A small sample feature extraction method based on EEG motor imagery

Abstract

A small sample feature extraction method based on EEG motor imagery relates to the technical field of EEG signal processing, and the method comprises: step 1, data extraction; step 2, data segmentation; step 3, feature extraction; step 4, feature vector weight extraction. The present invention utilizes as little EEG information as possible with key features to perform motor imagery classification, thereby further improving the efficiency of EEG motor imagery classification.

Description

Technical Field

[0001] The present invention relates to the technical field of electroencephalogram (EEG) signal processing, and in particular to a small sample feature extraction method based on EEG motor imagery. Background Art

[0002] Traditional EEG feature extraction methods are difficult to take into account the time domain and frequency spectrum density. To avoid this problem, in the research of EEG machine learning, people began to use the end-to-end model to achieve classification and recognition to meet the high real-time requirements of brain-computer interaction. To achieve end-to-end machine learning of EEG. To achieve end-to-end machine learning of EEG, from raw data input to final output classifier, recognition results, etc., without too much manual setting, using layer neural network to couple any nonlinear function, this is not only to pre-process the input data, but also to classify it according to prior knowledge using traditional network learning features; it is also possible to embed neural pattern conversion algorithms in nonlinear learning networks, such as attention mechanism, back propagation, decision tree, greedy pursuit and other algorithms, making the fitting of data more flexible, it has high robustness and noise signal fault tolerance, and can completely approximate complex nonlinear relationships. However, since brain wave information is easily disturbed and lost by the external environment and individual activity patterns vary greatly, the classification accuracy of motor imagery needs to be improved.

[0003] Current machine learning methods for EEG motor imagery classification and recognition generally have the problem of reduced classification accuracy when the training data is limited. Summary of the invention

[0004] In view of the above defects or deficiencies in the prior art, the purpose of the present invention is to provide a small sample feature extraction method based on EEG motor imagery. The method uses as little EEG information with key features as possible to perform motor imagery classification, thereby further improving the efficiency of EEG motor imagery classification.

[0005] In order to solve the above problems, in a first aspect, the present invention provides a small sample feature extraction method based on EEG motor imagery, the method comprising:

[0006] Data extraction: extract three motor imagery acquisition channels, including C3, Cz, and C4. Use wavelet packets to extract nine rhythm segments, including 4-8 Hz, 8-12 Hz, ..., 36-40 Hz EEG signals, in order to extract the motor imagery information in the μ band and β band.

[0007] Data segmentation uses double input, sliding with a step size of 100ms, and a time window of 3s. The starting time point of the two double inputs is 100ms later than the previous one.

[0008] Feature extraction, each input data of the qth rhythm segment Where p∈[1,2] represents dual input, q∈N[1,9] represents dual input, N s is the number of sampling points, N s =t×F s , t is the split time window size, F s is the sampling rate, C p is the total number of samples of the pth motor imagery signal under the segmentation strategy, N is a natural number, and the signals of channels C3, Cz, and C4 correspond to X pq For each row of the two-dimensional matrix, the correspondence between the dual-input training sample and the label is:

[0009] Feature vector weight extraction.

[0010] In a second aspect, an embodiment of the present application provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in the embodiment of the present application when executing the computer program.

[0011] In a third aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program is used to:

[0012] When the computer program is executed by a processor, the method described in the embodiment of the present application is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The embodiments of the present invention will be described below with reference to the accompanying drawings, wherein

[0014] Figure 1 A schematic diagram of a small sample feature extraction process based on EEG motor imagery of the present application is shown;

[0015] Figure 2 A schematic diagram of the digital capsule layer routing results of the present application is shown. DETAILED DESCRIPTION

[0016] In order to make the invention purpose, technical process and technical innovation of the present invention more clearly described, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in each embodiment of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0017] To achieve the above objectives, the present invention provides a small sample feature extraction method based on EEG motor imagery. The main process is as follows: Figure 1 As shown, the method includes:

[0018] Data extraction: extract three motor imagery acquisition channels, including C3, Cz, and C4. Use wavelet packets to extract nine rhythm segments, including 4-8 Hz, 8-12 Hz, ..., 36-40 Hz EEG signals, in order to extract the motor imagery information in the μ band and β band.

[0019] Data segmentation uses double input, sliding with a step size of 100ms, and a time window of 3s. The starting time point of the two double inputs is 100ms later than the previous one.

[0020] Feature extraction, each input data of the qth rhythm segment Where p∈[1,2] represents dual input, q∈N[1,9] represents dual input, N s is the number of sampling points, N s =t×F s , t is the split time window size, F s is the sampling rate, C 5 is the total number of samples of the pth motor imagery signal under the segmentation strategy, and N is a natural number. The signals of channels C3, Cz, and C4 correspond to X pq For each row of the two-dimensional matrix, the correspondence between the dual-input training sample and the label is:

[0021] Feature vector weight extraction.

[0022] In particular, the feature vector weight extraction step includes: Fusion with the main capsule of the capsule network, Mq is the value input to the capsule network after feature extraction, P is the main capsule channel, S1×S2 is the main capsule neuron, and the number of output capsules is 16. Digital capsules encapsulate the given information and output it to data structures with different types of information. Each capsule has a logical unit that represents the existence of an entity, and its output represents different attributes of the same entity. In addition to the convolutional layer that clusters the information, each capsule also has a capsule self-rescue mechanism, that is, the capsule multiplies its own feature matrix by the learnable overall relationship coefficient weight to achieve feature enhancement and clustering operations on the previous layer of information. In order to route the output of each capsule to the upper layer capsule that receives similar voting clusters, the expectation maximization algorithm is used to iterate the formation of adjacent capsules by back-propagating feature information between layers to obtain a feature matrix that can distinguish information.

[0023] The digital capsule layer routing results are as follows Figure 2 As shown in the figure, the dynamic routing of the capsule network is an iterative algorithm that can encapsulate the given information and output it into different types of data structures. Each capsule has a logical unit that represents the existence of an entity, and its output has different attributes. In addition to containing a convolutional layer to cluster information, each capsule also has a capsule self-rescue mechanism, that is, the capsule multiplies its own feature matrix by the learnable overall relationship number weight to achieve the information reinforcement and clustering operation of the previous layer, in order to route the output of each capsule to the upper layer that receives similar voting clusters and use the expected clustering operation, in order to route the output of each capsule to the upper layer capsule that receives similar voting clusters, and use the expectation maximization algorithm to iteratively form adjacent capsules by back-propagating feature information between layers to obtain a feature matrix that can distinguish information.

[0024] After analysis, it is believed that for the binary classification 4D structure capsule, the capsule layer output category 16 is the most appropriate. It is also demonstrated through experiments, taking the number of iterations in the capsule layer as 2 and the main capsule layer as 128 as an example. When the capsule layer classification attribute is 16, the classification accuracy is higher than when the number of capsules is 8 and 32, and under the same capsule structure, it will be 4% higher, which shows the sensitivity of the capsule layer to the output dimension. For the output category of the capsule layer, the experience is the capsule dimension power of the number of categories, that is: ClassNumber Dmension , Dmension refers to the dimension of the capsule structure matrix. Compared with the traditional CNNCNN, the capsule network has achieved better results in classification accuracy. However, after training, the capsule network has achieved better results in classification accuracy.

[0025] In order to better understand the present invention, the above is described in detail in conjunction with specific examples of the present invention, but it is not intended to limit the present invention. Any simple modification made to the above embodiments based on the technical essence of the present invention still falls within the scope of the technical solution of the present invention.

[0026] As another aspect, the present application further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in the embodiment of the present application when executing the computer program.

[0027] As another aspect, the present application further provides a computer-readable storage medium, which may be a computer-readable storage medium included in the aforementioned device in the above-mentioned embodiment; or a computer-readable storage medium that exists independently and is not assembled into the device. The computer-readable storage medium stores one or more programs, and the aforementioned programs are used by one or more processors to execute the method described in the embodiment of the present application.

[0028] Any reference to the storage medium used in the embodiments of the present application may include non-volatile, volatile memory. Suitable non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), optical disk (including CD-ROM read-only memory (CD-ROM) and digital versatile disk (DVD)), or flash memory. Volatile memory may include random access memory (RAM), which is used as an external high-speed cache memory. As an illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM) and memory bus dynamic RAM (RDRAM).

[0029] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (Programmable Gate Array; hereinafter referred to as: PGA), a field programmable gate array (Field Programmable Gate Array; hereinafter referred to as: FPGA), etc.

[0030] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0031] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of software and firmware. If the integrated module is implemented in the form of software and firmware and sold or used as an independent product, it can be transferred from a storage medium or a network to a computer with a dedicated hardware structure for functional implementation.

[0032] It should also be noted that the steps of performing the above series of processes can naturally be performed in chronological order according to the order of description, but do not necessarily need to be performed in chronological order. Some steps can be performed in parallel or independently of each other.

[0033] Although the embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary and cannot be understood as limiting the present application. Any person skilled in the art can change, modify, replace and modify the above embodiments without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention is not limited thereto, and should be based on the scope defined by the claims.

Claims

1. A small sample feature extraction method based on EEG motor imagery, the method comprising: Data extraction: extract three motor imagery acquisition channels, including C3, Cz, and C4. Use wavelet packets to extract nine rhythm segments, including 4-8 Hz, 8-12 Hz, ..., 36-40 Hz EEG signals, in order to extract the motor imagery information in the μ band and β band. Data segmentation uses double input, sliding with a step size of 100ms, and a time window of 3s. The starting time point of the two double inputs is 100ms later than the previous one. Feature extraction, each input data of the qth rhythm segment Where p∈[1,2] represents dual input, q∈[1,9] represents dual input, N s is the number of sampling points, N s =t×F s , t is the split time window size, F s is the sampling rate, C p is the total number of samples of the pth motor imagery signal under the segmentation strategy. The signals of channels C3, Cz, and C4 correspond to X pq For each row of the two-dimensional matrix, the correspondence between the dual-input training sample and the label is: Feature vector weight extraction, the feature vector weight extraction step includes: Fusion with the main capsule of the capsule network, M q is the value input to the capsule network after feature extraction, P is the main capsule channel, S1×S2 is the main capsule neuron, and the number of output capsules is 16; the digital capsule encapsulates the given information and outputs it to a data structure with different types of information. Each capsule has a logical unit, indicating the existence of an entity, and its output represents different attributes of the same entity. In addition to containing a convolutional layer to cluster information, each capsule also has a capsule self-rescue mechanism, that is, the capsule multiplies its own feature matrix by the learnable overall relationship coefficient weight to achieve feature enhancement and clustering operations on the previous layer of information. In order to route the output of each capsule to the upper-layer capsule that receives similar voting clusters, the expectation maximization algorithm is used to iterate the formation of adjacent capsules by back-propagating feature information between layers to obtain a feature matrix that can distinguish information.

2. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to claim 1 is implemented.

3. A computer-readable storage medium having a computer program stored thereon, the computer program being used to: When the computer program is executed by a processor, the method according to claim 1 is implemented.

Images