A method, device and electronic equipment for evaluating a movement intention

Abstract

This application provides a method, system, and electronic device for assessing motor intention, relating to the field of electroencephalogram (EEG) signal analysis and processing technology. The method includes acquiring raw EEG signals from 62 channels of a user; selecting channels from the raw EEG signals to determine a first EEG signal; preprocessing and feature construction of the first EEG signal to obtain an adjacency matrix of the EEG signal; and determining the causal relationship between the associated edges of the channels using a channel determination model based on the adjacency matrix of the EEG signal, thereby assisting in assessing the user's motor intention.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electroencephalogram signal analysis and processing, and in particular to a motor intention evaluation method, device and electronic equipment. BACKGROUND

[0002] Electroencephalogram signals provide a lot of information about a person, reflecting the causal influence characteristics between brain regions. The prior art mentions calculating the causal relationship between electroencephalogram signals and electromyogram signals based on Granger causality, and using the obtained causal relationship to evaluate and analyze the brain-muscle network. The prior art can only determine the causal relationship between different physiological signals, and cannot determine the causal relationship between signal collection channels, which weakens the interpretability of the evaluation result, thereby reducing the accuracy of the evaluation.

[0003] Therefore, a motor intention evaluation method, device and electronic equipment are proposed. SUMMARY

[0004] The present application provides a motor intention evaluation method, device and electronic equipment, which determines the causal relationship of the associated edges between channels through a channel determination model, assists in evaluating the motor intention of a demand user, and improves the accuracy of the evaluation.

[0005] The motor intention evaluation method provided by the present application adopts the following technical solution, which comprises the following steps:

[0006] Original electroencephalogram signals of 62 channels of a demand user are collected, the original electroencephalogram signals are subjected to channel selection, and a first electroencephalogram signal is determined;

[0007] The first electroencephalogram signal is preprocessed and feature constructed to obtain an adjacency matrix of the electroencephalogram signal;

[0008] According to the adjacency matrix of the electroencephalogram signal, the causal relationship of the associated edges between channels is determined through a channel determination model, and the motor intention of the demand user is assisted in evaluation.

[0009] Optionally, the preprocessing and feature construction of the first electroencephalogram signal to obtain the adjacency matrix of the electroencephalogram signal comprises:

[0010] The first electroencephalogram signal is resampled and denoised to obtain a second electroencephalogram signal;

[0011] The second electroencephalogram signal is segmented and frequency band power extracted to obtain electroencephalogram channel point features;

[0012] The adjacency matrix of the electroencephalogram signal is constructed according to the electroencephalogram channel point features.

[0013] Optionally, the construction of the adjacency matrix of the electroencephalogram signal according to the electroencephalogram channel point features comprises:

[0014] determining a geodesic distance value between two channels and a consistent correlation value between the electroencephalogram signals of the two channels;

[0015] when the consistent correlation value meets a preset correlation condition, obtaining a weight value of a correlation edge between the two channels according to the consistent correlation value and the geodesic distance value;

[0016] constructing an adjacency matrix of the electroencephalogram signals according to all the correlation edges and the corresponding weight values.

[0017] Optionally, the determining of the causal relationship of the correlation edge between the channels through the channel determination model according to the adjacency matrix of the electroencephalogram signals comprises:

[0018] performing a distillation operation on the adjacency matrix of the electroencephalogram signals through a correlation edge classification model;

[0019] training the adjacency matrix of the electroencephalogram signals according to the distilled adjacency matrix to obtain the causal relationship of the correlation edge between the channels.

[0020] Optionally, the distillation operation on the adjacency matrix of the electroencephalogram signals through the correlation edge classification model comprises:

[0021] sequentially deleting a correlation edge of the adjacency matrix according to the correlation edge classification model to obtain a causal relationship contribution value of each correlation edge, deleting the correlation edge with the smallest causal relationship contribution value according to a difference value of cross entropy, and sorting the remaining correlation edges;

[0022] deleting the correlation edges from bottom to top according to the size of the causal relationship contribution value, adjusting the weight values of the correlation edges according to the difference value of cross entropy, and the correlation edge classification model being a GNN graph neural network.

[0023] Optionally, the adjusting of the weight values of the correlation edges comprises:

[0024] if the cross entropy increases after the deletion of the correlation edge, the difference value of the increased cross entropy is filled into the distilled adjacency matrix as the weight value of the deleted correlation edge;

[0025] if the cross entropy decreases after the deletion of the correlation edge, the weight value of the deleted correlation edge is reset to 0.

[0026] Optionally, the training of the adjacency matrix of the electroencephalogram signals according to the distilled adjacency matrix to obtain the causal relationship of the correlation edge between the channels comprises:

[0027] The graph variational autoencoder is used for training the adjacency matrix of the electroencephalogram signal, and when the similarity degree of the trained adjacency matrix and the distilled adjacency matrix reaches a preset similarity threshold, the causal relationship of each correlation edge of the trained adjacency matrix is obtained, and the correlation degree of the channels corresponding to the correlation edge is determined.

[0028] The application provides a motion intention evaluation system, which adopts the technical scheme as follows:

[0029] The acquisition module is configured to acquire original electroencephalogram signals of 62 channels of a demand user, select channels of the original electroencephalogram signals, and determine first electroencephalogram signals.

[0030] The preprocessing and feature construction module is configured to preprocess and construct features of the first electroencephalogram signals, and obtain an adjacency matrix of the electroencephalogram signals.

[0031] The evaluation module is configured to determine a causal relationship of a correlation edge between channels according to the adjacency matrix of the electroencephalogram signals by using a channel determination model, and assist in evaluating a motion intention of the demand user.

[0032] Optionally, the preprocessing and feature construction module comprises:

[0033] The denoising sub-module is configured to resample and denoise the first electroencephalogram signals, and obtain second electroencephalogram signals.

[0034] The feature extraction sub-module is configured to segment and extract frequency band power of the second electroencephalogram signals, and obtain electroencephalogram channel point features.

[0035] The matrix determination sub-module is configured to construct the adjacency matrix of the electroencephalogram signals according to the electroencephalogram channel point features.

[0036] Optionally, the matrix determination sub-module comprises:

[0037] The first determination unit is configured to determine a geodesic distance value between two channels.

[0038] The second determination unit is configured to determine a coherence correlation value between electroencephalogram signals of two channels.

[0039] The weight value determination unit is configured to, when the coherence correlation value meets a preset correlation condition, obtain a weight value of a correlation edge between two channels according to the coherence correlation value and the geodesic distance value.

[0040] The matrix construction unit is configured to construct the adjacency matrix of the electroencephalogram signals according to all the correlation edges and corresponding weight values.

[0041] Optionally, the evaluation module comprises:

[0042] a distillation submodule configured to perform a distillation operation on an adjacency matrix of the electroencephalogram signal by using a correlation edge classification model;

[0043] a training submodule configured to train the adjacency matrix of the electroencephalogram signal according to the distilled adjacency matrix, and obtain a causal relationship of a correlation edge between channels.

[0044] Optionally, the distillation submodule comprises:

[0045] a sorting unit configured to sequentially delete a correlation edge of the adjacency matrix according to the correlation edge classification model, obtain a causal relationship contribution value of each correlation edge, delete a correlation edge with the smallest causal relationship contribution value according to a difference in cross-entropy, and sort the remaining correlation edges;

[0046] a weight adjustment unit configured to sequentially delete the correlation edges from bottom to top according to the size of the causal relationship contribution value, adjust a weight value of the correlation edge according to the difference in cross-entropy, and use the correlation edge classification model as a GNN graph neural network.

[0047] Optionally, the weight adjustment unit comprises:

[0048] a first weight adjustment subunit configured to, if the cross-entropy increases after the correlation edge is deleted, fill a difference in the increased cross-entropy as the weight value of the deleted correlation edge into the distilled adjacency matrix;

[0049] a second weight adjustment subunit configured to, if the cross-entropy decreases after the correlation edge is deleted, reset the weight value of the deleted correlation edge to 0.

[0050] Optionally, the training submodule comprises:

[0051] a channel correlation degree determination unit configured to train the adjacency matrix of the electroencephalogram signal by using a graph variational autoencoder, obtain a causal relationship of each correlation edge of the trained adjacency matrix when a similarity degree of the trained adjacency matrix and the distilled adjacency matrix reaches a preset similarity threshold, and determine a correlation degree of a channel corresponding to the correlation edge.

[0052] The specification also provides an electronic device, wherein the electronic device comprises:

[0053] a processor; and

[0054] a memory storing computer-executable instructions that, when executed, cause the processor to perform any of the above methods.

[0055] The specification also provides a computer-readable storage medium, wherein the computer-readable storage medium stores one or more programs, which when executed by a processor, implement any of the above methods.

[0056] In the present application, by collecting the original electroencephalogram signals of 62 channels of the demand user, the original electroencephalogram signals are selected to determine the first electroencephalogram signals; the first electroencephalogram signals are preprocessed and features are constructed to obtain an adjacency matrix of the electroencephalogram signals; according to the adjacency matrix of the electroencephalogram signals, the causal relationship of the associated edges between the channels is determined through a channel determination model to assist in evaluating the motion intention of the demand user and improve the accuracy of the evaluation. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 A principle diagram of a motion intention evaluation method provided by an embodiment of the present specification;

[0058] Figure 2 A structure diagram of a motion intention evaluation method system provided by an embodiment of the present specification;

[0059] Figure 3 A structure diagram of an electronic device provided by an embodiment of the present specification;

[0060] Figure 4 A principle diagram of a computer-readable medium provided by an embodiment of the present specification. DETAILED DESCRIPTION

[0061] The following description is provided to enable those skilled in the art to carry out the present application. The preferred embodiments in the following description are only examples and other obvious modifications can be made by those skilled in the art. The basic principles defined in the following description can be applied to other embodiments, modifications, improvements, equivalents and other technical solutions without departing from the spirit and scope of the present application.

[0062] Exemplary embodiments of the present application will now be described more fully with reference to the accompanying drawings. However, exemplary embodiments can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these exemplary embodiments are provided so that the present application will be thorough and complete, and will fully convey the inventive concept to those skilled in the art. Like reference numerals refer to like elements throughout the specification and drawings and repeated description of these elements will be omitted.

[0063] Features, structures, characteristics or other details described in a certain embodiment can be combined in suitable manner in one or more other embodiments without departing from the technical concept of the present application.

[0064] In the description of specific embodiments, the features, structures, characteristics or other details described are for the purpose of making the embodiments sufficiently understood by those skilled in the art. However, it does not exclude that one or more of the features, structures, characteristics or other details can not be practiced without the specific feature, structure, characteristic or other detail.

[0065] The flowchart shown in the accompanying drawings is only an exemplary illustration, and does not necessarily include all contents and operations / steps, nor does it necessarily execute in the order described. For example, some operations / steps can be further decomposed, and some operations / steps can be combined or partially combined, so the actual execution order can be changed according to the actual situation.

[0066] The block diagram shown in the accompanying drawings is only a functional entity, and does not necessarily correspond to a physically independent entity. That is, the functional entity can be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0067] The term "and / or" or "and / or" includes all combinations of any one or more of the associated listed items.

[0068] Figure 1 A schematic diagram of the principle of an evaluation method of motion intention provided by an embodiment of the present specification, the method comprising:

[0069] S1 collects 62-channel raw electroencephalogram signals of a demand user, selects channels for the raw electroencephalogram signals, and determines a first electroencephalogram signal;

[0070] S2 pre-processes and constructs features for the first electroencephalogram signal to obtain an adjacency matrix of the electroencephalogram signal;

[0071] S3 determines the causal relationship of the associated edges between the channels according to the adjacency matrix of the electroencephalogram signal through a channel determination model, and assists in evaluating the motion intention of the demand user.

[0072] The generation of movement includes the induction of brain movement intention to the control of muscle movement execution and the feedback of this closed loop. If any link of the closed loop is out of order, the movement execution will fail. Effective identification and evaluation of brain movement intention can help to improve the behavior, emotion and other conditions of the demand user, and timely find the abnormal behavior of the demand user. The existing network analysis is basically based on electroencephalogram signals and / or other physiological signals, which can only be classified and analyzed to a certain extent, and cannot obtain the accurate granger causality between the signal collection channels, and the explainability is weak. Based on this, an evaluation method of movement intention is proposed, which determines the causality of the associated edge between the channels through a channel determination model, assists in evaluating the movement intention of the demand user, provides an internal basis for the effectiveness of passive training, and improves the accuracy of evaluation.

[0073] Specifically, it comprises:

[0074] S1 collects the original electroencephalogram signals of 62 channels of the demand user, selects the channels of the original electroencephalogram signals, and determines the first electroencephalogram signals;

[0075] Since the electroencephalogram signals are relatively weak and are easily disturbed, the electroencephalogram signal collection device should be grounded as much as possible to reduce the interference of electromagnetic field and electrostatic field. In the process of collecting electroencephalogram signals, the electrodes are placed on the brain of the demand user according to the international standard lead 10-20 system, and the original electroencephalogram signals of 62 channels are collected.

[0076] Then the original electroencephalogram signals are selected to obtain the first electroencephalogram signals, and the number of channels of the first electroencephalogram signals is n, wherein n < 62. Specifically, the original electroencephalogram signals are selected according to the preset selection condition to obtain the first electroencephalogram signals; the EEG signals have different forms on different EEG channels, and the preset selection condition includes a preset channel. In an embodiment of the present disclosure, if the present disclosure is applied to the auxiliary evaluation decision of epileptic seizure, the preset selection condition is 8 preset channels which can best reflect the epileptic seizure. The first electroencephalogram signals are reduced from 62 channels to 8 channels according to the preset selection condition. Specifically, 4 pairs of bipolar electrodes are selected from each brain hemisphere, a total of 16 electrodes, and 8 preset channel electroencephalogram data is generated for each user as the first electroencephalogram signals. Specifically, the 8 preset channels include: F7-F3 channel, F8-F4 channel, T7-C3 channel, T8-C4 channel, P7-P3 channel, P8-P4 channel, O1-P3 channel and O2-P4 channel.

[0077] Wherein, two electrodes form a channel, and the signal of the channel is obtained by subtracting the signals of the two electrodes, and the coordinates of the channel are the coordinates of an electrode located between the two electrodes.

[0078] In another embodiment of the present specification, the specific channel can also be directly determined by the relevant personnel according to the actual use needs, and the original electroencephalogram is subjected to channel selection to obtain the first electroencephalogram. The first electroencephalogram is preprocessed and feature constructed to obtain the adjacency matrix of the electroencephalogram.

[0079] The electroencephalogram is a kind of high dynamic, nonlinear data, and the original electroencephalogram is large in quantity and has redundancy. In order to reduce the calculation data quantity of preprocessing, the original electroencephalogram is subjected to channel selection to determine the first electroencephalogram, and then the first electroencephalogram is preprocessed. Since the recorded electroencephalogram has high time-varying sensitivity and is easily disturbed, in order to improve the efficiency and accuracy of the later analysis, the first electroencephalogram is preprocessed and feature constructed to obtain the adjacency matrix of the electroencephalogram.

[0080] Firstly, the first electroencephalogram is preprocessed, that is, the first electroencephalogram is resampled and denoised to obtain the second electroencephalogram.

[0081] The electroencephalogram is a kind of non-stationary signal with strong randomness, which is easily contaminated by irrelevant noise. In order to improve the accuracy of the later analysis, firstly, the first electroencephalogram is resampled, and the frequency of the resampling is preferably 250 Hz; then the resampled data is filtered to remove noise to obtain the second electroencephalogram.

[0082] Secondly, the second electroencephalogram is feature constructed to obtain the adjacency matrix of the electroencephalogram, which specifically includes:

[0083] The second electroencephalogram is segmented and the frequency band power is extracted to obtain the electroencephalogram channel point feature.

[0084] The energy size and change of different frequency bands in the second electroencephalogram represent the state of the brain, therefore, the wave band corresponding to the frequency of the second electroencephalogram is determined by extracting the frequency band power of each wave band to determine the electroencephalogram channel point feature.

[0085] The second electroencephalogram is segmented.

[0086] In one embodiment of the present specification, the second electroencephalogram is divided into continuous and non-overlapping windows, the window length is 10 seconds, the frequency band energy of the electroencephalogram of each window is obtained by the power spectral density function, and the electroencephalogram wave segment corresponding to the frequency of the electroencephalogram of each window is determined. The wave bands of the electroencephalogram waves include six kinds, and the wave bands and corresponding frequency ranges of the six electroencephalogram waves are shown in Table 1:

[0087]

[0088] (Table 1)

[0089] S222 obtains the brain electrical channel point features based on the segmentation results;

[0090] The total band power of each channel in the second brain electrical signal is extracted from each wave band as the brain electrical channel point feature, and a shape feature matrix of the brain electrical signal of each window is obtained according to the brain electrical channel point features of each wave band, wherein the shape feature matrix of the brain electrical signal includes n channels*6 brain electrical channel point features.

[0091] In an embodiment of the present specification, each channel corresponds to a time sequence diagram of a brain electrical signal, a window of 10 is set according to the psd technology (power spectral density), the power of each window is calculated, and 6 wave bands of windows are taken, so that 6 brain electrical channel point features are obtained for each channel, and a total of 8 channels, so that an 8*6 (8 channels, each channel corresponding to 6-dimensional features) point feature matrix, i.e., a shape feature matrix, can be obtained.

[0092] S23 constructs an adjacency matrix of the brain electrical signal according to the brain electrical channel point features;

[0093] When placing the electrodes according to the standard 10-20 electrode configuration, two electrodes form a channel, each channel corresponds to a channel node in the adjacency matrix, and the weight value A of the edge between the two channel nodes is obtained. i,j .

[0094] S231 determines the geodesic distance value between channel i and channel j; wherein channel i and channel j refer to any two channels.

[0095] Specifically, the geodesic distance value in the adjacency matrix of the brain electrical signal is calculated through the coordinates of channel i and channel j

[0096]

[0097] A i,j is the inverse cosine value of the distance between channel i and channel j.

[0098] S232 determines the coherence correlation value between the brain electrical signals of channel i and channel j

[0099]

[0100] Wherein, S ii and S jj are power spectral densities, and S ij is a cross spectral density.

[0101] S233 obtains the weight value A of the associated edge between the two channels according to the coherence correlation value and the geodesic distance value.i,j the weight value R i,j the consistent correlation value the average of the geodesic distance values, i.e. wherein,

[0102] i.e., the associated edge between channel i and channel j is the ith row and jth column in the adjacency matrix, S244 constructs the adjacency matrix of the electroencephalogram signal according to the eight channels and the corresponding weight values A i,j constructs the adjacency matrix of the electroencephalogram signal.

[0103] S3 determines the causal relationship of the associated edge between channels by a channel determination model according to the adjacency matrix of the electroencephalogram signal, to assist in evaluating the motor intention of the demand user.

[0104] S31 performs a distillation operation on the adjacency matrix of the electroencephalogram signal by an associated edge classification model;

[0105] In an embodiment of the present specification, the adjacency matrix of the electroencephalogram signal is explained by a Granger causal relationship using an associated edge classification model. Specifically, according to the associated edge classification model, one associated edge of the adjacency matrix is sequentially deleted, the causal relationship contribution value of each associated edge is obtained by deleting the associated edge one by one, the difference value of the cross entropy with or without the associated edge is calculated, the associated edge with the minimum causal relationship contribution is determined according to the difference value of the cross entropy, the associated edge with the minimum causal relationship contribution is deleted, and the remaining associated edges are sorted.

[0106] According to the size of the causal relationship contribution value, the associated edges are sequentially deleted from bottom to top, and the weight value of the associated edge is adjusted according to the difference value of the cross entropy.

[0107] Specifically, if the adjacency matrix of the electroencephalogram signal is connected at this time, the associated edges are sequentially deleted from bottom to top according to the size of the causal relationship contribution value, and the weight value of the associated edge is adjusted according to the change of the cross entropy after the associated edge is deleted, wherein the model uncertainty before and after the edge is deleted is the size of the change of the cross entropy.

[0108] If the cross entropy increases after the associated edge is deleted, the difference value of the increased cross entropy is filled into the distilled adjacency matrix as the weight value of the deleted associated edge. If the cross entropy decreases after the associated edge is deleted, the weight value of the deleted associated edge is reset to 0, or the associated edge is deleted. In an embodiment of the present specification, if the deletion of the associated edge does not change the cross entropy, the weight value of the deleted associated edge remains unchanged.

[0109] If the adjacency matrix of the electroencephalogram signal at this time is not connected, find the largest connected subgraph of the adjacency matrix of the electroencephalogram signal, and perform the above operation to obtain the distilled adjacency matrix. The association edge classification model is a GNN graph neural network.

[0110] S32 trains the adjacency matrix of the electroencephalogram signal according to the distilled adjacency matrix to obtain the causal relationship between the association edges between channels.

[0111] The adjacency matrix of the electroencephalogram signal is trained using a graph variational autoencoder to obtain the causal relationship of each association edge of the trained adjacency matrix, and the association degree of the channel corresponding to the association edge is determined. In an embodiment of the present specification, the GNN in the graph variational autoencoder is used for encoding, and the inner product in the graph variational autoencoder is used for decoding. When the shape similarity of the trained adjacency matrix and the distilled adjacency matrix reaches a preset similarity threshold, the value in the decoded adjacency matrix is the Granger causal relationship contribution of each association edge. According to the Granger causal relationship contribution of each association edge, the association degree between channels is determined, the electroencephalogram signal is collected by using the channel with a large association degree, and the evaluation cost is reduced; by collecting as few channels as possible, accurate prediction is obtained to assist in evaluating the movement intention of the user, and the accuracy of the evaluation is improved.

[0112] Figure 2 An embodiment of the present specification provides a structural schematic diagram of a movement intention evaluation system, which comprises:

[0113] The acquisition module 201 is configured to acquire original electroencephalogram signals of 62 channels of a user in need, and select channels of the original electroencephalogram signals to obtain first electroencephalogram signals.

[0114] The preprocessing and feature construction module 202 is configured to preprocess and construct features of the first electroencephalogram signals to obtain an adjacency matrix of the electroencephalogram signals.

[0115] The evaluation module 203 is configured to determine a causal relationship of association edges between channels by a channel determination model according to the adjacency matrix of the electroencephalogram signals, and assist in evaluating a movement intention of the user in need.

[0116] Optionally, the preprocessing and feature construction module 202 comprises:

[0117] The denoising sub-module is configured to resample and denoise the first electroencephalogram signals to obtain second electroencephalogram signals.

[0118] The feature extraction sub-module is configured to segment and extract frequency band power of the second electroencephalogram signals to obtain electroencephalogram channel point features.

[0119] A matrix determining sub-module is configured to construct an adjacency matrix of the electroencephalogram signals according to the electroencephalogram channel point features.

[0120] Optionally, the matrix determining sub-module comprises:

[0121] A first determining unit is configured to determine a geodesic distance value between two channels.

[0122] A second determining unit is configured to determine a coherence correlation value between electroencephalogram signals of two channels.

[0123] A weight value determining unit is configured to, when the coherence correlation value meets a preset correlation condition, obtain a weight value of a correlation edge between two channels according to the coherence correlation value and the geodesic distance value.

[0124] A matrix constructing unit is configured to construct an adjacency matrix of the electroencephalogram signals according to all the correlation edges and corresponding weight values.

[0125] Optionally, the evaluation module 203 comprises:

[0126] A distillation sub-module is configured to perform a distillation operation on the adjacency matrix of the electroencephalogram signals by using a correlation edge classification model.

[0127] A training sub-module is configured to train the adjacency matrix of the electroencephalogram signals according to the distilled adjacency matrix, so as to obtain a causal relationship of the correlation edges between channels.

[0128] Optionally, the distillation sub-module comprises:

[0129] A sorting unit is configured to sequentially delete a correlation edge of the adjacency matrix according to the correlation edge classification model, obtain a causal relationship contribution value of each correlation edge, delete the correlation edge with the smallest causal relationship contribution value according to a difference value of cross entropy, and sort the remaining correlation edges.

[0130] A weight adjusting unit is configured to sequentially delete the correlation edges from bottom to top according to the size of the causal relationship contribution values, adjust the weight values of the correlation edges according to the difference value of cross entropy, and the correlation edge classification model is a GNN graph neural network.

[0131] Optionally, the weight adjusting unit comprises:

[0132] A first weight adjusting sub-unit is configured to, if the cross entropy increases after the correlation edge is deleted, fill the difference value of the increased cross entropy into the distilled adjacency matrix as the weight value of the deleted correlation edge.

[0133] A second weight adjusting sub-unit is configured to, if the cross entropy decreases after the correlation edge is deleted, reset the weight value of the deleted correlation edge to 0.

[0134] Optionally, the training sub-module comprises:

[0135] The channel correlation degree determination unit is configured to train an adjacency matrix of the electroencephalogram signal by using a graph variational autoencoder, and obtain a causal relationship of each correlation edge of the trained adjacency matrix when a similarity degree between the trained adjacency matrix and the distilled adjacency matrix reaches a preset similarity threshold, and determine a correlation degree of a channel corresponding to the correlation edge.

[0136] The functions of the system of the embodiments of the present application have been described in the method embodiments described above, and thus the descriptions of the embodiments of the present application are not described in detail, and the relevant descriptions in the foregoing embodiments can be referred to, and will not be described herein.

[0137] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can 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-ROMs, optical storage, etc.) containing computer-usable program code.

[0138] The electronic device embodiments of the present application are described below, which can be regarded as the physical form of the embodiments of the method and device embodiments of the present application described above. The details described in the electronic device embodiments of the present application should be regarded as a supplement to the method or device embodiments described above; for the details not disclosed in the electronic device embodiments of the present application, the foregoing method or device embodiments can be referred to for implementation.

[0139] Figure 3 is a structural block diagram of an exemplary embodiment of a joint login system device according to the present application. Figure 3 The computer device shown is merely an example, and should not impose any limitation on the functions and use range of the embodiments of the present application.

[0140] As shown in Figure 3 The computer device 300 of the exemplary embodiment is in the form of a general-purpose data processing device. The components of the computer device 300 can include, but are not limited to, at least one processor 310, at least one memory 320, a network interface 330, a display unit 340, an input component 350, etc.

[0141] The memory 320 stores a computer-readable program, which can be a source program or a code of a read-only program. The program can be executed by the processor 310, so that the processor 310 performs the steps of various embodiments of the present application. For example, the processor 310 can execute the program as shown in Figure 1The steps shown.

[0142] The memory 320 can include a readable medium in the form of volatile storage such as random access memory (RAM) and / or cache memory, and can further include non-volatile storage such as read only memory (ROM). The memory 320 can also include a program / utility, having a set (at least one) of program modules that include, but are not limited to: an operating system, one or more application programs, other program modules, and program data, each or a combination thereof, which may

[0143] Also included in the memory 320, and also available to programs being executed by the processing unit 310, are program components, being a graphical environment, and operating system software. The bus 340 serves as the main information highway interconnecting the various hardware components and also serves as a path for data communication amongst the components.

[0144] The computer device 300 can also communicate with one or more external devices such as a keyboard, a pointing device, a display, a network device, a Bluetooth device, and so on. These external devices can be connected to the computer device 300 through the input / output interface 350. The input / output interface 350 can include, among other things, a keyboard, a pointing device, a display, a network device, a Bluetooth device, and so on. The communication can be via the network interface 330, and also via a network adapter with one or more networks, such as a local area network (LAN), a wide area network (WAN), and / or the public network, such as the Internet. The network adapter can communicate with the other modules of the computer device 300 through the bus. It should be appreciated that although not shown, other hardware and / or software modules can be used in the computer device 300, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data archival storage systems, etc.

[0145] Figure 4 is a schematic diagram of one computer readable medium embodiment of the present application. As Figure 4As shown, the computer program can be stored in one or more computer readable media. The computer readable media can be a readable signal medium or a readable storage medium. The readable storage medium, for example, can be, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or apparatus, or any suitable combination of the above. More specific examples (a non-exhaustive list) of the readable storage medium include an electrical connection having one or more wires, a portable disc, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. When the computer program is executed by one or more data processing devices, the computer readable medium can implement the above method of the present application, that is, acquiring 62-channel raw electroencephalogram signals of a demand user, selecting channels of the raw electroencephalogram signals to determine a first electroencephalogram signal, pre-processing and feature construction of the first electroencephalogram signal to obtain an adjacency matrix of the electroencephalogram signal, and determining a causal relationship of an associated edge between channels through a channel determination model according to the adjacency matrix of the electroencephalogram signal to assist in evaluating a motor intention of the demand user.

[0146] Through the above description of the embodiments, those skilled in the art can easily understand that the exemplary embodiments described in the present application can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solutions according to the embodiments of the present application can be embodied in the form of a software product, which can be stored in a computer readable storage medium (which can be a CD-ROM, a U disk, a mobile hard disk, etc.) or a network, and includes a plurality of instructions to make a data processing device (which can be a personal computer, a server, or a network device, etc.) execute the above method according to the present application.

[0147] The computer readable storage medium can include a data signal carried in the baseband or as a part of a carrier wave propagating through the program code readable. Such a propagating data signal can take various forms, including but not limited to electro-magnetic, optical or any suitable combination of the above. The readable storage medium can also be any readable medium other than the readable storage medium, which can send, propagate or transmit programs for use by or in connection with an instruction execution system, device or apparatus. The program code contained on the readable storage medium can be transmitted by any suitable medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the above.

[0148] The program code may be executed by one or more programmable processing devices, which can include microprocessors, digital signal processors (DSPs), central processing units (CPUs), graphics processing units (GPUs) or field programmable gate arrays (FPGAs).

[0149] In summary, the present application can be implemented by a computer program method, device, electronic device or computer readable medium. Some or all of the functions of the present application can be implemented in practice using a general-purpose data processing device such as a microprocessor or a digital signal processor (DSP).

[0150] The above-described specific embodiments have further detailed the purposes, technical solutions and beneficial effects of the present application. It should be understood that the present application is not inherently related to any specific computer, virtual device or electronic device, and various general-purpose devices can implement the present application. The above-described specific embodiments are merely examples of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method of evaluating a movement intention, characterized by, The method comprises the following steps: Collecting the original electroencephalogram signals of 62 channels of a demand user, selecting channels of the original electroencephalogram signals, and determining a first electroencephalogram signal; Preprocessing and feature construction are performed on the first electroencephalogram signal to obtain an adjacency matrix of the electroencephalogram signal; specifically, resampling and denoising are performed on the first electroencephalogram signal to obtain a second electroencephalogram signal; The second electroencephalogram is segmented, and total band power of each channel in the second electroencephalogram is extracted from each wave segment; each channel corresponds to a channel node in an adjacency matrix; the geodesic distance value and the coherence value in the adjacency matrix of the electroencephalogram are calculated through the coordinates of any two channels 、 ; the weight value of the associated edge between the two channels is obtained according to the average value of the coherence value and the geodesic distance value; the adjacency matrix of the electroencephalogram is constructed according to all associated edges and corresponding weight values; wherein the geodesic distance value is the inverse cosine value of the distance between the channels 、 , the coherence value is , 、 , and is the cross spectral density. Distillation is performed on the adjacency matrix of the electroencephalogram signal; the adjacency matrix of the electroencephalogram signal is trained according to the distilled adjacency matrix to obtain the causal relationship of each correlation edge of the trained adjacency matrix, determine the correlation degree of the channel corresponding to the correlation edge, and collect the electroencephalogram signal by using the channel with a large correlation degree, thereby assisting in evaluating the motion intention of the demand user through a small number of collected channels.

2. A method of assessing motor intent according to claim 1, wherein, The preprocessing and feature construction performed on the first electroencephalogram signal to obtain the adjacency matrix of the electroencephalogram signal further comprises: Segmentation and frequency band power extraction are performed on the second electroencephalogram signal to obtain electroencephalogram channel point features; The adjacency matrix of the electroencephalogram signal is constructed according to the electroencephalogram channel point features.

3. The method of assessing motion intent of claim 1, wherein, The distillation performed on the adjacency matrix of the electroencephalogram signal comprises: According to the correlation edge classification model, one correlation edge of the adjacency matrix is sequentially deleted to obtain the causal relationship contribution value of each correlation edge, the correlation edge with the minimum causal relationship contribution value is deleted according to the difference value of cross entropy, and the remaining correlation edges are sorted; According to the size of the causal relationship contribution value, the correlation edges are sequentially deleted from bottom to top, the weight value of the correlation edge is adjusted according to the difference value of cross entropy, and the correlation edge classification model is a GNN graph neural network.

4. A method of assessing movement intention according to claim 3, wherein, The adjustment of the weight value of the correlation edge comprises: If the cross entropy increases after the correlation edge is deleted, the difference value of the increased cross entropy is taken as the weight value of the deleted correlation edge and filled into the distilled adjacency matrix; If the cross entropy decreases after the correlation edge is deleted, the weight value of the deleted correlation edge is reset to 0.

5. The method of assessing motion intent of claim 1, wherein, The training of the adjacency matrix of the electroencephalogram signal according to the distilled adjacency matrix to obtain the causal relationship of each correlation edge of the trained adjacency matrix comprises: The adjacency matrix of the electroencephalogram signal is trained using a graph variational autoencoder, and when the similarity degree of the trained adjacency matrix and the distilled adjacency matrix reaches a preset similarity threshold, the causal relationship of each correlation edge of the trained adjacency matrix is obtained.

6. An evaluation system of a movement intention, characterized by, The method comprises the following steps: A collection module is configured to collect original electroencephalogram signals of 62 channels of a demand user, select channels of the original electroencephalogram signals, and determine a first electroencephalogram signal; A preprocessing and feature construction module is configured to preprocess and construct features of the first electroencephalogram signal to obtain an adjacency matrix of the electroencephalogram signal; specifically, resampling and denoising are performed on the first electroencephalogram signal to obtain a second electroencephalogram signal; The second electroencephalogram is segmented, and total band power of each channel in the second electroencephalogram is extracted from each wave segment; each channel corresponds to a channel node in an adjacency matrix; the geodesic distance value and the coherence value in the adjacency matrix of the electroencephalogram are calculated through the coordinates of any two channels 、 ; the weight value of the associated edge between the two channels is obtained according to the average value of the coherence value and the geodesic distance value; the adjacency matrix of the electroencephalogram is constructed according to all associated edges and corresponding weight values; wherein the geodesic distance value is the inverse cosine value of the distance between the channels 、 , the coherence value is , 、 , and is the cross spectral density. An evaluation module is configured to perform distillation on the adjacency matrix of the electroencephalogram signal; the adjacency matrix of the electroencephalogram signal is trained according to the distilled adjacency matrix to obtain the causal relationship of each correlation edge of the trained adjacency matrix, determine the correlation degree of the channel corresponding to the correlation edge, collect the electroencephalogram signal by using the channel with a large correlation degree, and assist in evaluating the motion intention of the demand user through a small number of collected channels.

7. An electronic device, comprising: The electronic device comprises: a processor; and a memory storing computer-executable instructions that, when executed, cause the processor to perform the method of any of claims 1-5.

8. A computer readable storage medium, wherein, The computer-readable storage medium stores one or more programs that, when executed by a processor, implement the method of any of claims 1-5.

Images