Distracted driving state recognition method and system based on EEG complex network

Abstract

The present invention provides a distracted driving state recognition method and system based on complex EEG network, which belongs to the field of behavior recognition technology, and obtains EEG data of the driver in the driving state; processes the obtained EEG data based on a pre-trained recognition model to obtain the recognition result of whether the driver's driving state is normal driving or distracted driving; wherein, the training set for training the recognition model includes multiple groups of data, each group of data includes multiple groups of EEG data and labels indicating that the driving state corresponding to the group of EEG data is normal driving or distracted driving. The present invention takes into account the regional information of the brain, and the constructed EEG signal network includes 2 different functional networks, covering more comprehensive network information features; it can detect the global and local changes of the brain functional network in the distracted driving state from the network topology structure under different types of driving distraction, so as to achieve higher recognition accuracy and reliability.

Description

Technical Field

[0001] The present invention relates to the field of behavior recognition technology, and in particular to a distracted driving state recognition method and system based on complex electroencephalogram (EEG) networks. Background Art

[0002] Electroencephalogram (EEG) signals contain high-frequency neural activity information. Emerging neuroimaging techniques have shown that the human brain can be represented as an interconnected semi-regular network across multiple brain regions, generating a dataset of functional connectivity patterns. A new, multidisciplinary approach to studying complex network systems has emerged, called complex network analysis, which describes the characteristics of complex systems through the topological structure within the network. In addition, studies have found that during the driving process, the EEG signals of car drivers are affected by the driver's expectations and unconscious behaviors, and the network topology will change accordingly. As one of the main causes of road traffic accidents, it is necessary to use the dynamic changes of EEG waves to identify distracted driving states, especially for the EEG components of the frontal and left motor areas of the brain. Therefore, it is crucial to study the differences in multiple frequency band characteristics of distracted states from two aspects: graph theory properties and key connections, using EEG complex networks.

[0003] Most of the research on detecting distracted driving in terms of physiological signals focuses on analyzing behavioral data combined with physiological signals to find out the power changes in different brain regions. Due to the difference in distraction tasks, the research on distracted driving is more detailed than that on fatigue driving, because when dealing with simple distraction tasks, driving behavior will not change significantly, and only slight changes can be seen in EEG signals. In terms of EEG signal analysis and processing, most driving state detections are based on changes in power spectrum and entropy, without involving regional information of the brain, and the mechanism of distraction state cannot be studied comprehensively and systematically. EEG network analysis can often better reflect some changes in driving state, especially the state of attention of the brain during distraction, by analyzing some statistical indicators of network topology. However, limited network connections and basic network properties are insufficient to explain various types of driving distraction, and the evaluation criteria lack accuracy. Summary of the invention

[0004] The object of the present invention is to provide a distracted driving state identification method and system based on complex EEG networks, which can detect the global and local changes of brain functional networks under distracted driving states from the network topology structure under different types of distracted driving, so as to achieve higher recognition accuracy and reliability, so as to solve at least one technical problem existing in the above-mentioned background technology.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] On the one hand, the present invention provides a distracted driving status identification method based on an EEG complex network, comprising: obtaining EEG data of a driver in a driving state; processing the obtained EEG data based on a pre-trained recognition model to obtain an identification result of whether the driver's driving state is normal driving or distracted driving; wherein, a training set for training the recognition model includes multiple groups of data, each group of data includes multiple groups of EEG data and a label indicating that the driving state corresponding to each group of EEG data is normal driving or distracted driving.

[0007] Preferably, training the recognition model includes: acquiring EEG data of normal driving and distracted driving by an acquisition device; preprocessing the EEG data, removing redundant eye movement artifacts, reducing corresponding noise, and dividing the EEG signals into four frequency bands of δ wave, θ wave, ɑ wave and β wave according to frequency; treating each channel as a network node, calculating the coherence and phase locking values ​​during task-free driving and distracted driving for every two channels, representing the correlation and synchronization between EEG signals, respectively, to obtain a brain functional connection matrix based on correlation and synchronization; selecting a certain range of sparsity threshold space, comparing each element value in the functional connection matrix with the threshold, converting the connection matrix into a binary adjacency matrix of normal and distracted states through the threshold, and sequentially constructing the adjacency matrix; calculating the complex network topological structure attributes under each threshold according to the adjacency matrix, and after obtaining the attributes calculated under all networks in the threshold space, selecting the curve area (AUC) of the network attributes under all sparsity spaces as the overall feature, and inputting it into the trained support vector machine to obtain the classification result of the driving state as normal driving or distracted driving state.

[0008] Preferably, the EEG data is preprocessed, including: removing high-frequency interference through a 0.5-40 Hz bandpass filter, then removing eye movement artifacts through independent component analysis, and reconstructing sub-band waveforms according to the frequency range, wherein the delta wave frequency is 0.5-4 Hz, the theta wave frequency is 4-8 Hz, the alpha wave frequency is 8-13 Hz, and the beta wave frequency is 13-30 Hz.

[0009] Preferably, obtaining the correlation and synchronization between EEG signals includes: using the coherence coefficient to calculate the correlation of the EEG signals recorded on the two scalp electrodes, the coherence is mostly used for frequency domain calculations, and can provide more information based on frequency; using the phase locking value to calculate the synchronization of the EEG signals recorded on the two scalp electrodes, that is, the calculated value of the phase synchronization information in the time domain.

[0010] Preferably, the coherence refers to the magnitude squared coherence, which is a measure of the linear correlation between two variables x(t) and y(t). Coherence is mostly used for frequency domain calculations:

[0011]

[0012] Where 0≤COH xy (f)≤1, 0 means there is no linear correlation between x(t) and y(t) at frequency f, and 1 means there is a correlation between x(t) and y(t) at frequency f; where K xy (f) is the coherence function of frequency f, representing the cross-power spectral density;

[0013] S xy (f) S xx (f) and S yy (f) is the power spectral density between x(t) and y(t) and their respective expressions:

[0014]

[0015] The phase-locked value is used to characterize the synchronization of the EEG signals recorded on the two scalp electrodes, that is, the calculated value of the phase synchronization information in the time domain, which measures the inter-sequence variability of the phase difference at time t:

[0016]

[0017] in, and are the phases of the x scalp electrode signal and the y scalp electrode signal at time t, j is the unit of imaginary number, and the value range of PLV is

[01] . The larger the value, the stronger the phase synchronization between the two signals.

[0018] Preferably, the complex network topology structure metrics include path length, clustering coefficient, global efficiency and local efficiency;

[0019] The clustering coefficient indicates the degree of aggregation of complex network nodes, which can reflect the degree of groupization and functional differentiation of brain functional networks:

[0020]

[0021] Among them, k i is the total number of neighboring nodes of node i in the network, e i is the total number of actual connection edges between node i and its neighboring nodes; C is the arithmetic mean of the clustering coefficients of all nodes in the network.

[0022] The characteristic path length represents the average value of the shortest path length between any two nodes in the network, and usually reflects the information transmission capacity within the network:

[0023]

[0024] Where L is the characteristic path length; L ijis the shortest path length between node i and node j, that is, the number of shortest edges from node i to node j; N is the number of nodes;

[0025] Global efficiency represents the shortest path length L ij The average of the reciprocals of is used to measure the ability of the brain functional network to transmit and process information;

[0026]

[0027] Local efficiency is used to measure local information transmission and processing capabilities:

[0028]

[0029] Among them, E global (G i ) is the global efficiency of Gi, where Gi refers to the subgraph consisting of the neighbors of node i.

[0030] Preferably, the input feature of the support vector machine is the overall attribute of the basic metric value of the complex network, that is, the overall characteristics of each attribute within the threshold range, which is characterized by the area under the attribute curve:

[0031]

[0032] Among them, Y represents the basic attribute measurement value of the complex network, ΔS represents the interval between two sparsity thresholds, and is the step size of the sparsity progression, S1 and S n Represent the upper and lower bounds of the sparsity threshold respectively.

[0033] In a second aspect, the present invention provides a distracted driving state recognition system based on complex EEG network, comprising:

[0034] An acquisition module, used to acquire the EEG data of the driver in a driving state;

[0035] A processing module is used to process the acquired EEG data based on a pre-trained recognition model to obtain a recognition result of whether the driver's driving state is normal driving or distracted driving; wherein, the training set for training the recognition model includes multiple groups of data, each group of data includes multiple groups of EEG data and a label indicating that the driving state corresponding to each group of EEG data is normal driving or distracted driving.

[0036] In a third aspect, the present invention provides a non-transitory computer-readable storage medium, which is used to store computer instructions. When the computer instructions are executed by a processor, the distracted driving state identification method based on complex EEG network as described above is implemented.

[0037] In a fourth aspect, the present invention provides a computer program product, comprising a computer program, which, when executed on one or more processors, is used to implement the distracted driving state identification method based on complex EEG network as described above.

[0038] In a fifth aspect, the present invention provides an electronic device, comprising: a processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes instructions for implementing the distracted driving state identification method based on complex EEG networks as described above.

[0039] The beneficial effects of the present invention are as follows: taking into account the regional information of the brain, the constructed EEG signal network includes two different functional networks, and the network information characteristics covered are more comprehensive; under different types of driving distraction, the global and local changes of the brain functional network under the state of driving distraction can be detected from the network topology structure to achieve higher recognition accuracy and reliability.

[0040] Additional advantages of the present invention will be more clearly presented in the following description or learned through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0042] Figure 1 This is a flow chart of a distracted driving state recognition method based on complex EEG network according to an embodiment of the present invention. DETAILED DESCRIPTION

[0043] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below by the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be interpreted as limiting the present invention.

[0044] It should be understood by those skilled in the art that unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those skilled in the art to which this invention belongs.

[0045] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a", "an", "said" and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements and / or groups thereof.

[0046] To facilitate understanding of the present invention, the present invention is further explained below with reference to specific embodiments in conjunction with the accompanying drawings, and the specific embodiments do not constitute a limitation on the embodiments of the present invention.

[0047] Those skilled in the art should understand that the drawings are merely schematic diagrams of embodiments, and the components in the drawings are not necessarily necessary for implementing the present invention.

[0048] Example 1

[0049] In this embodiment 1, a distracted driving state recognition system based on an EEG complex network is first provided, which includes: an acquisition module, used to acquire EEG data of the driver in the driving state; a processing module, used to process the acquired EEG data based on a pre-trained recognition model to obtain a recognition result of whether the driver's driving state is normal driving or distracted driving; wherein, the training set for training the recognition model includes multiple groups of data, each group of data includes multiple groups of EEG data and a label indicating that the driving state corresponding to each group of EEG data is normal driving or distracted driving.

[0050] In this embodiment 1, the above-mentioned system is used to implement a distracted driving state recognition method based on an EEG complex network, including: using an acquisition module to acquire EEG data of the driver in a driving state; using a processing module to process the acquired EEG data based on a pre-trained recognition model to obtain a recognition result of whether the driver's driving state is normal driving or distracted driving; wherein, the training set for training the recognition model includes multiple groups of data, each group of data includes multiple groups of EEG data and a label indicating that the driving state corresponding to each group of EEG data is normal driving or distracted driving.

[0051] Among them, training the recognition model includes: acquiring EEG data of normal driving and distracted driving by an acquisition device; preprocessing the EEG data, removing redundant eye movement artifacts, reducing corresponding noise, and dividing the EEG signals into four frequency bands of δ wave, θ wave, ɑ wave and β wave according to frequency; treating each channel as a network node, calculating the coherence and phase locking values ​​during task-free driving and distracted driving for every two channels, representing the correlation and synchronization between EEG signals, respectively, to obtain a brain functional connection matrix based on correlation and synchronization; selecting a certain range of sparsity threshold space, comparing each element value in the functional connection matrix with the threshold, converting the connection matrix into a binary adjacency matrix of normal and distracted states through the threshold, and constructing the adjacency matrix in turn; calculating the complex network topological structure attributes under each threshold according to the adjacency matrix, and after obtaining the attributes calculated under all networks in the threshold space, selecting the curve area (AUC) of the network attributes under all sparsity spaces as the overall feature, inputting it into the trained support vector machine, and obtaining the classification result of the driving state as normal driving or distracted driving.

[0052] The EEG data were preprocessed, including: removing high-frequency interference through a 0.5-40 Hz bandpass filter, then removing eye movement artifacts through independent component analysis, and reconstructing sub-band waveforms according to the frequency range, where the delta wave frequency is 0.5-4 Hz, the theta wave frequency is 4-8 Hz, the alpha wave frequency is 8-13 Hz, and the beta wave frequency is 13-30 Hz.

[0053] Obtaining the correlation and synchronization between EEG signals includes: using the coherence coefficient to calculate the correlation of the EEG signals recorded on the two scalp electrodes. Coherence is mostly used for frequency domain calculations and can provide more information based on frequency; using the phase-locked value to calculate the synchronization of the EEG signals recorded on the two scalp electrodes, that is, the calculated value of the phase synchronization information in the time domain.

[0054] Coherence refers to the squared amplitude coherence, which measures the linear correlation between two variables x(t) and y(t). Coherence is mostly used in frequency domain calculations:

[0055]

[0056] Where 0≤COH xy (f)≤1, 0 means there is no linear correlation between x(t) and y(t) at frequency f, and 1 means there is a correlation between x(t) and y(t) at frequency f; where K xy (f) is the coherence function of frequency f, representing the cross-power spectral density;

[0057] S xy (f) S xx (f) and S yy(f) is the power spectral density between x(t) and y(t) and their respective expressions:

[0058]

[0059] The phase-locked value is used to characterize the synchronization of the EEG signals recorded on the two scalp electrodes, that is, the calculated value of the phase synchronization information in the time domain, which measures the inter-sequence variability of the phase difference at time t:

[0060]

[0061] in, and are the phases of the x scalp electrode signal and the y scalp electrode signal at time t, j is the unit of imaginary number, and the value range of PLV is

[01] . The larger the value, the stronger the phase synchronization between the two signals.

[0062] Complex network topology metrics include path length, clustering coefficient, global efficiency, and local efficiency;

[0063] The clustering coefficient indicates the degree of aggregation of complex network nodes, which can reflect the degree of groupization and functional differentiation of brain functional networks:

[0064]

[0065] Among them, k i is the total number of neighboring nodes of node i in the network, e i is the total number of actual connection edges between node i and its neighboring nodes; C is the arithmetic mean of the clustering coefficients of all nodes in the network.

[0066] The characteristic path length represents the average value of the shortest path length between any two nodes in the network, and usually reflects the information transmission capacity within the network:

[0067]

[0068] Where L is the characteristic path length; L ij is the shortest path length between node i and node j, that is, the number of shortest edges from node i to node j; N is the number of nodes;

[0069] Global efficiency E g Indicates the shortest path length L ij The average of the reciprocals of is used to measure the ability of the brain functional network to transmit and process information;

[0070]

[0071] Local efficiency is used to measure local information transmission and processing capabilities:

[0072]

[0073] Among them, E global (G i ) is the global efficiency of Gi, where Gi refers to the subgraph consisting of the neighbors of node i.

[0074] The input features of the support vector machine are the overall attributes of the basic metrics of the complex network, that is, the overall characteristics of each attribute within the threshold range, which are characterized by the area under the attribute curve:

[0075]

[0076] Among them, Y represents the basic attribute measurement value of the complex network, ΔS represents the interval between two sparsity thresholds, and is the step size of the sparsity progression, S1 and S n Represent the upper and lower bounds of the sparsity threshold respectively.

[0077] Example 2

[0078] In this embodiment 2, the present invention provides a method for identifying distracted driving status based on complex EEG network, comprising the following steps:

[0079] Step 1: Design a distracted driving scenario with a secondary task, select several drivers to perform driving simulation, and use EEG signal acquisition equipment to obtain EEG data of non-secondary task driving and distracted driving stages;

[0080] Step 2: Preprocess the EEG signal to remove unnecessary eye movement artifacts, reduce the corresponding noise, and divide the EEG signal into four frequency bands, namely δ wave, θ wave, α wave and β wave;

[0081] Step 3: Treat each channel as a network node, calculate the coherence and phase locking values ​​during task-free driving and distracted driving for every two channels, representing the correlation and synchronization between EEG signals, respectively, and obtain the brain functional connection matrix based on correlation and synchronization in each frequency band;

[0082] Step 4: Select a certain range of sparsity threshold space and construct the adjacency matrix in sequence with a step size of 1%, that is, compare each element value in the functional connection matrix with the threshold, and convert the connection matrix into a binary adjacency matrix of normal and distracted states through the threshold. Calculate the complex network topology structure attribute value under each threshold, and select the curve area (AUC) measurement value of the network attribute under all sparsity spaces as the overall feature, input it into the trained support vector machine, and obtain the classification result of the driving state as normal driving or distracted driving state.

[0083] Among them, in step 1, the driving experiment including the secondary task includes: the simulated driving process is divided into five stages, in one stage, only the main task is to keep following the vehicle, and in the other four stages, the secondary task is added while maintaining the main task to interfere with the driver's distraction and keep following the vehicle.

[0084] Wherein, in step 2, the specific process of the pretreatment is:

[0085] 1) The collected EEG data was band-pass filtered from 0.5 Hz to 40 Hz.

[0086] 2) Eliminate damaged electrode channels and supplement them with interpolation of adjacent electrodes.

[0087] 3) Finally, the EEG data were set as the average reference and the eye movement artifacts were removed by ICA analysis.

[0088] 4) Reconstruct the sub-band waveform according to the frequency range through wavelet packet decomposition, where the frequency of delta wave is 0.5-4Hz, the frequency of theta wave is 4-8Hz, the frequency of alpha wave is 8-13Hz, and the frequency of beta wave is 13-30Hz.

[0089] The correlation and synchronization between the EEG signals in step 3 are obtained by the following steps:

[0090] 1) The correlation between the EEG signals recorded on the two scalp electrodes is calculated using the coherence coefficient;

[0091] 2) The synchronization of the EEG signals recorded at the two scalp electrodes was calculated using the phase-locking value;

[0092] The specific method for obtaining the adjacency matrix in step 4 is as follows: select a sparsity threshold in the range of (13%, 47%), and construct a binary network in sequence with a step size of 1% to obtain the channel connection relationship in the normal state and distracted state time series respectively. If the element value is greater than or equal to the connection strength threshold, the two channels are connected, and the value is 1; otherwise, there is no connection between the two channels, and the value is 0.

[0093] Among them, after all the adjacency matrices are obtained in step 4, the appropriate EEG complex network graph theory parameters can be calculated and selected through the adjacency matrix, including the basic metric clustering coefficient, which can reflect the degree of groupization and functional differentiation of the brain functional network; the characteristic path length, which is the average value of the shortest path length between any two nodes in the network, measures the information transmission capacity of the network; and the global efficiency used to measure the ability of the brain functional network to transmit and process information; and the local efficiency used to measure the local information transmission and processing capabilities. This type of parameter can well characterize the differences between different networks. The overall characteristics of each attribute within the sparsity range, that is, the area under the curve of each attribute, are input into the trained support vector machine to achieve the effect of identifying the distracted state.

[0094] Among them, coherence refers to the amplitude square coherence, which measures the linear correlation between two variables x(t) and y(t). Coherence is mostly used for frequency domain calculations and can provide more information based on frequency. Its formula is:

[0095]

[0096] Where 0≤COH xy (f)≤1, 0 means there is no linear correlation between x(t) and y(t) at frequency f, and 1 means there is a correlation between x(t) and y(t) at frequency f. xy (f) is the coherence function of frequency f, representing the cross-power spectral density, S xy (f) S xx (f) and S yy (f) is the power spectral density between x(t) and y(t) and their respective power spectral density, which is expressed as:

[0097]

[0098] The phase-locked value is used to characterize the synchronization of the EEG signals recorded on the two scalp electrodes, that is, the calculated value of the phase synchronization information in the time domain. It measures the inter-sequence variability of the phase difference at time t, and can provide more information based on the signal time domain. The specific relationship is as follows:

[0099]

[0100] in, and are the phases of the x scalp electrode signal and the y scalp electrode signal at time t, j is the unit of an imaginary number, N is the number of nodes, and the value range of PLV is

[01] . The larger the value, the stronger the phase synchronization between the two signals.

[0101] The complex network topology metrics include path length (L), clustering coefficient (C), global efficiency (E g ) and local efficiency (E local ), the formulas are:

[0102]

[0103]

[0104]

[0105]

[0106] The overall characteristics of each attribute input into the support vector machine are the area under the curve (AUC) of each attribute, and the formula is:

[0107]

[0108] Among them, ΔS represents the interval between two sparsity thresholds, is the step size of sparsity progression, and its value is 0.01. S1 and S n They represent the upper and lower bounds of the sparsity threshold, which are 0.13 and 0.47 respectively.

[0109] Example 3

[0110] In this embodiment 3, a distracted driving state recognition method based on complex EEG network is provided, comprising the following steps:

[0111] Step 1: By designing a car-following driving simulation experiment with subtasks, the driving stage is divided into normal car-following driving and distracted driving with subtasks. The distracted subtasks include math calculations, clock angle judgment, digital visual interference, and mobile phone operation. The first two represent cognitive distraction, and the last two represent visual distraction. The EEG signals of the driver during normal driving and distracted driving are collected synchronously.

[0112] Step 2.1: The collected EEG data is filtered through a 0.5 Hz to 40 Hz bandpass filter, and the sinusoidal noise of a specific frequency is removed.

[0113] Step 2.2: Remove damaged electrode channels and supplement them with interpolation of adjacent electrodes.

[0114] Step 2.3: The EEG data were averaged and referenced, and the artifact components were identified and removed using the ADJUST method after independent component analysis (ICA).

[0115] Step 2.4: Reconstruct the sub-band waveform according to the frequency range through wavelet packet transform, where the frequency of delta wave is 0.5-4 Hz, the frequency of theta wave is 4-8 Hz, the frequency of alpha wave is 8-13 Hz, and the frequency of beta wave is 13-30 Hz.

[0116] Step 3: Based on each frequency band, the 63 electrode channel nodes were regarded as network nodes, and the coherence and phase-locking values ​​during task-free driving and distracted driving were calculated for every two channels, representing the correlation and synchronization between EEG signals, respectively, to obtain a 63×63 brain functional connection matrix based on correlation and synchronization.

[0117] Step 4.1: The sparsity threshold is defined by the small-world network property. A sparsity range is selected and the network is reconstructed in 1% steps. Sparsity is the ratio of the actual number of edges in the brain functional network to the estimated maximum number of edges. The sparsity threshold applies a (relative) threshold to the connection weights so that the weakest connections are removed, thereby making the number of connections the same across subjects.

[0118] Step 4.2: Determine the appropriate sparsity threshold range, specifically (13% to 47%), and characterize it according to the small-world property of the complex network. Compare the strength threshold with the value of each element in the functional connection matrix. If the element value is greater than or equal to the connection strength threshold, the two channels are connected, and the value is 1. Otherwise, there is no connection between the two channels, and the value is 0. The binary network matrix under different thresholds of the four frequency bands is obtained.

[0119] Step 4.3: The adjacency matrix of normal and distracted states is obtained based on the connection relationship between the two channels. Then, the adjacency matrix is ​​used to calculate and select the appropriate metric value that characterizes the difference in the topological structure of the complex network, and the overall characteristics of each attribute within the sparsity range under different frequency bands, that is, the area under the curve of each attribute, is input into the trained support vector machine to obtain the classification result of the driving state as normal driving or distracted driving;

[0120] In this embodiment, the correlation coefficient is used in step 3 to calculate the correlation of the EEG signals recorded on the two scalp electrodes. Coherence refers to the squared amplitude coherence, which measures the linear correlation between two variables x(t) and y(t). Coherence is mostly used for frequency domain calculations and can provide more information based on frequency. The specific relationship is:

[0121] COH xy (f)=|K xy (f)| 2

[0122] Where 0≤COH xy (f)≤1, 0 means there is no linear correlation between x(t) and y(t) at frequency f, and 1 means there is a correlation between x(t) and y(t) at frequency f. xy (f) is the coherence function of frequency f, which represents the cross-power spectral density. Sxy(f), Sxx(f) and Syy(f) are the power spectral densities between x(t) and y(t) and their respective expressions:

[0123]

[0124] In step 3, the phase locking value is used to calculate the synchronization of the EEG signals recorded on the two scalp electrodes. The phase locking value is used to characterize the synchronization of the EEG signals recorded on the two scalp electrodes, that is, the calculated value of the phase synchronization information in the time domain. What is measured is the sequence variability of the phase difference at time t. The specific relationship is as follows:

[0125]

[0126] in, and are the phases of the x scalp electrode signal and the y scalp electrode signal at time t, j is the unit of imaginary number, and the value range of PLV is

[01] . The larger the value, the stronger the phase synchronization between the two signals.

[0127] In this embodiment, after step 4, a suitable threshold is selected. The threshold is selected as the sparsity threshold. Given a suitable sparsity range (13%, 47%), an adjacency matrix is ​​constructed with a step size of 1%. The corresponding complex network graph theory parameters are calculated through the adjacency matrix. The network characteristics such as path length, clustering coefficient, global efficiency and local efficiency in the network are used as basic parameters, and the overall characteristics of the network characteristics within the sparsity range are used as input classification features of the support vector machine.

[0128] The clustering coefficient of node i is the possibility of mutual connection (edge) between nodes adjacent to i, which can reflect the degree of groupization and functional differentiation of the brain functional network:

[0129]

[0130] Among them, k i is the total number of neighboring nodes of node i in the network, e i is the total number of actual connecting edges between node i and its neighboring nodes.

[0131] The characteristic path length of node i represents the average value of the shortest path length between any two nodes in the network, which can measure the information transmission capacity of the network:

[0132]

[0133] Where L is the characteristic path length; L ij is the shortest path length between node i and node j, that is, the number of shortest edges used from node i to node j; N is the number of nodes.

[0134] The global efficiency of the brain functional network can measure the ability of the brain functional network to transmit and process information:

[0135]

[0136] The local efficiency of the brain functional network can measure the local information transmission and processing capabilities:

[0137]

[0138] Among them, E global (G i ) is the global efficiency of node i, that is, E g .

[0139] According to the above steps, the input features of the support vector machine are the overall attributes of the basic metrics of the complex network, that is, the overall characteristics of each attribute within the threshold range, which are characterized by the area under the attribute curve:

[0140]

[0141] Among them, Y represents the basic attribute measurement value of the complex network, ΔS represents the interval between two sparsity thresholds, and is the step size of the sparsity progression, S1 and S n Represent the upper and lower bounds of the sparsity threshold respectively.

[0142] According to the above steps, the method described in this embodiment performs well in the classification results of the support vector machine under normal and distracted driving conditions, indicating that the overall network features have a relatively accurate recognition capability for distracted driving conditions.

[0143] Example 4

[0144] This embodiment 4 provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the distracted driving state recognition method based on the complex EEG network as described above is implemented. The method includes:

[0145] Obtaining the driver's EEG data during driving;

[0146] The acquired EEG data are processed based on a pre-trained recognition model to obtain a recognition result of whether the driver's driving state is normal driving or distracted driving; wherein, the training set for training the recognition model includes multiple groups of data, each group of data includes multiple groups of EEG data and a label indicating that the driving state corresponding to the group of EEG data is normal driving or distracted driving.

[0147] Example 5

[0148] This embodiment 5 provides a computer program product, including a computer program. When the computer program is executed on one or more processors, the computer program is used to implement the distracted driving state identification method based on EEG complex network as described above. The method includes:

[0149] Obtaining the driver's EEG data during driving;

[0150] The acquired EEG data are processed based on a pre-trained recognition model to obtain a recognition result of whether the driver's driving state is normal driving or distracted driving; wherein, the training set for training the recognition model includes multiple groups of data, each group of data includes multiple groups of EEG data and a label indicating that the driving state corresponding to the group of EEG data is normal driving or distracted driving.

[0151] Example 6

[0152] This embodiment 6 provides an electronic device, including: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to execute instructions for implementing the distracted driving state recognition method based on complex EEG network as described above, the method comprising:

[0153] Obtaining the driver's EEG data during driving;

[0154] The acquired EEG data are processed based on a pre-trained recognition model to obtain a recognition result of whether the driver's driving state is normal driving or distracted driving; wherein, the training set for training the recognition model includes multiple groups of data, each group of data includes multiple groups of EEG data and a label indicating that the driving state corresponding to the group of EEG data is normal driving or distracted driving.

[0155] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0156] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0157] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0158] These computer program instructions can also be loaded onto a computer or other programmable data processing device, and a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0159] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative work on the basis of the technical solution disclosed in the present invention should be included in the scope of protection of the present invention.

Claims

1. A distracted driving state recognition method based on EEG complex network, characterized in that: include: Obtaining the driver's EEG data during driving; The acquired EEG data are processed based on a pre-trained recognition model to obtain a recognition result of whether the driver's driving state is normal driving or distracted driving; wherein the training set for training the recognition model includes multiple groups of data, each group of data includes multiple groups of EEG data and a label indicating that the driving state corresponding to the group of EEG data is normal driving or distracted driving; Training the recognition model includes: acquiring EEG data of normal driving and distracted driving by using an acquisition device; preprocessing the EEG data, removing redundant eye movement artifacts, reducing corresponding noise, and dividing the EEG signals into four frequency bands of δ wave, θ wave, ɑ wave and β wave according to frequency; treating each channel as a network node, calculating the coherence and phase locking values ​​during task-free driving and distracted driving for every two channels, representing the correlation and synchronization between EEG signals, respectively, to obtain a brain functional connection matrix based on correlation and synchronization; selecting a sparsity threshold space in the range of 13% to 47%, comparing each element value in the functional connection matrix with the threshold, converting the connection matrix into a binary adjacency matrix of normal and distracted states through the threshold, and sequentially constructing the adjacency matrix; calculating the complex network topological structure attributes under each threshold according to the adjacency matrix, and after obtaining the attributes calculated under all networks in the threshold space, selecting the curve area of ​​the network attributes under all sparsity spaces as the overall feature, and inputting it into the trained support vector machine to obtain the classification result of the driving state as normal driving or distracted driving; in, Coherence refers to the magnitude-squared coherence, which measures the linear correlation between two variables x(t) and y(t): COH xy (f)=|K xy (f)| 2 Where 0≤COH xy (f)≤1, 0 means there is no linear correlation between x(t) and y(t) at frequency f, and 1 means there is a correlation between x(t) and y(t) at frequency f; where K xy (f) is the coherence function of frequency f, representing the cross-power spectral density; S xy (f) S xx (f) and S yy (f) is the power spectral density expression between x(t) and y(t) and their respective power spectral density expressions: The phase-locked value is used to characterize the synchronization of the EEG signals recorded on the two scalp electrodes, that is, the calculated value of the phase synchronization information in the time domain, which measures the inter-sequence variability of the phase difference at time t: in, and are the phases of the x scalp electrode signal and the y scalp electrode signal at time t, j is an imaginary unit, and the value range of PLV is [0, 1]. The larger the value, the stronger the phase synchronization between the two signals; The complex network topology structure metrics include path length, clustering coefficient, global efficiency and local efficiency; the clustering coefficient indicates the degree of aggregation of complex network nodes, reflecting the degree of groupization and functional differentiation of brain functional networks: Among them, k i is the total number of neighboring nodes of node i in the network, e i is the total number of actual connection edges between node i and its neighboring nodes; C is the arithmetic mean of the clustering coefficients of all nodes in the network; The characteristic path length represents the average value of the shortest path length between any two nodes in the network, reflecting the information transmission capacity within the network: Where L is the characteristic path length; L ij is the shortest path length between node i and node j, that is, the number of shortest edges from node i to node j; N is the number of nodes; Global efficiency E g Indicates the shortest path length L ij The average of the reciprocals of is used to measure the ability of the brain functional network to transmit and process information; Local efficiency is used to measure local information transmission and processing capabilities: Among them, E g (G i ) is the global efficiency of Gi, where Gi refers to the subgraph consisting of the neighbors of node i.

2. The distracted driving state recognition method based on EEG complex network according to claim 1 is characterized in that: The EEG data were preprocessed, including: removing high-frequency interference through a 0.5-40 Hz bandpass filter, then removing eye movement artifacts through independent component analysis, and reconstructing sub-band waveforms according to the frequency range, where the delta wave frequency is 0.5-4 Hz, the theta wave frequency is 4-8 Hz, the alpha wave frequency is 8-13 Hz, and the beta wave frequency is 13-30 Hz.

3. The distracted driving state recognition method based on EEG complex network according to claim 1, characterized in that: Obtaining the correlation and synchronization between EEG signals includes: calculating the correlation of the EEG signals recorded on two scalp electrodes using the coherence coefficient; calculating the synchronization of the EEG signals recorded on the two scalp electrodes using the phase locking value, that is, the calculated value of the phase synchronization information in the time domain.

4. The distracted driving state recognition method based on EEG complex network according to claim 1 is characterized in that: The input features of the support vector machine are the overall attributes of the basic metrics of the complex network, that is, the overall characteristics of each attribute within the threshold range, which are characterized by the area under the attribute curve: Among them, Y represents the basic attribute measurement value of the complex network, ΔS represents the interval between two sparsity thresholds, and is the step size of the sparsity progression, S1 and S n Represent the upper and lower bounds of the sparsity threshold respectively.

5. A distracted driving state recognition system based on complex EEG network, characterized in that: include: An acquisition module, used to acquire the EEG data of the driver in a driving state; The processing module is used to process the acquired EEG data based on a pre-trained recognition model to obtain a recognition result of whether the driver's driving state is normal driving or distracted driving; wherein the training set for training the recognition model includes multiple groups of data, each group of data includes multiple groups of EEG data and labels indicating that the driving state corresponding to the group of EEG data is normal driving or distracted driving; training the recognition model includes: acquiring EEG data of normal driving and distracted driving stages by an acquisition device; pre-processing the EEG data, removing redundant eye movement artifacts, reducing corresponding noise, and dividing the EEG signals into four frequency bands of δ wave, θ wave, ɑ wave and β wave according to frequency; treating each channel as a network node, and calculating the task-free driving for every two channels and distracted driving, representing the correlation and synchronization between EEG signals, respectively, to obtain a brain functional connection matrix based on correlation and synchronization; selecting a sparsity threshold space ranging from 13% to 47%, comparing each element value in the functional connection matrix with the threshold, converting the connection matrix into a binary adjacency matrix of normal and distracted states through the threshold, and constructing the adjacency matrix in turn; calculating the complex network topological structure attributes under each threshold according to the adjacency matrix, obtaining the attributes calculated under all networks in the threshold space, selecting the curve area of ​​the network attributes under all sparsity spaces as the overall feature, and inputting it into the trained support vector machine to obtain the classification result of the driving state as normal driving or distracted driving; in, Coherence refers to the magnitude-squared coherence, which measures the linear correlation between two variables x(t) and y(t): COH xy (f)=|K xy (f)| 2 Where 0≤COH xy (f)≤1, 0 means there is no linear correlation between x(t) and y(t) at frequency f, and 1 means there is a correlation between x(t) and y(t) at frequency f; where K xy (f) is the coherence function of frequency f, representing the cross-power spectral density; S xy (f) S xx (f) and S yy (f) is the power spectral density expression between x(t) and y(t) and their respective power spectral density expressions: The phase-locked value is used to characterize the synchronization of the EEG signals recorded on the two scalp electrodes, that is, the calculated value of the phase synchronization information in the time domain, which measures the inter-sequence variability of the phase difference at time t: in, and are the phases of the x scalp electrode signal and the y scalp electrode signal at time t, j is an imaginary unit, and the value range of PLV is [0, 1]. The larger the value, the stronger the phase synchronization between the two signals; The complex network topology structure metrics include path length, clustering coefficient, global efficiency and local efficiency; the clustering coefficient indicates the degree of aggregation of complex network nodes, reflecting the degree of groupization and functional differentiation of brain functional networks: Among them, k i is the total number of neighboring nodes of node i in the network, e i is the total number of actual connection edges between node i and its neighboring nodes; C is the arithmetic mean of the clustering coefficients of all nodes in the network; The characteristic path length represents the average value of the shortest path length between any two nodes in the network, reflecting the information transmission capacity within the network: Where L is the characteristic path length; L ij is the shortest path length between node i and node j, that is, the number of shortest edges from node i to node j; N is the number of nodes; Global efficiency E g Indicates the shortest path length L ij The average of the reciprocals of is used to measure the ability of the brain functional network to transmit and process information; Local efficiency is used to measure local information transmission and processing capabilities: Among them, E g (G i ) is the global efficiency of Gi, where Gi refers to the subgraph consisting of the neighbors of node i.

6. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by the processor, the distracted driving state identification method based on the complex EEG network as described in any one of claims 1-4 is implemented.

7. An electronic device, characterized in that: include: A processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes instructions for implementing the distracted driving state recognition method based on EEG complex network as described in any one of claims 1-4.

Images