Multi-channel electroencephalogram acquisition system based on Hessian matrix analysis and signal processing method

Abstract

The invention provides a multi-channel electroencephalogram acquisition system based on Hessian matrix analysis and a signal processing method. The system comprises a multi-channel brain electrode array, a signal acquisition unit, a signal processing unit and a visual display unit, the multi-channel brain electrode array comprises a central electrode and at least eight surrounding electrodes surrounding the central electrode to form a Hessian electrode array; brain potential signals of at least nine electrode points of the Hessian electrode array are synchronously collected, and a Hessian matrix representing the local curvature of a potential field at the position of a central electrode is constructed; and performing eigenvalue decomposition on the Hessian matrix to obtain at least one main eigenvalue and a main eigenvector corresponding to the main eigenvalue, and calculating a main direction angle according to the main eigenvalue to represent the intensity of the electroencephalogram activity collected at the position of the central electrode and the direction of an electroencephalogram signal source. According to the method, accurate analysis of the directivity of the electroencephalogram activity source can be achieved, good spatial filtering capacity and high signal-to-noise ratio are achieved, and weak local electroencephalogram signals can be captured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wearable and implantable brain-computer interface, and particularly relates to a multi-channel electroencephalogram acquisition system and signal processing method based on a Hessian matrix solution. BACKGROUND

[0002] A brain-computer interface (BCI) is a direct communication path established between the brain and external devices. It converts the neural signals generated by brain activity into control commands through acquisition and analysis, realizing the interaction between the brain and computers or intelligent devices. According to the contact mode of the electrode and the brain, the brain-computer interface is mainly divided into two categories: one is an invasive brain-computer interface in which the electrode is implanted into the cerebral cortex or attached to the cerebral cortex or the dura mater, which has high signal quality but has the risk of surgery; the other is a non-invasive brain-computer interface in which the electrode is placed on the scalp, which is represented by electroencephalogram (EEG) technology, has the advantages of safety and convenience, and is widely used.

[0003] However, whether it is scalp EEG or the electroencephalogram signals collected in the cerebral cortex or outside the dura mater, they all face a common physical challenge - volume conduction effect. The conductive properties of biological tissues make the electric field generated by neuronal activity diffuse to the surrounding, resulting in the signal recorded on the electrode being a mixed superposition of a large number of neuronal activities and noise. This effect seriously reduces the spatial resolution of the electroencephalogram signal, making it difficult to accurately determine the true source of the signal.

[0004] In order to overcome the volume conduction effect, Laplacian electrode technology has been developed in the prior art. By calculating the second-order spatial derivative of the electric potential between a central electrode and its surrounding electrodes (i.e. the Laplacian value), this technology can effectively filter out the common-mode signals from a distant source, thereby acting as a "spatial filter". This significantly improves the sensitivity to local electroencephalogram activity under the electrode and enhances the spatial resolution of the signal.

[0005] Although the Laplacian electrode has obvious advantages in improving spatial resolution, it has certain limitations: the Laplacian electrode can only solve the problem of collecting local signal sources, i.e. providing a scalar value to reflect the intensity of local activity. When there are multiple or moving signal sources around the electrode, the traditional Laplacian electrode cannot determine the accurate direction of the signal source and cannot realize the directional acquisition of electroencephalogram activity. The lack of this information limits our understanding of brain neurodynamics (such as the path of information flow).

[0006] According to the search, the Chinese invention application with publication number CN119679431A discloses a wireless high-density electroencephalogram and electromyogram mixed synchronous acquisition monitoring system, which comprises a high-density electroencephalogram electrode array, a signal acquisition and transmission unit, a base station system and an upper computer system, can realize synchronous acquisition, wireless transmission and preprocessing (such as filtering and noise reduction) of high-density electroencephalogram signals, and extract time-frequency domain features of electroencephalogram signals through a decoding algorithm module, for motion intention recognition and other applications. However, it mainly solves the problems of synchronous acquisition of electroencephalogram and electromyogram mixed signals, limited channel number and poor signal quality, and does not mention the signal source direction recognition requirement.

[0007] Therefore, there is an urgent need in the art for a new electroencephalogram signal acquisition and processing technology, which can not only suppress far-end interference and improve spatial resolution, but also further analyze the directionality information of local electroencephalogram activity, which will provide technical support for neuroscience research and high-precision brain-computer interaction paradigm research. SUMMARY

[0008] In view of the limitations in the prior art, the present application provides a multi-channel electroencephalogram acquisition system and signal processing method based on Hessian matrix analysis, which solves the problem that the signal source direction cannot be determined under volume conduction effect, and realizes the directionality analysis of local electroencephalogram activity.

[0009] In a first aspect of the present application, a multi-channel electroencephalogram acquisition system based on Hessian matrix analysis is provided, comprising: a multi-channel electroencephalogram electrode array, comprising at least one physical electrode group, the physical electrode group comprising a center electrode and at least eight surrounding electrodes, forming a square or circular Hessian electrode array; a signal acquisition unit, which synchronously acquires and digitizes at least nine channels of electroencephalogram potential signals in the multi-channel electroencephalogram electrode array; a signal processing unit, which constructs a Hessian matrix representing the local curvature of the potential field at the position of the center electrode based on the electrode signals acquired by the signal acquisition unit and the predetermined geometric relationship between them, performs eigenvalue decomposition on the Hessian matrix to obtain at least one principal eigenvalue and a principal eigenvector corresponding thereto, and calculates a principal direction angle according to the principal eigenvalue to represent the intensity of the acquired electroencephalogram activity and the direction of the electroencephalogram signal source at the position of the center electrode; a visual display unit, which visualizes the electroencephalogram potential acquired by the signal acquisition unit and indicates the direction of the electroencephalogram signal source according to the results of the signal processing unit.

[0010] Optionally, the multi-channel electroencephalogram electrode array, wherein: the square-hexagon electrode array is a square-hexagon matrix of three rows and three columns, and the lateral and longitudinal distances between adjacent electrodes are both preset values h; and the circular-hexagon electrode array is a circular-hexagon array composed of one central electrode and eight surrounding electrodes, and the eight surrounding electrodes are located on a circumference with the central electrode as the center and h as the radius.

[0011] Optionally, the signal processing unit further comprises: determining the local source intensity of the electroencephalogram activity based on the absolute value of the main eigenvalue; and / or determining the local source propagation direction or the principal axis direction of the electroencephalogram activity based on the direction of the main eigenvector.

[0012] Optionally, the multi-channel electroencephalogram electrode array comprises a plurality of preset physical electrode groups, each of which is composed of at least nine physical electrodes selected from more than nine physical electrodes, each of which constitutes a nine-point electrode array with a different physical electrode as the center, and the signal processing unit independently performs hexagon matrix calculation and eigenvalue decomposition on each of the physical electrode groups to obtain the electroencephalogram activity direction vector field covering the target acquisition region of the head. Optionally, the target acquisition region comprises a region that can be arranged on the scalp to realize non-invasive electroencephalogram acquisition, a region that can be arranged epidurally to realize semi-invasive electroencephalogram acquisition, and a region that can be arranged on the cerebral cortex to realize invasive electroencephalogram acquisition.

[0013] Optionally, the visualization display unit comprises: displaying a direction arrow determined by the main eigenvector at the position of each of the central electrodes on the electroencephalogram topographic map, and the visual attributes of the arrow, including length, color or thickness, are determined by the main eigenvalue.

[0014] In a second aspect, the application provides a multi-channel electroencephalogram signal processing method based on hexagon matrix analysis, comprising: defining at least one hexagon electrode array on a target acquisition region, wherein the hexagon electrode array comprises one central electrode and at least eight surrounding electrodes around the central electrode; extracting the electroencephalogram potential signals of at least nine channels synchronously acquired by the hexagon electrode array; calculating a hexagon matrix representing the local curvature of the potential field at the position of the central electrode using the known geometric positions between the electrodes of the hexagon electrode array and the acquired electroencephalogram potential signals; performing eigenvalue decomposition on the hexagon matrix to obtain at least one main eigenvalue and a main eigenvector, which are used to determine the intensity of the electroencephalogram signal source and the main direction of the signal source acquired at the position of the central electrode.

[0015] Optionally, the defining at least one hexagon electrode array on a target acquisition region comprises: from a cap containing more than nine electrodes, one or more of said Hessian electrode arrays are defined virtually in software, wherein each virtual Hessian electrode array is centered on a different physical electrode.

[0016] Optionally, said using the acquired electroencephalography signals and the known geometry between the electrodes of said Hessian electrode array to calculate a Hessian matrix representing the local curvature of the potential field at the position of said center electrode comprises: i) defining the electroencephalography signal of said center electrode as V C ; defining the electroencephalography signals of four surrounding electrodes located in the East (E), West (W), South (S), and North (N) directions as V E , V W , V S , and V N , respectively; and defining the electroencephalography signals of four surrounding electrodes located in the North-East (NE), North-West (NW), South-West (SW), and South-East (SE) directions as V NE , V NW , V SW , and V SE , respectively; said surrounding electrodes being at a distance h from said center electrode; ii) calculating the second order partial derivative of the potential field along a first direction, i.e. the x-axis, V xx : ; calculating the second order partial derivative of the potential field along a second direction, i.e. the y-axis, V yy : ; and calculating the mixed second order partial derivative V xy : for a square Hessian electrode array, V xy is: ; for a circular Hessian electrode array, V xy is: ; iii) constructing the calculated V xx , V yy , and V xy into a second order symmetric Hessian matrix, said Hessian matrix at each time point t having the form: .

[0017] Optionally, the eigenvalue decomposition of the Hessian matrix is performed to obtain at least one principal eigenvalue and a principal eigenvector, so as to determine the intensity of the brain electrical signal source collected at the central electrode position and the principal direction of the signal source, comprising: i) for each time point Hessian matrix H(V(t)), solve its eigenvalue λ and eigenvector v, the characteristic equation is: ; Solve the characteristic equation to obtain two eigenvalues: ; ii) λ1 is λ max , which is the largest eigenvalue, representing the largest principal curvature; λ2 is λ min , which is the smallest eigenvalue, representing the smallest principal curvature; iii) substitute λ max into the characteristic equation: ; The corresponding principal eigenvector v max is obtained: ; iv) using the components v x and v y of the principal eigenvector, the principal direction angle θ of the brain electrical signal source is calculated: .

[0018] Optionally, the method further comprises a calibration module or a calibration step, specifically: before the Hessian matrix calculation, the brain electrical potential signals of the at least nine electrodes are subjected to inter-channel gain calibration and / or time delay calibration to compensate for the amplitude and phase inconsistency of each channel due to physical differences.

[0019] Compared with the prior art, the present application can achieve at least one of the following beneficial effects: The multi-channel brain electrical signal acquisition system based on Hessian matrix analysis provided by the present application has good spatial filtering capability and high signal-to-noise ratio, and can capture weak local brain electrical signals. In the process of calculating the Hessian matrix, the essence is to perform a second-order differential operation on the potential field, which better inherits and develops the idea of Laplace filtering. It can efficiently suppress far-field noise caused by volume conduction effect and spatially widely distributed common-mode interference, greatly improving the signal-to-noise ratio. Therefore, the present application can stably and reliably extract the local brain electrical signals of the brain region where the electrode is located in the noisy background activity.

[0020] The multi-channel electroencephalogram signal processing method based on the analysis of the Hessian matrix provided in the application can realize accurate analysis of the directionality of the electroencephalogram activity source. The existing technology such as the Laplace electrode can only provide a scalar value to reflect the presence or absence or strength of the local signal. The application can determine the principal axis direction of the local electroencephalogram activity source by calculating the Hessian matrix of the potential field and decomposing the principal eigenvector thereof. This makes it possible to study the complex neural dynamics such as the information flow of the brain, the propagation path of the neural oscillation wave, and provides a new dimension for the analysis of brain function.

[0021] The multi-channel electroencephalogram signal processing method based on the analysis of the Hessian matrix provided in the application has high flexibility and can be compatible with many existing high-density electroencephalogram acquisition devices. The selection of at least 9 electrode points is not limited to a specific physical hardware design, and the core method can be realized in a software manner to virtually and dynamically define a plurality of Hessian electrode arrays on any standard and sufficiently dense multi-channel electroencephalogram acquisition array. This means that the user does not need to customize new electrodes, and can upgrade the existing high-density electroencephalogram acquisition system through software to convert it into an electroencephalogram acquisition system capable of Hessian matrix analysis, greatly reducing the technical application threshold and maximizing the value of existing scientific research and clinical equipment. BRIEF DESCRIPTION OF DRAWINGS

[0022] Other features, objects and advantages of the application will become more apparent from the following detailed description of non-limiting embodiments with reference to the attached drawings. The illustrative embodiments of the application and their description serve to explain the application. In the drawings: Figure 1 The schematic diagram of the architecture of the multi-channel electroencephalogram acquisition system based on the analysis of the Hessian matrix provided in an embodiment of the application is shown; Figure 2 The schematic diagram of two preferred geometric configurations of the Hessian electrode array in an embodiment of the application is shown, wherein Fig. (a) shows a square array configuration of the electrode array, and Fig. (b) shows a circular array configuration of the electrode array; Figure 3 The simulation result of the response and extraction of a single target square wave signal by the Hessian electrode array in a strong interference background in an embodiment of the application is shown, wherein Fig. (a) is a waveform diagram of a sine interference signal and a single square wave signal to be measured; Fig. (b) is a curve diagram of the eigenvalues solved in real time; Fig. (c) is a global view of the visual display unit interface; and Fig. (d) is a local view of the visual display unit interface; Figure 4Fig. 1 is a schematic diagram of simulation results of responses of a Hansen electrode array in an embodiment of the present application to multiple target square wave signals in a strong interference background, wherein Fig. (a) is a waveform diagram of a sine interference signal and multiple square wave signals to be measured; Fig. (b) is a curve diagram of eigenvalues solved in real time; Fig. (c) is a global view of an interface of a visual display unit; and Fig. (d) is a partial view of the interface of the visual display unit; Figure 5 Fig. 2 is a schematic diagram of application of an overlapping Hansen electrode array constructed by multiple physical electrode groups and parallel processing to obtain a brain electrical activity directional vector field covering a target region in an embodiment of the present application; Figure 6 Fig. 3 is a flowchart of a multi-channel electroencephalogram signal processing method based on Hansen matrix analysis in an embodiment of the present application; Reference signs: 101 - multi-channel electroencephalogram electrode array, 102 - signal acquisition unit, 103 - signal processing unit, 104 - visual display unit; 201 - center electrode point of a square Hansen array, 202 - surrounding electrode point of the square Hansen array, 203 - center electrode point of a circular Hansen array, 204 - surrounding electrode point of the circular Hansen array; 301 - directional arrow of a principal eigenvector determination, 302 - position of a source to be measured, 303 - waveform of a sine interference signal, 304 - waveform of a single square wave pulse signal to be measured, 305 - curve of a maximum eigenvalue λ max , and 306 - position of an interference source in a global view interface of a visual display unit; 401 - waveforms of multiple square wave pulse signals to be measured, and 402 - curve of a maximum eigenvalue λ max ; 501, 502, and 503 - center electrode points of different square Hansen electrode arrays virtually defined in a software manner. DETAILED DESCRIPTION

[0023] To make the objectives, technical solutions, and advantages of the present application clearer, the present application will be described in detail below with specific embodiments. The following embodiments will help those skilled in the art further understand the present application, but do not limit the present application in any form. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These all belong to the protection scope of the present application.

[0024] In the following embodiments of the present application, a multi-channel electroencephalogram acquisition system and a signal processing method based on Hansen matrix analysis are provided, aiming to solve the problem that it is difficult to analyze the directionality of local brain electrical activity sources in the prior art. Meanwhile, the system has high spatial resolution and high signal-to-noise ratio, and the signal processing method is compatible with many existing high-density electroencephalogram acquisition devices.

[0025] Referring to Figure 1 The multi-channel electroencephalogram acquisition system based on the analysis of the Hessian matrix provided in the embodiments comprises a multi-channel electrode array 101, a signal acquisition unit 102, a signal processing unit 103, and a visual display unit 104.

[0026] The multi-channel electrode array 101 comprises at least one physical electrode group, which comprises a center electrode and at least eight surrounding electrodes, forming a square or circular Hessian electrode array. The signal acquisition unit 102 synchronously acquires and digitizes at least nine-channel electroencephalogram potential signals in the multi-channel electrode array. The signal processing unit 103 constructs a Hessian matrix representing the local curvature of the potential field at the position of the center electrode based on the electrode signals acquired by the signal acquisition unit and the predetermined geometric relationship therebetween, performs eigenvalue decomposition on the Hessian matrix to obtain at least one principal eigenvalue and a principal eigenvector corresponding thereto, and calculates a principal direction angle based on the principal eigenvalue to represent the intensity of the acquired electroencephalogram activity and the direction of the electroencephalogram signal source at the position of the center electrode. The visual display unit 104 visualizes the electroencephalogram potential acquired by the signal acquisition unit and indicates the direction of the electroencephalogram signal source based on the result of the signal processing unit.

[0027] The multi-channel electroencephalogram acquisition system of the above embodiments of the present application, by using the Hessian electrode array and constructing the Hessian matrix and then performing eigenvalue decomposition and other processing, can be applied to electroencephalogram acquisition on the cerebral cortex, epidural electroencephalogram acquisition, or scalp electroencephalogram acquisition, and can not only have good suppression ability for far-field noise and common-mode interference, but also can analyze the directionality information of local electroencephalogram activity in a complex electrophysiological environment. Meanwhile, the signal processing unit 103 of the above embodiments of the present application can realize accurate analysis of the directionality of the electroencephalogram activity source, and by calculating the Hessian matrix of the potential field and decomposing the principal eigenvector thereof, the principal axis direction of the local electroencephalogram activity source can be determined.

[0028] In some embodiments of the present application, the multi-channel electrode array 101 is used to non-invasively, semi-invasively, or invasively acquire electroencephalogram signals. It can be arranged on the scalp to realize non-invasive electroencephalogram acquisition (EEG), arranged epidurally to realize semi-invasive electroencephalogram acquisition, or directly arranged on the cerebral cortex to realize invasive electroencephalogram acquisition.

[0029] Specifically, the multichannel EEG array consists of multiple physical electrode groups, each containing at least nine physical electrodes, or at least nine electrodes selected from a larger pool of physical electrodes. Each group forms a nine-point electrode array centered around a different physical electrode, known as a Hessian electrode array. The signal processing unit independently performs Hessian matrix calculation and eigenvalue decomposition on each physical electrode group to obtain the EEG activity direction vector field covering the target acquisition area of ​​the head.

[0030] Reference Figure 2 There are two preferred geometric configurations for the Hessian electrode array: Square Hessian array, such as Figure 2 As shown in (a), the nine electrodes are arranged in a three-row, three-column square matrix, with the central electrode point being 201 and the surrounding electrode points being 202. The horizontal and vertical spacing between adjacent electrode points is a preset value h. In a specific embodiment, the value of h is preferably between 2 mm and 10 mm to balance spatial resolution and signal acquisition range. Each individual electrode point can be designed as circular, square, or other shapes, and there is a preferred relationship between its characteristic dimensions (such as diameter or side length) d and the spacing h, for example, d=h / 2 or d=h / 3, to ensure sufficient electrical insulation between the electrodes while maximizing the contact area of ​​a single electrode.

[0031] Circular Hessian array, such as Figure 2 As shown in (b): it consists of a central electrode point 203 and eight surrounding electrode points 204, wherein the eight surrounding electrode points are conformally arranged on a circle with the central electrode point 203 as the center and a preset value h as the radius. Similarly, the value of h is preferably in the range of 2 mm to 10 mm. Each individual electrode point can be designed as a circle, square or other shape, and there is a preferred relationship between its characteristic dimension (such as diameter or side length) d and the spacing h, for example, d=h / 2 or d=h / 3.

[0032] In specific embodiments of this application, any of the preferred geometric configurations described above can be used. Of course, in other embodiments, Hessian electrode arrays with geometric configurations different from those described above can also be used.

[0033] To ensure good electrical conductivity, biocompatibility and low contact impedance, the electrode preferably but not limited to adopts silver / silver chloride (Ag / AgCl) material, gold material, platinum-iridium alloy material, conductive carbon material such as graphene, conductive hydrogel material, etc. In the manufacturing process, it is preferred but not limited to screen printing Ag / AgCl conductive ink on the flexible substrate to form electrode contacts and leads, or using sputtering or electroplating process to deposit metal layer (such as gold or platinum-iridium alloy) on the flexible or rigid substrate, or using conductive carbon materials such as graphene, conductive hydrogel materials, etc. to prepare electrodes by patterning process. In addition, in order to further reduce the skin contact impedance, the electrode surface can be designed as a microneedle structure, a corrugated structure, etc., a concentric ring structure, etc., or electroplated with platinum black, iridium oxide, etc. on the metal layer. The leads of all electrodes are finally collected to a connector interface for connection with the signal acquisition unit 102.

[0034] Referring to Figure 1 In the above multi-channel electroencephalogram acquisition system based on the analysis of the Hessian matrix, the signal acquisition unit 102 is electrically connected with the multi-channel electroencephalogram electrode array 101, and is responsible for synchronously collecting the electroencephalogram potential signals of the nine electrodes of the Hessian electrode array. In some embodiments, the collected analog signals are amplified by the built-in preamplifier, filtered by the band-pass filter, and finally converted into digital signals by the analog-to-digital converter (ADC) for subsequent processing.

[0035] In some embodiments of the present application, the signal acquisition unit 102 and the multi-channel electroencephalogram electrode array 101 are electrically connected through a low-noise shielded cable or a flexible printed circuit (FPC) to perform high-fidelity synchronous collection of weak electroencephalogram potential signals from the nine electrodes.

[0036] Specifically, in the signal acquisition unit 102, each electrode channel is configured with a set of independent signal conditioning circuit, and the specific process is as follows: Preamplification: The analog electroencephalogram signals of each channel are first input into a high-performance instrument amplifier. The amplifier has very high input impedance (>1 GΩ) to match the high impedance characteristics of the electrode, and has a high common-mode rejection ratio to effectively suppress 50 Hz or 60 Hz power frequency interference and other common-mode noise from the human body. The gain of the amplifier is usually set between 100 and 1000 times to amplify the original microvolt (uV) level signal to the millivolt (mV) level.

[0037] Filtering: The amplified signal is then passed through a hardware band-pass filter. The filter is composed of a high-pass filter and a low-pass filter in cascade. The cut-off frequency of the high-pass filter is usually set at 0.1 Hz or 0.5 Hz to eliminate the baseline drift caused by electrode polarization and unstable skin contact. The cut-off frequency of the low-pass filter is set according to the application requirements, for example, it can be set at 30 Hz when collecting slow wave sleep, or 100 Hz or higher when collecting high frequency band, the main purpose is to meet the Nyquist theorem, to prevent signal aliasing.

[0038] Analog-to-digital conversion: The conditioned analog signal is finally sent to a high-precision analog-to-digital converter (ADC). The resolution of the ADC is preferably 24 bits, 16 bits or 12 bits to ensure fine quantization of weak brain electrical signals; its sampling rate is determined according to the setting of the low-pass filter, usually 250 Hz, 500 Hz, 1 kHz, 5 kHz, 20 kHz or 30 kHz, to ensure the complete information of the signal is collected. The ADC sampling of all nine channels must be strictly synchronized by a unified master clock to ensure that the time alignment accuracy between channels reaches the microsecond level.

[0039] In some embodiments of the present application, the signal processing unit 103 is the core computing module, which can be physically implemented as a high-performance digital signal processor (DSP), a field programmable gate array (FPGA) that can realize parallel computing, or a special software module running on a general-purpose computer (host computer). The unit receives nine-channel synchronous digital signal streams from the signal acquisition unit 102, and its main function configuration is: based on the nine Heisenberg electrode potential signals and their predetermined geometric relationship, real-time calculation of the Heisenberg matrix representing the local curvature of the potential field at the center electrode position, eigenvalue decomposition of the Heisenberg matrix to obtain the main eigenvalue (λ max ) and the main eigenvector (v max ) corresponding to it. Further, based on the absolute value of the main eigenvalue λ max , the local source strength of the brain electrical activity is determined; and / or, based on the direction of the main eigenvector v max , the main direction angle θ is calculated to determine the local source propagation direction or main axis direction of the brain electrical activity.

[0040] The signal processing unit 103 in the embodiments of the present application has good spatial filtering capability and high signal-to-noise ratio, and can capture weak local brain electrical signals. In the process of calculating the Heisenberg matrix, it is essentially a second-order differential operation on the potential field, which can effectively suppress the far-field noise caused by volume conduction effect and the widely distributed common-mode interference in space, greatly improving the signal-to-noise ratio, so it can stably and reliably extract the local brain electrical signals of the brain region where the electrode is located in the noisy background activity.

[0041] In some embodiments of the present application, a visualization display unit 104, which is usually implemented on the host computer, is used to present the analysis results to the user in an intuitive way. It can display the acquired raw / processed electroencephalogram waveforms in real time, and visualize the direction information calculated by the signal processing unit. Referring to Figure 3 , the unit can display a direction arrow 301 determined by the principal eigenvector at the position of the central electrode on the electroencephalogram topography or electrode topology, pointing to the position 302 of the source to be detected. The visual attributes of the arrow, such as length, color or thickness, can be associated with the size of the principal eigenvalue, so as to simultaneously show the intensity and direction of the local electroencephalogram activity.

[0042] To demonstrate the beneficial effects of the multi-channel electroencephalogram acquisition system based on the analysis of the Hessian matrix in the above embodiments of the present application, a simulation experiment was conducted.

[0043] Referring to Figure 3 , the simulation simulates the scenario of detecting a weak single square wave pulse signal 304 to be detected in the background of a strong sinusoidal interference signal 303, as shown in Figure 3 (a). As shown in Figure 3 (b), the curve 305 of the maximum eigenvalue λ max calculated in the embodiments of the present application can accurately and with high signal-to-noise ratio indicate the occurrence time of the pulse event. Figure 3 (c) shows the global view interface of the visualization display unit 104, and 306 represents the position of the interference source. As can be seen from the figure, the interference source randomly appears at different positions in the visualization interface.

[0044] Similarly, referring to Figure 4 , the simulation simulates the scenario of detecting multiple square wave pulse signals 401 to be detected in the background of a strong sinusoidal interference signal, as shown in Figure 4 (a). As shown in Figure 4 (b), the curve 402 of the maximum eigenvalue λ max calculated in the embodiments of the present application can clearly respond to each square wave event. Figure 4 (c) and Figure 4 (d) are the global view interface and the local view interface of the visualization display unit, respectively, showing the random position of the interference source and the direction arrow pointing to the source to be detected.

[0045] Referring to Figure 5The preferred embodiment of the present application supports the monitoring of directional field of large brain areas. In a multi-channel brain electrode array (such as an EEG cap) containing more than nine physical electrodes, a plurality of overlapping physical electrode groups, such as 501, 502 and 503, can be defined virtually in software, each group constituting a Hessian electrode array centered on different physical electrodes. The signal processing unit 103 performs parallel Hessian calculation and eigenvalue decomposition on each group to obtain a brain electrical activity directional vector field covering the target acquisition area. This enables researchers to observe the dynamic propagation path of neural activity in real time.

[0046] Based on the same technical concept, in another embodiment of the present application, a multi-channel brain electrical signal processing method based on Hessian matrix analysis is provided, which aims to convert the acquired original multi-channel brain electrical data into local source intensity and directional information with clear physical meaning through a series of calibration, preprocessing and core algorithms.

[0047] Specifically, with reference to Figure 6 In a more preferred embodiment, the multi-channel brain electrical signal processing method based on Hessian matrix analysis includes the following steps S601-S609, which are described in detail as follows.

[0048] Step S601: Defining a Hessian electrode array: One or more Hessian electrode arrays are defined from a multi-channel brain electrode array. Specifically, at least one Hessian electrode array is defined on a target acquisition area, including: from a multi-channel brain cap containing more than nine electrodes, one or more Hessian electrode arrays are defined virtually in software, wherein each virtual Hessian electrode array is centered on a different physical electrode. As described above, the target acquisition area can include an area that can be placed on the scalp to achieve non-invasive brain electrical acquisition, an area that can be placed epidurally to achieve semi-invasive brain electrical acquisition, or an area that can be placed on the cerebral cortex to achieve invasive brain electrical acquisition.

[0049] In this embodiment, the selection of at least 9 electrode points is not limited to a specific physical hardware design, and can be implemented in software. A plurality of Hessian electrode arrays can be defined virtually and dynamically on any standard, sufficiently dense (channel number greater than 9) multi-channel brain electrical acquisition array, so that users do not need to customize new electrodes, but can upgrade the existing high-density brain electrical acquisition system through software to convert it into a brain electrical acquisition system capable of Hessian matrix analysis, greatly reducing the technical application threshold.

[0050] Step S602: Extracting brain electrical signals of the Hessian electrode array: Extract the brain electric potential signals collected by the nine channels in step S601. Specifically, refer to the description of the signal acquisition unit 102 above to obtain the brain electric potential signals collected by the nine channels of the Hessen electrode array.

[0051] Step S603: Gain calibration and / or time delay calibration.

[0052] This step is an optional step. As a preferred embodiment, before formal electroencephalogram data analysis, a one-time system calibration process is first performed to compensate for the inherent differences between physical channels. This step is preferably performed under controlled conditions, such as in a PBS conductive solution or through a dedicated multi-channel signal simulator.

[0053] Specifically, the calibration includes the following: Amplitude calibration: A reference signal of a known amplitude and frequency (e.g., a 10 mV peak-to-peak sine wave at 20 Hz) is simultaneously input to all nine electrode channels. After collecting a stable signal, the root mean square (RMS) value or peak-to-peak value of each channel signal is calculated. Taking the center electrode or other specified channel as a reference (with a calibration coefficient of 1), the amplitude ratio of the other eight channels relative to the reference channel is calculated to obtain a set of nine amplitude calibration coefficients Cal_Amp (including Cal_Amp_1 to Cal_Amp_9). These coefficients will be permanently stored and used for subsequent data processing to normalize the amplitude of each channel.

[0054] Delay calibration: Similarly, using the reference signal described above (e.g., a 10 mV peak-to-peak sine wave at 20 Hz), the time offset corresponding to the peak of the cross-correlation function between each channel signal and the reference channel signal is found by calculating the cross-correlation function. This offset is the phase delay Cal_Delay (including Cal_Delay_1 to Cal_Delay_9) of each channel relative to the reference channel. This set of delay data is also stored and used for subsequent accurate time alignment of each channel signal.

[0055] Data reading and applying calibration coefficients: Read the original data file and match the corresponding electrode order (e.g., Center, E, NE, N, NW, W, SW, S, and SE) according to the relative positions of the electrodes. For each nine-channel data vector V_raw(t) at sampling time point t, perform amplitude correction: multiply each channel data in V_raw(t) by its corresponding amplitude calibration coefficient Cal_Amp point by point, i.e., ; The nine-channel data vector V_raw(t) of each sampling time point t is time-aligned: according to the phase delay Cal_Delay of each channel, the time series of each channel is shifted by a resampling interpolation method, to ensure that the signals of all channels are accurately aligned on the time axis, and the nine-channel calibrated data V_cal(t) used for calculation is obtained.

[0056] Step S604: Calculate the second-order partial derivative V xx , yy and V xy : This step is performed in real time in the signal processing unit 103.

[0057] The electroencephalogram signal of the central electrode is defined as V C ; the electroencephalogram signals of the four surrounding electrodes located in the east (E), west (W), south (S), and north (N) directions are defined as V E , V W , V S , and V N , respectively; and the electroencephalogram signals of the four surrounding electrodes located in the northeast (NE), northwest (NW), southwest (SW), and southeast (SE) directions are defined as V NE , V NW , V SW , and V SE , respectively; the distance between the surrounding electrodes and the central electrode is h; The second-order partial derivative V xx of the potential field along the first direction (x-axis) is calculated by combining the electroencephalogram signals of the electrodes: ; The second-order partial derivative V yy of the potential field along the second direction (y-axis) is calculated: ; and the mixed second-order partial derivative V xy is calculated. For a square-shaped hexagonal electrode array, V xy is: ; For a circular hexagonal electrode array, V xy is: ; Step S605: Construct the Hessian matrix H: This step is performed in real time in the signal processing unit 103.

[0058] The calculated V xx , V yy , and V xyThe Hessian matrix is constructed as a symmetric 2nd order matrix, at each time point t, the Hessian matrix is in the form of: .

[0059] Step S606: Eigenvalue decomposition of the Hessian matrix H is performed: This step is also performed in real time in the signal processing unit 103.

[0060] For each time point's Hessian matrix H(V(t)), its eigenvalues λ and eigenvectors v are solved. The eigen equation is: ; Solving the eigen equation, two eigenvalues are obtained: ; Step S607: Extract the principal eigenvalue λ max and the principal eigenvector v max : This step is performed in real time in the signal processing unit 103.

[0061] λ1is λ max , the largest eigenvalue, representing the largest principal curvature; λ2is λ min , the smallest eigenvalue, representing the smallest principal curvature.

[0062] Substitute λ max into the eigen equation: ; The corresponding principal eigenvector v max is obtained: ; Step S608: Calculate the principal direction angle according to the principal eigenvalue λ max : This step is performed in real time in the signal processing unit 103.

[0063] Using the components v x and v y of the principal eigenvector, the principal direction angle θ of the EEG source is calculated: .

[0064] Step S609: Output and visualize the EEG signal intensity and direction information: The calculated intensity time series S(t) and direction time series θ(t) are stored as a result file. At the same time, the results can be presented in real time or offline on the visualization display unit 104.

[0065] The steps S604-S609 are core processing parts of the whole signal processing method, can realize accurate analysis of the direction of the brain electrical activity source, and can determine the principal axis direction of the local brain electrical activity source by calculating the Hessian matrix of the potential field and decomposing the principal eigenvector.

[0066] The multi-channel electroencephalogram signal processing method based on Hessian matrix analysis in the above embodiment has high flexibility, is compatible with many existing high-density electroencephalogram acquisition devices, can realize accurate analysis of the direction of the brain electrical activity source, has good spatial filtering capability and high signal-to-noise ratio, and can capture weak local electroencephalogram signals.

[0067] The above describes some specific embodiments of the present application. It should be understood that the present application is not limited to the above specific embodiments, and various modifications or changes can be made by those skilled in the art within the scope of the claims, which does not affect the essential content of the present application. The above preferred features can be combined in any way without conflict.

Claims

1. A multi-channel electroencephalogram acquisition system based on analysis of the Hessian matrix, characterized in that, The method comprises the following steps: a multi-channel brain electrode array comprising at least one physical electrode group, the physical electrode group comprising a center electrode and at least eight surrounding electrodes, forming a square or circular Hessian electrode array; a signal acquisition unit synchronously acquiring and digitizing at least nine channels of brain electric potential signals in the multi-channel brain electrode array; a signal processing unit constructing a Hessian matrix representing the local curvature of the potential field at the position of the center electrode based on the electrode signals acquired by the signal acquisition unit and the geometric relationship between them, performing eigenvalue decomposition on the Hessian matrix to obtain at least one principal eigenvalue and a principal eigenvector corresponding thereto, and calculating a principal direction angle based on the principal eigenvalue to represent the intensity of the brain electric activity and the direction of the brain electric signal source at the position of the center electrode; a visual display unit visualizing the brain electric potential acquired by the signal acquisition unit and indicating the direction of the brain electric signal source according to the result of the signal processing unit.

2. The multi-channel electroencephalogram acquisition system based on the analysis of the Hessian matrix according to claim 1, characterized in that, The multi-channel brain electrode array, wherein: the square Hessian electrode array is a square Hessian array with three rows and three columns, and the horizontal and vertical distances between adjacent electrodes are both a preset value h; the circular Hessian electrode array is a circular Hessian array formed by a center electrode and eight surrounding electrodes, and the eight surrounding electrodes are located on a circumference with the center electrode as the center and h as the radius.

3. The multi-channel electroencephalogram acquisition system based on the analysis of the Hessian matrix according to claim 1, characterized in that, The signal processing unit further comprises: determining the local source intensity of the brain electric activity based on the absolute value of the principal eigenvalue; and / or determining the local source propagation direction or principal axis direction of the brain electric activity based on the direction of the principal eigenvector.

4. The Hessian matrix based analytical multi-channel electroencephalogram acquisition system according to claim 1, wherein, The multi-channel brain electrode array comprises a plurality of preset physical electrode groups, each of which is formed by selecting at least nine physical electrodes from more than nine physical electrodes, and each of which forms a nine-point electrode array with a different physical electrode as the center. The signal processing unit independently performs Hessian matrix calculation and eigenvalue decomposition on each of the physical electrode groups to obtain a brain electric activity direction vector field covering a target acquisition region of the head; The target acquisition region comprises any one of the following: a region arranged on the scalp to realize non-invasive brain electric acquisition; a region arranged epidurally to realize semi-invasive brain electric acquisition; a region arranged on the cerebral cortex to realize invasive brain electric acquisition.

5. The Hessian matrix based analytical multi-channel electroencephalogram acquisition system according to any one of claims 1-4, characterized in that, The visual display unit comprises: at the position of each center electrode on the brain electric topographic map, a direction arrow determined by the principal eigenvector is displayed, and the visual attributes of the arrow include length, color or thickness, which are determined by the principal eigenvalue.

6. A multi-channel electroencephalogram signal processing method based on analysis of a Hessian matrix, characterized in that, The method comprises the following steps: defining at least one Hessian electrode array on a target acquisition region, the Hessian electrode array comprising a center electrode and at least eight surrounding electrodes; extracting at least nine channels of brain electric potential signals synchronously acquired by the Hessian electrode array; using the acquired brain electric potential signals and the known geometric positions between the electrodes of the Hessian electrode array, a Hessian matrix representing the local curvature of the potential field at the position of the center electrode is calculated; performing eigenvalue decomposition on the Hessian matrix to obtain at least one principal eigenvalue and principal eigenvector, which are used to determine the strength of the source of the collected electroencephalogram signal at the center electrode position and the principal direction of the source.

7. The method of claim 6, wherein, The defining of at least one Hessian electrode array on the target acquisition region comprises: virtually defining one or more Hessian electrode arrays in software from an electroencephalogram cap containing more than nine electrodes, wherein each virtual Hessian electrode array is centered on a different physical electrode.

8. The method of claim 6, wherein the method is characterized in that, The calculating of the Hessian matrix representing the local curvature of the potential field at the center electrode position using the collected electroencephalogram potential signals and the known geometric positions between the electrodes of the Hessian electrode array comprises: i) define the brain electric potential signal of the center electrode as V C ; define the brain electric potential signals of the four surrounding electrodes located in the east (E), west (W), south (S), and north (N) directions as V E , V W , V S , and V N , respectively; and define the brain electric potential signals of the four surrounding electrodes located in the northeast (NE), northwest (NW), southwest (SW), and southeast (SE) directions as V NE , V NW , V SW , and V SE , respectively; the distance between the surrounding electrodes and the center electrode is h; ii) calculating a second order partial derivative V of the potential field along a first direction, i.e. the x-axis, by combining the electroencephalic potential signals of the electrodes xx : ； calculating a second order partial derivative of the electric potential field in the second direction, i.e. the y-axis, Vyy yy : ； and computing the mixed second partial derivative V xy : For a square array of the Hensen electrodes, V xy is: ； For a circular array of the Hansen electrodes, V xy is: ； iii) calculating V xx , V yy , and V xy are constructed as a second order symmetric Hessian matrix, which at each time point t has the form 。 9. The method for processing multichannel electroencephalogram signals based on the analysis of the Hessian matrix according to claim 8, characterized in that, The performing of eigenvalue decomposition on the Hessian matrix to obtain at least one principal eigenvalue and principal eigenvector, which are used to determine the strength of the source of the collected electroencephalogram signal at the center electrode position and the principal direction of the source comprises: i) for each time point of the Hessian matrix H(V(t)), solving its eigenvalue λ and eigenvector v, and the characteristic equation is: ； Solving the characteristic equation to obtain two eigenvalues: ； ii) λ1is λ max is the largest eigenvalue, representing the largest principal curvature; λ2is λ min is the smallest eigenvalue, representing the smallest principal curvature; iii) λ max into the characteristic equation: ； obtaining a corresponding principal eigenvector v max : ； iv) using components v of the principal eigenvector x and v y the principal direction angle Θ of the source of the electroencephalographic signal is calculated: 。 10. The Hessian matrix based analytical multi-channel electroencephalogram signal processing method according to any one of claims 6-9, characterized in that, It also comprises a calibration module or calibration step, which is specifically: Before the Hessian matrix calculation, performing gain calibration and / or time delay calibration between channels of the electroencephalogram potential signals of the at least nine electrodes to compensate for the amplitude and phase inconsistency of each channel due to physical differences.

Images