform time series convolution processing to obtain the original EEG signal after resonance elimination :

[0094] ;

[0095] ;

[0096] in:

[0097] represents the convolution parameter matrix, where ;

[0098] Represents convolution calculation;

[0099] Represents the original EEG signal Signal value A signal value sequence centered at and having a length of Len; in the embodiment of the present invention, Len is set to 5.

[0100] Optionally, constructing a multi-channel EEG signal recognition model in step S4 includes:

[0101] A multi-channel EEG signal recognition model is constructed, and the original EEG signal after the resonance is eliminated is recognized by using the multi-channel EEG signal recognition model to obtain the recognition result of the original EEG signal;

[0102] The multi-channel EEG signal recognition model includes an input layer, a spatiotemporal information extraction layer, a spatiotemporal information fusion layer and an output layer. The input layer is used to receive the original EEG signal after resonance elimination. The spatiotemporal information extraction layer is used to extract the time characteristics and spatial characteristics of the signal respectively. The spatiotemporal information fusion layer is used to fuse the time characteristics and the spatial characteristics to obtain fused characteristics. The output layer is used to convert the fused characteristics into label information of the original EEG signal as the recognition result of the original EEG signal.

[0103] Optionally, the identifying the original EEG signal after the resonance is eliminated by using a multi-channel EEG signal identification model includes:

[0104] The original EEG signal after the resonance is eliminated is identified by using a multi-channel EEG signal recognition model. The identification process is:

[0105] S41: The input layer receives the original EEG signal after resonance elimination ;

[0106] S42: The spatiotemporal information extraction layer extracts signals respectively Temporal and spatial characteristics:

[0107] ;

[0108] ;

[0109] ;

[0110] ;

[0111] ;

[0112] in:

[0113] Indicates signal The time characteristics of Indicates the signal value Time information;

[0114] Indicates signal spatial characteristics;

[0115] Indicates signal The covariance matrix of , T represents the transpose;

[0116] Represents the covariance matrix The eigenvector corresponding to the kth largest eigenvalue obtained by eigendecomposition ;

[0117] K represents the dimension of spatial features;

[0118] S43: Fusion of temporal features And spatial characteristics , get the fusion features :

[0119] ;

[0120] in:

[0121] Representing time characteristics The standard deviation of

[0122] S44: The output layer will fuse features Convert to raw EEG signal The label information is used as the original EEG signal The recognition result, where the original EEG signal The label information is:

[0123] ;

[0124] in:

[0125] Represents the original EEG signal The probability of label being label, label represents any label in the label set, Represents the mapping matrix corresponding to the label label;

[0126] The original EEG signal The probability of different labels as the original EEG signal The label information, that is, the original EEG signal The recognition result.

[0127] In order to solve the above problem, the present invention provides an electronic device, the electronic device comprising:

[0128] A memory storing at least one instruction;

[0129] Communication interface, enabling electronic equipment to communicate; and

[0130] The processor executes the instructions stored in the memory to implement the multi-dimensional EEG signal recognition method described above.

[0131] In order to solve the above problem, the present invention also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the above-mentioned multi-dimensional EEG signal recognition method.

[0132] Compared with the prior art, the present invention proposes a multi-dimensional EEG signal recognition method, which has the following advantages:

[0133] First, this scheme proposes a randomness-enhanced training data construction method and a model training method. The model parameter training process based on the training data set data is as follows: Construct a model parameter training objective function based on the training data set data:

[0134] ;

[0135] in: represents the L2 norm; represents the model parameter training objective function, Represents the model parameters to be trained; Represents model parameters based on The constructed EEG signal resonance elimination model; Indicates that the signal Input into the EEG signal resonance elimination model, the model outputs the signal after resonance elimination; initialize and generate U groups of whale positions, each group of whale positions corresponds to a model parameter, then the u-th group of whale positions generated by initialization is ; Set the current iteration number of the whale position to c and the maximum iteration number to C, then the result of the cth iteration of the uth group of whale positions is , the initial value of c is 0; the U group of whale positions obtained in the cth iteration are used as the model parameters in the model parameter training objective function, and the output result of the model parameter training objective function is used as the fitness function value of the whale position, then the whale position The fitness function value is ; Select the whale position with the lowest fitness function value among the U group of whale positions obtained in the cth iteration as the optimal whale position for the cth iteration ; Set the current predation behavior coefficient of the whale position, where the whale position The current predation behavior coefficient of is:

[0136] ;

[0137] in: Indicates the location of the whale The current predation behavior coefficient; Represents a random number between 0 and 1; iterates the whale position based on the current predation behavior coefficient of the whale position, where the whale position The iteration formula is:

[0138] ;

[0139] in: Represents a random number between -1 and 1; represents any whale position in the U group of whale positions obtained in the cth iteration; the iterative result of the whale position is perturbed by Cauchy mutation, where the whale position The Cauchy variation perturbation formula is:

[0140] ;

[0141] in: It means element-by-element addition; represents a random perturbation sequence that conforms to the standard Cauchy distribution; let c=c+1 until the maximum number of iterations C is reached, and the U group of whale positions obtained in the Cth iteration are used as the model parameters in the model parameter training objective function, and the whale position with the lowest fitness function value is selected as the model parameter training result of the EEG signal resonance elimination model . This scheme collects EEG signals and adds resonance information to the collected EEG signals to obtain unstable EEG signals with enhanced randomness, which constitute a training data set to train the EEG signal resonance elimination model, so that the model can effectively eliminate the resonance of the EEG signals collected in various scenarios, and enhance the robustness of the model in eliminating the resonance of EEG signals. During the training process, the whale position iteration strategy combined with the whale's predation behavior is used to perform random search and global search model parameter iteration, and the Cauchy perturbation method is used to jump out of the local optimal solution, quickly solve the optimal solution, and construct a usable EEG signal resonance elimination model.

[0142] At the same time, this scheme proposes a resonance elimination method combined with power spectrum information and an EEG signal recognition method to collect the original EEG signal to be recognized. ,in Represents the original EEG signal The signal value corresponding to the i-th signal sampling moment in the training is used to eliminate the original EEG signal using the EEG signal resonance elimination model. Resonance elimination is performed, where the original EEG signal The resonance elimination process is as follows: the input layer receives the original EEG signal ; The power spectrum information extraction layer extracts the original EEG signal Power spectrum information:

[0143] ;

[0144] in: Represents the original EEG signal The power spectrum information of ; Represents the original EEG signal The frequency of the temporal convolution layer is based on the power spectrum information Raw EEG signals Perform time series convolution processing to obtain the original EEG signal after resonance elimination :

[0145] ;

[0146] ;

[0147] in: represents the convolution parameter matrix, where ; Represents convolution calculation; Represents the original EEG signal Signal value A signal value sequence with Len as the center and length as Len; a multi-channel EEG signal recognition model is constructed, and the original EEG signal after resonance elimination is recognized by the multi-channel EEG signal recognition model to obtain the recognition result of the original EEG signal; the multi-channel EEG signal recognition model includes an input layer, a spatiotemporal information extraction layer, a spatiotemporal information fusion layer and an output layer, the input layer is used to receive the original EEG signal after resonance elimination, the spatiotemporal information extraction layer is used to extract the time feature and the space feature of the signal respectively, the spatiotemporal information fusion layer is used to fuse the time feature and the space feature to obtain the fusion feature, and the output layer is used to convert the fusion feature into the label information of the original EEG signal as the recognition result of the original EEG signal. This scheme uses the EEG signal resonance elimination model to perform time convolution processing combined with the power spectrum on the original EEG signal to be identified, thereby combining the frequency domain power spectrum information and time domain information of the signal to filter the noise signal and the interference signal to obtain a pure original EEG signal. In combination with feature decomposition and neighboring signal value information extraction methods, the spatial and temporal characteristics of the original EEG signal are constructed to obtain the probability of the original EEG signal under different labels as the label information of the original EEG signal, that is, the recognition result of the original EEG signal, thereby realizing the removal of interference information in the EEG signal and the recognition of EEG signal labels combined with time and space dimension information. BRIEF DESCRIPTION OF THE DRAWINGS

[0148] Figure 1 A schematic diagram of a flow chart of a multi-dimensional EEG signal recognition method provided by an embodiment of the present invention;

[0149] Figure 2 A schematic diagram of the structure of an electronic device for implementing a multi-dimensional EEG signal recognition method provided by an embodiment of the present invention.

[0150] In the figure: 1 electronic device, 10 processor, 11 memory, 12 program, 13 communication interface.

[0151] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0152] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0153] The embodiment of the present application provides a multi-dimensional EEG signal recognition method. The execution subject of the multi-dimensional EEG signal recognition method includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present application. In other words, the multi-dimensional EEG signal recognition method can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0154] Embodiment 1:

[0155] S1: Collect EEG signals at different positions, add resonance information to the EEG signals for random resonance processing, obtain EEG signals after resonance enhancement, and form a training data set with the collected EEG signals and the EEG signals after resonance enhancement.

[0156] In the step S1, EEG signals at different positions are collected, and resonance information is added to the EEG signals for stochastic resonance processing, including:

[0157] The electrodes are placed on different areas of the head of the EEG signal collector, and the acquisition board connected to the electrodes is used to record the potential changes in the brain in different areas as EEG signals. A total of N groups of original EEG signals are collected, where the collection of the original EEG signals is expressed as:

[0158] ;

[0159] in:

[0160] represents the nth group of EEG signals collected, and t represents the time sequence information;

[0161] Represents the original EEG signal The signal values ​​corresponding to the L signal sampling moments in , Represents the original EEG signal The signal value corresponding to the i-th signal sampling moment in , , the time interval corresponding to adjacent signal sampling times is ;

[0162] Resonance information is added to the original EEG signal for random resonance processing to obtain the original EEG signal after resonance enhancement. The stochastic resonance processing flow is:

[0163] S11: Combining raw EEG signals , construct the generalized Gaussian signal probability density function based on the signal value:

[0164] ;

[0165] ;

[0166] ;

[0167] in:

[0168] represents the constructed generalized Gaussian signal probability density function, and x represents the signal value of the input generalized Gaussian signal probability density function;

[0169] represents the preset standard deviation, represents the adjustment parameter, ;

[0170] Represents the original EEG signal The mean of the signal values ​​in ;

[0171] represents the gamma function;

[0172] represents an exponential function with a natural constant as base;

[0173] S12: Calculate the original EEG signal The resonance probability of any signal value in , where the original EEG signal The signal value corresponding to the i-th signal sampling time in The resonance probability is:

[0174] ;

[0175] ;

[0176] in:

[0177] Represents the original EEG signal The signal value corresponding to the i-th signal sampling time in The resonance probability of

[0178] represents the probability density function of the generalized Gaussian signal The corresponding generalized Gaussian signal probability distribution value;

[0179] S13: Calculate the original EEG signal The resonance range of any signal value in , where the original EEG signal The signal value corresponding to the i-th signal sampling time The resonance range is:

[0180] ;

[0181] in:

[0182] Represents the original EEG signal The signal value corresponding to the i-th signal sampling time The resonance range of

[0183] Represents the original EEG signal Any signal value and mean The difference sequence of

[0184] Represents a difference sequence The mean of

[0185] Represents a difference sequence The standard deviation of

[0186] S14: The original EEG signal after resonance enhancement :

[0187] ;

[0188] in:

[0189] Represents the original EEG signal The resonance enhancement results;

[0190] Indicates the signal value The stochastic resonance processing result, where the stochastic resonance processing rule is: with the resonance probability For signal value Perform stochastic resonance processing to The probability of the signal value No stochastic resonance processing is performed; if the signal value Perform random resonance processing, then within the resonance range Randomly select any value and superimpose the selected value on the signal value The signal value The results of stochastic resonance processing Otherwise, the signal value is directly As a result of stochastic resonance .

[0191] The collected original EEG signals and the original EEG signals after resonance enhancement are used to form a training data set, including:

[0192] The collected original EEG signals and the original EEG signals after resonance enhancement constitute a training data set, where the training data set is represented as follows:

[0193] ;

[0194] in:

[0195] Represents the constructed training data set;

[0196] Represents the nth group of training data in the training data set data, represents the nth group of EEG signals collected, Represents the original EEG signal The resonance enhancement result.

[0197] S2: constructing an EEG signal resonance elimination model and using a training data set to train model parameters, wherein the EEG signal resonance elimination model takes the original EEG signal as input and takes the original EEG signal after resonance elimination as output.

[0198] The step S2 constructs a brain wave signal resonance elimination model, including:

[0199] Constructing an EEG signal resonance elimination model, wherein the EEG signal resonance elimination model takes the original EEG signal as input and takes the original EEG signal after resonance elimination as output, wherein a temporal convolutional neural network integrating an instantaneous prior power spectrum is the main implementation method of the resonance elimination model, and a whale optimization algorithm based on Cauchy variation perturbation is the main implementation method of the model parameter training;

[0200] The EEG signal resonance elimination model includes an input layer, a power spectrum information extraction layer, a time convolution layer and an output layer. The input layer is used to receive the original EEG signal, the power spectrum information extraction layer is used to extract the power spectrum information of the original EEG signal, the time convolution layer is used to perform time series convolution processing on the original EEG signal based on the power spectrum information to obtain the original EEG signal after the resonance is eliminated, and the output layer is used to output the original EEG signal after the resonance is eliminated;

[0201] Using the training dataset The model parameters of the EEG signal resonance elimination model are trained.

[0202] The use of training data sets The model parameters of the EEG signal resonance elimination model are trained, including:

[0203] The model parameter training process based on the training data set data is:

[0204] S21: Construct the model parameter training objective function based on the training data set data:

[0205] ;

[0206] in:

[0207] represents the L2 norm;

[0208] represents the model parameter training objective function, Represents the model parameters to be trained;

[0209] Represents model parameters based on The constructed EEG signal resonance elimination model;

[0210] Indicates that the signal Input into the EEG signal resonance elimination model, and the model outputs the signal after the resonance is eliminated;

[0211] S22: Initialize and generate U groups of whale positions. Each group of whale positions corresponds to a model parameter. The uth group of whale positions generated by initialization is ;

[0212] S23: Set the current iteration number of the whale position to c, and the maximum iteration number to C. Then the result of the cth iteration of the uth group of whale positions is , the initial value of c is 0;

[0213] S24: The U group of whale positions obtained in the cth iteration are used as the model parameters in the model parameter training objective function, and the output result of the model parameter training objective function is used as the fitness function value of the whale position, then the whale position The fitness function value is ; Select the whale position with the lowest fitness function value among the U group of whale positions obtained in the cth iteration as the optimal whale position for the cth iteration ;

[0214] S25: Set the current predation behavior coefficient of the whale position, where the whale position The current predation behavior coefficient of is:

[0215] ;

[0216] in:

[0217] Indicates the location of the whale The current predation behavior coefficient;

[0218] Represents a random number between 0 and 1;

[0219] S26: Iterate the whale position based on the current predation behavior coefficient of the whale position, where the whale position The iteration formula is:

[0220] ;

[0221] in:

[0222] Represents a random number between -1 and 1;

[0223] represents any whale position in the U group of whale positions obtained in the cth iteration;

[0224] S27: Cauchy mutation perturbation is performed on the iterative result of the whale position, where the whale position The Cauchy variation perturbation formula is:

[0225] ;

[0226] in:

[0227] It means element-by-element addition;

[0228] represents a random perturbation sequence that conforms to the standard Cauchy distribution;

[0229] S28: Let c=c+1, return to step S24, until the maximum number of iterations C is reached, and use the U group of whale positions obtained in the Cth iteration as the model parameters in the model parameter training objective function, and select the whale position with the lowest fitness function value as the model parameter training result of the EEG signal resonance elimination model .

[0230] S3: Collect the original EEG signal to be identified, and use the trained EEG signal resonance elimination model to eliminate the resonance of the original EEG signal to obtain the original EEG signal after the resonance is eliminated.

[0231] In the step S3, the original EEG signal to be identified is collected, and the EEG signal resonance elimination model obtained by training is used to eliminate the resonance of the original EEG signal, including:

[0232] Collect the original EEG signal to be identified ,in Represents the original EEG signal The signal value corresponding to the i-th signal sampling moment in the training is used to eliminate the original EEG signal using the EEG signal resonance elimination model. Resonance elimination is performed, where the original EEG signal The resonance elimination process is:

[0233] S31: The input layer receives the original EEG signal ;

[0234] S32: Power spectrum information extraction layer extracts raw EEG signals Power spectrum information:

[0235] ;

[0236] in:

[0237] Represents the original EEG signal The power spectrum information of

[0238] j represents the imaginary unit, ;

[0239] Represents the original EEG signal frequency;

[0240] S33: Temporal convolution layer based on power spectrum information Raw EEG signals Perform time series convolution processing to obtain the original EEG signal after resonance elimination :

[0241] ;

[0242] ;

[0243] in:

[0244] represents the convolution parameter matrix, where ;

[0245] Represents convolution calculation;

[0246] Represents the original EEG signal Signal value A signal value sequence centered at and having a length of Len; in the embodiment of the present invention, Len is set to 5.

[0247] S4: Construct a multi-channel EEG signal recognition model to recognize the original EEG signal after the resonance is eliminated, and obtain the recognition result of the original EEG signal.

[0248] The multi-channel EEG signal recognition model is constructed in step S4, including:

[0249] A multi-channel EEG signal recognition model is constructed, and the original EEG signal after the resonance is eliminated is recognized by using the multi-channel EEG signal recognition model to obtain the recognition result of the original EEG signal;

[0250] The multi-channel EEG signal recognition model includes an input layer, a spatiotemporal information extraction layer, a spatiotemporal information fusion layer and an output layer. The input layer is used to receive the original EEG signal after resonance elimination. The spatiotemporal information extraction layer is used to extract the time characteristics and spatial characteristics of the signal respectively. The spatiotemporal information fusion layer is used to fuse the time characteristics and the spatial characteristics to obtain fused characteristics. The output layer is used to convert the fused characteristics into label information of the original EEG signal as the recognition result of the original EEG signal.

[0251] The method of using a multi-channel EEG signal recognition model to recognize the original EEG signal after the resonance is eliminated includes:

[0252] The original EEG signal after the resonance is eliminated is identified by using a multi-channel EEG signal recognition model. The identification process is:

[0253] S41: The input layer receives the original EEG signal after resonance elimination ;

[0254] S42: The spatiotemporal information extraction layer extracts signals respectively Temporal and spatial characteristics:

[0255] ;

[0256] ;

[0257] ;

[0258] ;

[0259] ;

[0260] in:

[0261] Indicates signal The time characteristics of Indicates the signal value Time information;

[0262] Indicates signal spatial characteristics;

[0263] Indicates signal The covariance matrix of , T represents the transpose;

[0264] Represents the covariance matrix The eigenvector corresponding to the kth largest eigenvalue obtained by eigendecomposition ;

[0265] K represents the dimension of spatial features;

[0266] S43: Fusion of temporal features And spatial characteristics , get the fusion features :

[0267] ;

[0268] in:

[0269] Representing time characteristics The standard deviation of

[0270] S44: The output layer will fuse features Convert to raw EEG signal The label information is used as the original EEG signal The recognition result, where the original EEG signal The label information is:

[0271] ;

[0272] in:

[0273] Represents the original EEG signal The probability of label being label, label represents any label in the label set, Represents the mapping matrix corresponding to the label label;

[0274] The original EEG signal The probability of different labels as the original EEG signal The label information, that is, the original EEG signal The recognition result.

[0275] Embodiment 2:

[0276] like Figure 2 , which is a schematic diagram of the structure of an electronic device for implementing a multi-dimensional EEG signal recognition method provided by an embodiment of the present invention.

[0277] The electronic device 1 may include a processor 10 , a memory 11 , a communication interface 13 and a bus, and may also include a computer program stored in the memory 11 and executable on the processor 10 , such as a program 12 .

[0278] The memory 11 includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a mobile hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a magnetic memory, a disk, an optical disk, etc. In some embodiments, the memory 11 may be an internal storage unit of the electronic device 1, such as a mobile hard disk of the electronic device 1. In other embodiments, the memory 11 may also be an external storage device of the electronic device 1, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 1. Further, the memory 11 may also include both an internal storage unit of the electronic device 1 and an external storage device. The memory 11 may not only be used to store application software and various types of data installed in the electronic device 1, such as the code of the program 12, etc., but also be used to temporarily store data that has been output or is to be output.

[0279] The processor 10 may be composed of an integrated circuit in some embodiments, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and a combination of various control chips, etc. The processor 10 is the control core (Control Unit) of the electronic device, and uses various interfaces and lines to connect various components of the entire electronic device, and executes or executes programs or modules (such as programs 12 for realizing multi-dimensional EEG signal recognition) stored in the memory 11, and calls data stored in the memory 11 to execute various functions of the electronic device 1 and process data.

[0280] The communication interface 13 may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device 1 and other electronic devices, and to achieve connection and communication between internal components of the electronic device.

[0281] The bus may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to realize connection and communication between the memory 11 and at least one processor 10, etc.

[0282] Figure 2 Only an electronic device with components is shown, and those skilled in the art will understand that Figure 2 The structure shown does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0283] For example, although not shown, the electronic device 1 may also include a power source (such as a battery) for supplying power to various components. Preferably, the power source may be logically connected to the at least one processor 10 through a power management device, so that the power management device can realize functions such as charging management, discharging management, and power consumption management. The power source may also include one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The electronic device 1 may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0284] Optionally, the electronic device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), or a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode) touch device. The display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device 1 and to display a visual user interface.

[0285] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0286] The program 12 stored in the memory 11 in the electronic device 1 is a combination of multiple instructions. When running in the processor 10, it can achieve:

[0287] Collecting EEG signals at different positions, adding resonance information to the EEG signals for random resonance processing, obtaining EEG signals after resonance enhancement, and forming a training data set with the collected EEG signals and the EEG signals after resonance enhancement;

[0288] Construct an EEG signal resonance elimination model and use the training data set to train the model parameters;

[0289] Collecting the original EEG signal to be identified, and using the trained EEG signal resonance elimination model to eliminate the resonance of the original EEG signal to obtain the original EEG signal after the resonance is eliminated;

[0290] A multi-channel EEG signal recognition model is constructed to recognize the original EEG signal after the resonance is eliminated, and the recognition result of the original EEG signal is obtained.

[0291] Specifically, the specific implementation method of the processor 10 for the above instructions can refer to Figure 1 to Figure 2 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0292] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. And the terms "including", "comprising" or any other variants thereof in this article are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "including a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0293] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0294] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A multi-dimensional EEG signal recognition method, characterized in that: The method comprises: S1: Collect original EEG signals at different positions, add resonance information to the original EEG signals for stochastic resonance processing, obtain original EEG signals after resonance enhancement, and form a training data set with the collected original EEG signals and the original EEG signals after resonance enhancement; S2: constructing an EEG signal resonance elimination model and using a training data set to train model parameters, wherein the EEG signal resonance elimination model takes the original EEG signal as input and takes the original EEG signal after resonance elimination as output; S3: collecting the original EEG signal to be identified, and using the trained EEG signal resonance elimination model to eliminate the resonance of the original EEG signal to obtain the original EEG signal after the resonance is eliminated; S4: Construct a multi-channel EEG signal recognition model to recognize the original EEG signal after the resonance is eliminated, and obtain the recognition result of the original EEG signal.

2. A multi-dimensional EEG signal recognition method as claimed in claim 1, characterized in that: In the step S1, original EEG signals at different positions are collected, and resonance information is added to the original EEG signals for stochastic resonance processing, including: The electrodes are placed on different positions of the head of the EEG signal collector, and the acquisition board connected to the electrodes is used to record the potential changes in the brain in different positions as EEG signals. A total of N groups of original EEG signals are collected, where the collection of the original EEG signals is expressed as: ； in: represents the nth group of EEG signals collected, and t represents the time sequence information; Represents the original EEG signal The signal values ​​corresponding to the L signal sampling moments in , Represents the original EEG signal The signal value corresponding to the i-th signal sampling moment in , , the time interval corresponding to adjacent signal sampling times is ; Resonance information is added to the original EEG signal for random resonance processing to obtain the original EEG signal after resonance enhancement. The stochastic resonance processing flow is: S11: Combining raw EEG signals , construct the generalized Gaussian signal probability density function based on the signal value: ； ； ； in: represents the constructed generalized Gaussian signal probability density function, and x represents the signal value of the input generalized Gaussian signal probability density function; represents the preset standard deviation, represents the adjustment parameter, ; Represents the original EEG signal The mean of the signal values ​​in ; represents the gamma function; represents an exponential function with a natural constant as base; S12: Calculate the original EEG signal The resonance probability of any signal value in , where the original EEG signal The signal value corresponding to the i-th signal sampling time in The resonance probability is: ； ； in: Represents the original EEG signal The signal value corresponding to the i-th signal sampling time in The resonance probability of represents the probability density function of the generalized Gaussian signal The corresponding generalized Gaussian signal probability distribution value; S13: Calculate the original EEG signal The resonance range of any signal value in , where the original EEG signal The signal value corresponding to the i-th signal sampling time The resonance range is: ； in: Represents the original EEG signal The signal value corresponding to the i-th signal sampling time The resonance range of Represents the original EEG signal Any signal value and mean The difference sequence of Represents a difference sequence The mean of Represents a difference sequence The standard deviation of S14: The original EEG signal after resonance enhancement : ； in: Represents the original EEG signal The resonance enhancement results of Indicates the signal value The stochastic resonance processing result, where the stochastic resonance processing rule is: with the resonance probability For signal value Perform stochastic resonance processing to The probability of the signal value No stochastic resonance processing is performed; if the signal value Perform random resonance processing, then within the resonance range Randomly select any value and superimpose the selected value on the signal value The signal value The results of stochastic resonance processing Otherwise, the signal value is directly As a result of stochastic resonance .

3. A multi-dimensional EEG signal recognition method as claimed in claim 2, characterized in that: The collected EEG signals and the resonance-enhanced EEG signals are used to form a training data set, including: The collected original EEG signals and the original EEG signals after resonance enhancement constitute a training data set, where the training data set is represented as follows: ； in: Represents the constructed training data set; Represents the nth group of training data in the training data set data, represents the nth group of EEG signals collected, Represents the original EEG signal The resonance enhancement result.

4. A multi-dimensional EEG signal recognition method as claimed in claim 1, characterized in that: The step S2 constructs a brain wave signal resonance elimination model, including: Constructing an EEG signal resonance elimination model, wherein the EEG signal resonance elimination model takes the original EEG signal as input and takes the original EEG signal after resonance elimination as output; The EEG signal resonance elimination model includes an input layer, a power spectrum information extraction layer, a time convolution layer and an output layer. The input layer is used to receive the original EEG signal, the power spectrum information extraction layer is used to extract the power spectrum information of the original EEG signal, the time convolution layer is used to perform time series convolution processing on the original EEG signal based on the power spectrum information to obtain the original EEG signal after the resonance is eliminated, and the output layer is used to output the original EEG signal after the resonance is eliminated; Using the training dataset The model parameters of the EEG signal resonance elimination model are trained.

5. A multi-dimensional EEG signal recognition method as claimed in claim 4, characterized in that: The use of training data sets The model parameters of the EEG signal resonance elimination model are trained, including: The model parameter training process based on the training data set data is: S21: Construct the model parameter training objective function based on the training data set data: ； in: represents the L2 norm; represents the model parameter training objective function, Represents the model parameters to be trained; Represents model parameters based on The constructed EEG signal resonance elimination model; Indicates that the signal Input into the EEG signal resonance elimination model, and the model outputs the signal after resonance elimination; S22: Initialize and generate U groups of whale positions. Each group of whale positions corresponds to a model parameter. The uth group of whale positions generated by initialization is ; S23: Set the current iteration number of the whale position to c, and the maximum iteration number to C. Then the result of the cth iteration of the uth group of whale positions is , the initial value of c is 0; S24: The U group of whale positions obtained in the cth iteration are used as the model parameters in the model parameter training objective function, and the output result of the model parameter training objective function is used as the fitness function value of the whale position, then the whale position The fitness function value is ; Select the whale position with the lowest fitness function value among the U group of whale positions obtained in the cth iteration as the optimal whale position for the cth iteration ; S25: Set the current predation behavior coefficient of the whale position, where the whale position The current predation behavior coefficient of is: ； in: Indicates the location of the whale The current predation behavior coefficient; Represents a random number between 0 and 1; S26: Iterate the whale position based on the current predation behavior coefficient of the whale position, where the whale position The iteration formula is: ； in: Represents a random number between -1 and 1; represents any whale position in the U group of whale positions obtained in the cth iteration; S27: Cauchy mutation perturbation is performed on the iterative result of the whale position, where the whale position The Cauchy variation perturbation formula is: ； in: It means element-by-element addition; represents a random perturbation sequence that conforms to the standard Cauchy distribution; S28: Let c=c+1, return to step S24, until the maximum number of iterations C is reached, and use the U group of whale positions obtained in the Cth iteration as the model parameters in the model parameter training objective function, and select the whale position with the lowest fitness function value as the model parameter training result of the EEG signal resonance elimination model .

6. A multi-dimensional EEG signal recognition method as claimed in claim 5, characterized in that: In the step S3, the original EEG signal to be identified is collected, and the EEG signal resonance elimination model obtained by training is used to eliminate the resonance of the original EEG signal, including: Collect the original EEG signal to be identified ,in Represents the original EEG signal The signal value corresponding to the i-th signal sampling moment in the training is used to eliminate the original EEG signal using the EEG signal resonance elimination model. Resonance elimination is performed, where the original EEG signal The resonance elimination process is: S31: The input layer receives the original EEG signal ; S32: Power spectrum information extraction layer extracts raw EEG signals Power spectrum information: ； in: Represents the original EEG signal The power spectrum information of j represents the imaginary unit, ; Represents the original EEG signal frequency; S33: Temporal convolution layer based on power spectrum information Raw EEG signals Perform time series convolution processing to obtain the original EEG signal after resonance elimination : ； ； in: represents the convolution parameter matrix, where ; Represents convolution calculation; Represents the original EEG signal Signal value A signal value sequence with a center of and a length of Len.

7. A multi-dimensional EEG signal recognition method as claimed in claim 1, characterized in that: The multi-channel EEG signal recognition model is constructed in step S4, including: A multi-channel EEG signal recognition model is constructed, and the original EEG signal after the resonance is eliminated is recognized by using the multi-channel EEG signal recognition model to obtain the recognition result of the original EEG signal; The multi-channel EEG signal recognition model includes an input layer, a spatiotemporal information extraction layer, a spatiotemporal information fusion layer and an output layer. The input layer is used to receive the original EEG signal after resonance elimination. The spatiotemporal information extraction layer is used to extract the time characteristics and spatial characteristics of the signal respectively. The spatiotemporal information fusion layer is used to fuse the time characteristics and the spatial characteristics to obtain fused characteristics. The output layer is used to convert the fused characteristics into label information of the original EEG signal as the recognition result of the original EEG signal.

8. A multi-dimensional EEG signal recognition method as claimed in claim 7, characterized in that: The method of using a multi-channel EEG signal recognition model to recognize the original EEG signal after the resonance is eliminated includes: The original EEG signal after the resonance is eliminated is identified by using a multi-channel EEG signal recognition model. The identification process is: S41: The input layer receives the original EEG signal after resonance elimination ; S42: The spatiotemporal information extraction layer extracts signals respectively Temporal and spatial characteristics: ； ； ； ； ； in: Indicates signal The time characteristics of Indicates the signal value Time information; Indicates signal spatial characteristics; Indicates signal The covariance matrix of , T represents the transpose; Represents the covariance matrix The eigenvector corresponding to the kth largest eigenvalue obtained by eigendecomposition ; K represents the dimension of spatial features; S43: Fusion of temporal features And spatial characteristics , get the fusion features : ； in: Representing time characteristics The standard deviation of S44: The output layer will fuse features Convert to raw EEG signal The label information is used as the original EEG signal The recognition result, where the original EEG signal The label information is: ； in: Represents the original EEG signal The probability of label being label, label represents any label in the label set, Represents the mapping matrix corresponding to the label label; The original EEG signal The probability of different labels as the original EEG signal The label information, that is, the original EEG signal The recognition result.

Images