A method and device for single-channel EEG blind source separation

Abstract

The present invention provides a method and apparatus for single-channel EEG blind source separation, the method comprising collecting EEG data through a single-channel device, filtering the frequency band in which the electromyographic signal in the EEG data is located to obtain a first filtered signal, performing a secondary filtering on the first filtered signal to obtain the electromyographic signal, removing drift interference in the electromyographic signal, filtering the frequency band in which the EEG signal and the electrooculographic signal in the electromyographic signal are located to obtain a second filtered signal, and performing slow wave criterion, empirical mode decomposition, and layered signal superposition on the second filtered signal to obtain the final EEG signal. The present invention innovatively proposes a method for separating eye movement signals from single-channel EEG signals, separating the electrooculographic signal and the electromyographic signal, which were originally noise interference, to assist in the analysis of single-channel EEG signals. Based on the difference in the time at which the eye movement signal and the slow wave signal appear during sleep stages, the EEG signal and the eye movement signal are effectively distinguished, greatly reducing the interference between the signals.

Description

Technical Field

[0001] The present invention belongs to the technical field of electroencephalogram (EEG) data analysis, and in particular relates to a method and device for single-channel EEG blind source separation. Background Art

[0002] The electroencephalogram (EEG) is a graph obtained by amplifying and recording the spontaneous biopotentials of the brain from the scalp using sophisticated electronic instruments. It is the spontaneous, rhythmic electrical activity of groups of brain cells recorded by electrodes.

[0003] Electroencephalography (EEG) is a fundamental tool in daily clinical practice and plays a crucial role in sleep health analysis. Currently, research based on multi-lead EEG data from PSG (Polysomnography) for signal processing, sleep staging, and sleep health is relatively mature.

[0004] However, research on single-channel EEG data signal processing for daily health monitoring, sleep staging, and consumer-side proactive health intervention is still in the ascendant. Single-channel EEG signal acquisition can greatly reduce EEG acquisition costs and facilitate people's daily use.

[0005] When using EEG for sleep staging or sleep health research, electrooculography (EOG) and electromyography (EMG) are often needed to achieve better results. Therefore, when conducting such research or engineering applications, solutions are often explored using PSG data, and few solutions can be used for research using single-channel EEG data. However, single-channel EEG signals are often mixed with EOG and EMG signals. In practical applications of single-channel EEG, EOG and EMG signals often cause noise interference. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention proposes a method for single-channel EEG blind source separation, which comprises:

[0007] EEG data were collected through a single-channel device;

[0008] Filtering the frequency band of the electromyographic signal in the electroencephalogram data to obtain a first filtered signal;

[0009] Performing secondary filtering on the first filtered signal to obtain the electromyographic signal;

[0010] removing drift interference from the electromyographic signal, filtering the frequency band of the electroencephalogram signal and the electrooculogram signal in the electromyographic signal to obtain a second filtered signal;

[0011] The second filtered signal is subjected to a slow wave criterion to distinguish the slow wave part from the non-slow wave part, the second filtered signal of the slow wave part is subjected to empirical mode decomposition and high-frequency signal superposition, and the second filtered signal of the non-slow wave part is subjected to empirical mode decomposition and low-frequency signal superposition to obtain the final EEG signal.

[0012] Preferably, the “performing secondary filtering on the first filtered signal to obtain the electromyographic signal” includes:

[0013] Step A, constructing wavelet basis;

[0014] Step B: performing fitting using the wavelet basis based on a burr noise case in the first filtered signal to obtain the fitted wavelet basis;

[0015] Step C, convolving the first filtered signal with the fitted wavelet basis to obtain a characteristic signal;

[0016] Step D: obtaining a characteristic spectrum at a scale based on the characteristic signal;

[0017] Step E, repeating steps B to D until the characteristic spectra at all different scales are obtained, integrating the characteristic spectra at different scales to obtain a normalized total characteristic spectrum;

[0018] Step F: setting a threshold, and extracting a noise signal interval based on the normalized total characteristic spectrum and the threshold;

[0019] Step G: obtaining a noise signal according to the noise signal interval, performing single-channel blind source separation on the noise signal and the first filtered signal, and adjusting the amplitude to obtain the denoised electromyographic signal.

[0020] Specifically, the wavelet basis is constructed by the following formula:

[0021]

[0022] Wherein, ψ(t) represents the wavelet basis, and C represents the normalization constant during reconstruction;

[0023] The step C is implemented by the following formula:

[0024]

[0025] Wherein, ψ′(t) represents the wavelet basis after fitting, x(t) represents the first filtered signal, and x′(t) represents the characteristic signal obtained by convolution;

[0026] The step D is implemented by the following formula:

[0027] X(t)=|x′(t)|2

[0028] Wherein, X(t) represents the characteristic spectrum at a scale;

[0029] The total characteristic spectrum in step E is obtained by summing the characteristic spectra at different scales, and the "obtaining a normalized total characteristic spectrum" is achieved by the following formula:

[0030]

[0031] Among them, X N (t) represents the normalized total characteristic spectrum, X all (t) represents the total characteristic spectrum;

[0032] The step F is implemented by the following formula:

[0033]

[0034] Wherein, ND(t) represents the noise signal interval, and δ represents the threshold;

[0035] The step of “obtaining a noise signal according to the noise signal interval” in step G is implemented by the following formula:

[0036] Noise(t)=ND(t)·x(t)

[0037] Wherein, Noise(t) represents the noise signal.

[0038] Furthermore, the final EEG signal includes a first EEG signal and a second EEG signal, wherein the first EEG signal is obtained by performing empirical mode decomposition on the second filtered signal of the slow wave part and superimposing high-frequency signals, and the “performing empirical mode decomposition on the second filtered signal of the slow wave part and superimposing high-frequency signals” includes:

[0039] Performing K=5 empirical mode decomposition on the second filtered signal of the slow wave part to obtain five layers of modes and residual terms;

[0040] The first and second modal layers obtained by decomposition and the residual term are superimposed to obtain EOG; the third, fourth and fifth modal layers obtained by decomposition are superimposed to obtain the first EEG signal.

[0041] Specifically, the five-layer modal and the residual term are obtained by the following formula:

[0042]

[0043] Among them, IMF k (t) represents the kth mode, r(t) represents the residual term;

[0044] EOG is obtained by the following formula:

[0045]

[0046] The first EEG signal is obtained by the following formula:

[0047]

[0048] Wherein, EEG(t) represents the first electroencephalogram signal.

[0049] Furthermore, the second EEG signal is obtained by performing empirical mode decomposition on the second filtered signal of the non-slow wave portion and superimposing low-frequency signals, and the “performing empirical mode decomposition on the second filtered signal of the non-slow wave portion and superimposing low-frequency signals” includes:

[0050] Performing K=5 empirical mode decomposition on the second filtered signal of the non-slow wave part to obtain five layers of modes and residual terms;

[0051] The third, fourth and fifth layer modalities obtained by decomposition and the residual term are superimposed to obtain EOG; the first layer and second layer modalities obtained by decomposition are superimposed to obtain the second EEG signal.

[0052] Specifically, the five-layer modal and the residual term are obtained by the following formula:

[0053]

[0054] Among them, IMF k (t) represents the kth mode, r(t) represents the residual term;

[0055] EOG is obtained by the following formula:

[0056]

[0057] The second EEG signal is obtained by the following formula:

[0058]

[0059] Wherein, EEG(t) represents the second EEG signal.

[0060] The present invention also proposes a device for single-channel EEG blind source separation, which is used to implement the method described above, and includes:

[0061] Acquisition module, used to collect EEG data through a single-channel device;

[0062] A first filtering module, configured to filter the frequency band of the electromyographic signal in the electroencephalogram data to obtain a first filtered signal;

[0063] A first acquisition module, configured to perform secondary filtering on the first filtered signal to obtain the electromyographic signal;

[0064] A second filtering module is used to remove drift interference in the electromyographic signal, filter the frequency band of the electroencephalogram signal and the electrooculogram signal in the electromyographic signal, and obtain a second filtered signal;

[0065] The second acquisition module is used to perform a slow wave criterion on the second filtered signal to distinguish the slow wave part from the non-slow wave part, perform empirical mode decomposition and high-frequency signal superposition on the second filtered signal of the slow wave part, and perform empirical mode decomposition and low-frequency signal superposition on the second filtered signal of the non-slow wave part to obtain the final EEG signal.

[0066] Preferably, the first acquisition module includes:

[0067] Construction unit, used to construct wavelet basis;

[0068] a fitting unit, configured to perform fitting using the wavelet basis based on a burr noise case in the first filtered signal to obtain the fitted wavelet basis;

[0069] a convolution unit, configured to convolve the first filtered signal with the fitted wavelet basis to obtain a characteristic signal;

[0070] A characteristic spectrum acquisition unit, configured to obtain a characteristic spectrum at a certain scale based on the characteristic signal;

[0071] an integration unit, configured to cause the fitting unit, the convolution unit, and the characteristic spectrum acquisition unit to repeatedly execute preset functions until the characteristic spectra at all different scales are obtained, and then integrate the characteristic spectra at different scales to obtain a normalized total characteristic spectrum;

[0072] a noise signal interval acquisition unit, configured to set a threshold value and extract a noise signal interval based on the normalized total characteristic spectrum and the threshold value;

[0073] A third acquisition unit is configured to obtain a noise signal according to the noise signal interval, perform single-channel blind source separation on the noise signal and the first filtered signal, and adjust the amplitude to obtain the denoised electromyographic signal.

[0074] Furthermore, the final EEG signal includes a first EEG signal and a second EEG signal, and the second acquisition module includes:

[0075] a first acquisition unit, configured to perform K=5 empirical mode decomposition on the second filtered signal of the slow wave portion to obtain five layers of modes and residual terms; superimpose the first and second layers of modes obtained by the decomposition and the residual terms to obtain EOG; and superimpose the third, fourth, and fifth layers of modes obtained by the decomposition to obtain the first electroencephalogram signal;

[0076] The second acquisition unit is used to perform K=5 empirical mode decomposition on the second filtered signal of the non-slow wave part to obtain five layers of modes and residual terms; superimpose the third, fourth, and fifth layers of modes and the residual terms obtained by the decomposition to obtain EOG; and superimpose the first and second layers of modes obtained by the decomposition to obtain the second EEG signal.

[0077] The present invention has at least the following beneficial effects:

[0078] The method proposed in this invention can separate eye movement signals, myoelectric signals and pure EEG signals from single-channel EEG signals, reducing the interference between the various signals. It can separate the eye movement signals and myoelectric signals, which are originally noise interference, to assist in the analysis of single-channel EEG signals, providing great help for subsequent sleep analysis based on single-channel EEG.

[0079] Furthermore, the method proposed in the present invention introduces specific wavelet denoising technology as a method for removing glitch noise in electromyographic signals, and effectively distinguishes EEG signals from eye movement signals based on the difference in the appearance time of eye movement signals and slow wave signals in the sleep stages.

[0080] Therefore, the present invention provides a method and device for single-channel EEG blind source separation. The present invention innovatively proposes a method for separating eye movement signals from single-channel EEG signals, separating the eye movement signals and myoelectric signals that were originally noise interference to assist in the analysis of single-channel EEG signals. Based on the difference in the appearance time of eye movement signals and slow wave signals in the sleep stages, the EEG signals and eye movement signals are effectively distinguished, greatly reducing the interference between the signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0082] Figure 1 A schematic diagram of the overall process of a single-channel EEG blind source separation method provided in Example 1;

[0083] Figure 2A schematic diagram of a process for performing secondary filtering on the first filtered signal to obtain an electromyographic signal;

[0084] Figure 3(a) is a schematic diagram of a single-channel EEG signal during the awake stage, and Figures 3(b) to (d) are schematic diagrams of myoelectric signals, eye contact signals, and pure EEG signals, respectively, differentiated from the single-channel EEG signal during the awake stage;

[0085] Figure 4(a) is a schematic diagram of a single-channel EEG signal during the eye movement phase, and Figures 4(b) to (d) are schematic diagrams of the myoelectric signal, eye contact signal, and pure EEG signal distinguished from the single-channel EEG signal during the eye movement phase, respectively;

[0086] Figure 5(a) is a schematic diagram of a single-channel EEG signal in sleep stage N3. Figures 5(b) to (d) are schematic diagrams of myoelectric signals, eye movement signals, and pure EEG signals, respectively, differentiated from the single-channel EEG signal in sleep stage N3.

[0087] Figure 6 This is a schematic diagram of the module structure of a single-channel EEG blind source separation device proposed in Example 2;

[0088] Figure 7 This is a schematic diagram of the module structure of the first acquisition module.

[0089] Reference numerals:

[0090] 1-acquisition module; 2-first filtering module; 3-first acquisition module; 4-second filtering module; 5-second acquisition module; 31-construction unit; 32-fitting unit; 33-convolution unit; 34-feature spectrum acquisition unit; 35-integration unit; 36-noise signal interval acquisition unit; 37-third acquisition unit; 51-first acquisition unit; 52-second acquisition unit. DETAILED DESCRIPTION

[0091] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0092] Hereinafter, various embodiments of the present invention will be described more fully. The present invention can have various embodiments, and modifications and variations can be made therein. However, it should be understood that there is no intention to limit the various embodiments of the present invention to the specific embodiments disclosed herein, but rather that the present invention should be construed to encompass all modifications, equivalents, and / or alternatives falling within the spirit and scope of the various embodiments of the present invention.

[0093] Hereinafter, the terms "include" or "may include" used in various embodiments of the present invention indicate the presence of disclosed functions, operations, or elements, and do not limit the addition of one or more functions, operations, or elements. In addition, as used in various embodiments of the present invention, the terms "include", "have" and their cognates are intended only to indicate specific features, numbers, steps, operations, elements, components, or combinations of the foregoing, and should not be understood as excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations of the foregoing or the possibility of adding one or more features, numbers, steps, operations, elements, components, or combinations of the foregoing.

[0094] In various embodiments of the present invention, the expression "or" or "at least one of A or / and B" includes any or all combinations of the words listed simultaneously. For example, the expression "A or B" or "at least one of A or / and B" may include A, may include B, or may include both A and B.

[0095] The expressions (such as "first", "second", etc.) used in the various embodiments of the present invention may modify the various constituent elements in the various embodiments, but may not limit the corresponding constituent elements. For example, the above expressions do not limit the order and / or importance of the elements. The above expressions are only used to distinguish one element from other elements. For example, a first user device and a second user device indicate different user devices, although both are user devices. For example, without departing from the scope of the various embodiments of the present invention, a first element may be referred to as a second element, and similarly, a second element may also be referred to as a first element.

[0096] It should be noted that, in the present invention, unless otherwise expressly specified or defined, terms such as "mounted," "connected," and "fixed" should be understood broadly. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediary; and internal communication between two components. Those skilled in the art will understand the specific meanings of these terms in the present invention based on specific circumstances.

[0097] In the present invention, those skilled in the art need to understand that the terms indicating orientation or positional relationships herein are based on the orientation or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0098] The terms used in various embodiments of the present invention are only used to describe the purpose of specific embodiments and are not intended to limit the various embodiments of the present invention. As used herein, the singular form is intended to also include the plural form, unless the context clearly indicates otherwise. Unless otherwise limited, all terms used here (including technical terms and scientific terms) have the same meaning as those of ordinary skill in the art generally understood by the various embodiments of the present invention. The terms (such as those defined in generally used dictionaries) will be interpreted as having the same meaning as the contextual meaning in the relevant technical field and will not be interpreted as having idealized meaning or too formal meaning, unless clearly defined in various embodiments of the present invention.

[0099] Example 1

[0100] This embodiment proposes a method for single-channel EEG blind source separation. This method innovatively proposes a method for separating eye movement signals, myoelectric signals and pure EEG signals from single-channel EEG signals. It innovatively separates the eye movement signals and myoelectric signals that were originally noise interference to assist in the analysis of single-channel EEG signals, and introduces specific wavelet denoising technology as a method for removing burr noise in myoelectric signals. It also effectively distinguishes EEG signals from eye movement signals based on the difference in the time when eye movement signals and slow wave signals appear in the sleep stage, greatly reducing the interference between the various signals and providing great help for subsequent sleep analysis based on single-channel EEG. Please refer to Figure 1 , the method comprising:

[0101] S100. Collect EEG data through a single-channel device.

[0102] Specifically, a single-channel device needs to collect EEG data according to a preset sampling frequency and sampling duration.

[0103] S200: Filter the frequency band of the electromyographic signal in the electroencephalogram data to obtain a first filtered signal.

[0104] In this embodiment, the frequency band of the electromyographic signal in the electroencephalogram data is filtered by band-pass filtering, and the frequency of the band-pass filtering is preferably 25 Hz to 80 Hz.

[0105] S300: Perform secondary filtering on the first filtered signal to obtain an electromyographic signal.

[0106] S400 , removing drift interference in the electromyographic signal, filtering the frequency bands of the electroencephalographic signal and the electrooculographic signal in the electromyographic signal, and obtaining a second filtered signal.

[0107] In this embodiment, drift interference in the EEG data is removed by detrending, and the frequency bands of the EEG signals and EOG signals in the EMG signals are filtered by low-pass filtering. The frequency range of the low-pass filter is 30Hz to 60Hz. In this embodiment, the frequency of the low-pass filter is preferably 45Hz.

[0108] S500. Perform a slow wave criterion on the second filtered signal to distinguish the slow wave part from the non-slow wave part, perform empirical mode decomposition and high-frequency signal superposition on the second filtered signal of the slow wave part, and perform empirical mode decomposition and low-frequency signal superposition on the second filtered signal of the non-slow wave part to obtain the final EEG signal.

[0109] It should be noted that the function of the slow wave criterion in this embodiment is to distinguish slow waves from electrooculogram signals.

[0110] In this embodiment, the final EEG signal includes a first EEG signal and a second EEG signal. The first EEG signal is obtained by performing empirical mode decomposition on the second filtered signal of the slow wave portion and superimposing high-frequency signals. The second EEG signal is obtained by performing empirical mode decomposition on the second filtered signal of the non-slow wave portion and superimposing low-frequency signals. Step S500 specifically includes:

[0111] The second filtered signal of the slow wave part is subjected to K=5 empirical mode decomposition to obtain five layers of modes and residual terms; the first and second layers of modes and residual terms obtained by decomposition are superimposed to obtain EOG; the third, fourth and fifth layers of modes obtained by decomposition are superimposed to obtain the first EEG signal

[0112] The five-layer modes and residual terms are obtained by the following formula:

[0113]

[0114] Among them, IMF k (t) represents the kth mode, r(t) represents the residual term;

[0115] EOG is obtained by the following formula:

[0116]

[0117] Wherein, EOG(t) represents the electrooculogram signal.

[0118] The first EEG signal is obtained by the following formula:

[0119]

[0120] Wherein, EEG(t) represents the first electroencephalogram signal.

[0121] Step S500 further includes:

[0122] The second filtered signal of the non-slow wave part is subjected to empirical mode decomposition with K=5 to obtain five layers of modes and residual terms; the third, fourth, and fifth layers of modes and residual terms obtained by the decomposition are superimposed to obtain EOG; the first and second layers of modes obtained by the decomposition are superimposed to obtain the second EEG signal.

[0123] Among them, the five-layer mode and residual terms are obtained by the following formula:

[0124]

[0125] Among them, IMF k (t) represents the kth mode, r(t) represents the residual term;

[0126] EOG is obtained by the following formula:

[0127]

[0128] Wherein, EOG(t) represents the electrooculogram signal.

[0129] The second EEG signal is obtained by the following formula:

[0130]

[0131] Wherein, EEG(t) represents the second electroencephalogram signal.

[0132] For further information, please refer to Figure 2 , step S300 includes:

[0133] Step A, constructing wavelet basis;

[0134] Step B: Based on the burr noise case in the first filtered signal, use a wavelet basis to perform fitting to obtain a fitted wavelet basis;

[0135] Step C, convolving the first filtered signal with the fitted wavelet basis to obtain a characteristic signal;

[0136] Step D: obtaining a characteristic spectrum at a scale based on the characteristic signal;

[0137] Step E: repeat steps B to D until all characteristic spectra at different scales are obtained, and integrate the characteristic spectra at different scales to obtain a normalized total characteristic spectrum;

[0138] Step F: setting a threshold, and extracting the noise signal interval based on the normalized total characteristic spectrum and the threshold;

[0139] Step G: Obtain a noise signal according to the noise signal interval, perform single-channel blind source separation on the noise signal and the first filtered signal, and adjust the amplitude to obtain a de-noised electromyographic signal.

[0140] It should be noted that the method of performing single-channel blind source separation on the noise signal and the first filtered signal in step G belongs to the prior art and will not be described in detail in this embodiment.

[0141] Specifically, the Mexican hat wavelet basis ψ(t) is constructed by the following formula:

[0142]

[0143] Where ψ(t) represents the wavelet basis, and C represents the normalization constant during reconstruction;

[0144] Step C is implemented by the following formula:

[0145]

[0146] Where ψ′(t) represents the fitted wavelet basis, x(t) represents the first filtered signal, and x′(t) represents the characteristic signal obtained by convolution;

[0147] Step D obtains a single characteristic spectrum X(t) at one scale by performing a modulo-squaring operation on the characteristic signal x′(t) obtained by convolution, which is specifically achieved by the following formula:

[0148] X(t)=|x′(t)| 2

[0149] Among them, X(t) represents the characteristic spectrum at one scale;

[0150] In step E, the total characteristic spectrum is obtained by summing the characteristic spectra at different scales. The "obtaining a normalized total characteristic spectrum" is achieved by the following formula:

[0151]

[0152] Among them, X N (t) represents the normalized total characteristic spectrum, X all (t) represents the total characteristic spectrum;

[0153] Step F is implemented by the following formula:

[0154]

[0155] Where ND(t) represents the noise signal interval, and δ represents the threshold;

[0156] In step G, "obtaining the noise signal according to the noise signal interval" is achieved by the following formula:

[0157] Noise(t)=ND(t)·x(t)

[0158] Wherein, Noise(t) represents the noise signal.

[0159] The method proposed in this embodiment can provide great help for subsequent sleep analysis based on single-channel EEG. See Figures 3 to 5. Figures 3 to 5 show the effect of single-channel EEG blind source separation achieved by the method proposed in this embodiment on single-channel EEG data in the wakefulness stage, eye movement stage and sleep N3 stage.

[0160] It should be noted that the sleep process is divided into N1 stage, N2 stage, N3 stage and REM stage (eye movement stage) according to the different characteristics of EEG changes. The N1 stage accounts for about 5%, the N2 stage accounts for about 50%, the N3 stage accounts for about 20%, and the REM stage accounts for about 25%.

[0161] Generally speaking, the wakefulness period generally enters the sleep period in the order of N1 stage, N2 stage, N3 stage, and REM stage. The sleep depth from N1 stage to N3 stage gradually deepens. Among them, N1 stage is the transition period between wakefulness and sleep period. N2 stage sleep is deeper than N1 stage. The EEG is a low-amplitude mixed frequency wave. N3 stage is what is generally called deep sleep. N3 stage is equivalent to a very deep sleep stage, which has the best effect on restoring limb and internal organ functions. It is not easy to be awakened during this period.

[0162] Example 2

[0163] See also Figure 6 This embodiment proposes a single-channel EEG blind source separation device, which is used to implement the single-channel EEG blind source separation method proposed in Example 1. The device includes:

[0164] Acquisition module 1, used to collect EEG data through a single-channel device;

[0165] A first filtering module 2 is used to filter the frequency band of the electromyographic signal in the EEG data to obtain a first filtered signal;

[0166] A first acquisition module 3, configured to perform secondary filtering on the first filtered signal to obtain an electromyographic signal;

[0167] The second filtering module 4 is used to remove drift interference in the electromyographic signal and filter the frequency band of the electroencephalogram signal and the electrooculogram signal in the electromyographic signal to obtain a second filtered signal;

[0168] The second acquisition module 5 is used to perform slow wave judgment on the second filtered signal to distinguish the slow wave part from the non-slow wave part, perform empirical mode decomposition and high-frequency signal superposition on the second filtered signal of the slow wave part, and perform empirical mode decomposition and low-frequency signal superposition on the second filtered signal of the non-slow wave part to obtain the final EEG signal.

[0169] Specifically, the acquisition module 1 enables the single-channel device to collect EEG data according to the preset sampling frequency and sampling duration. The second filtering module 4 removes drift interference in the EEG data by detrending. The second acquisition module 5 performs slow wave judgment to distinguish slow waves from electrooculogram signals.

[0170] Specifically, the first filtering module 2 filters the frequency band of the electromyographic signal in the electroencephalographic data through band-pass filtering, and the frequency of the band-pass filtering is preferably 25Hz~80Hz; the second filtering module 4 filters the frequency band of the electroencephalographic signal and the electrooculographic signal in the electromyographic signal through low-pass filtering, and the optional range of the low-pass filtering is 30Hz~60Hz. In this embodiment, the frequency of the low-pass filtering is preferably 45Hz.

[0171] In this embodiment, the final EEG signal includes a first EEG signal and a second EEG signal. The first EEG signal is obtained by performing empirical mode decomposition on the second filtered signal of the slow wave part and superimposing the high-frequency signal. The second EEG signal is obtained by performing empirical mode decomposition on the second filtered signal of the non-slow wave part and superimposing the low-frequency signal. The second acquisition module 5 includes:

[0172] The first acquisition unit 51 is used to perform K=5 empirical mode decomposition on the first filtered signal determined to be a slow wave, wherein the high-frequency signals and residual terms of the first and second layers are superimposed to obtain an eye movement signal, and the low-frequency signals of the third, fourth and fifth layers are superimposed to obtain a pure EEG signal.

[0173] The five-layer modes and residual terms are obtained by the following formula:

[0174]

[0175] Among them, IMF k (t) represents the kth mode, r(t) represents the residual term;

[0176] EOG is obtained by the following formula:

[0177]

[0178] Wherein, EOG(t) represents the electrooculogram signal.

[0179] The first EEG signal is obtained by the following formula:

[0180]

[0181] Wherein, EEG(t) represents the first electroencephalogram signal.

[0182] The second acquisition module 5 further includes:

[0183] The second acquisition unit 52 is used to perform K=5 empirical mode decomposition on the second filtered signal of the slow wave part to obtain five layers of modes and residual terms; superimpose the first and second layers of modes and residual terms obtained by the decomposition to obtain EOG; and superimpose the third, fourth and fifth layers of modes obtained by the decomposition to obtain the first EEG signal.

[0184] Among them, the five-layer mode and residual terms are obtained by the following formula:

[0185]

[0186] Among them, IMF k (t) represents the kth mode, r(t) represents the residual term;

[0187] EOG is obtained by the following formula:

[0188]

[0189] Wherein, EOG(t) represents the electrooculogram signal.

[0190] The second EEG signal is obtained by the following formula:

[0191]

[0192] Wherein, EEG(t) represents the second electroencephalogram signal.

[0193] Preferably, please refer to Figure 7 , the first acquisition module 3 includes:

[0194] A construction unit 31 is used to construct a wavelet basis;

[0195] A fitting unit 32 is configured to perform fitting using a wavelet basis based on the burr noise case in the first filtered signal to obtain a fitted wavelet basis;

[0196] A convolution unit 33 is configured to convolve the first filtered signal with the fitted wavelet basis to obtain a characteristic signal;

[0197] A characteristic spectrum acquisition unit 34 is configured to obtain a characteristic spectrum at a certain scale based on the characteristic signal;

[0198] The integration unit 35 is used to make the fitting unit 32, the convolution unit 33 and the characteristic spectrum acquisition unit 34 repeatedly perform preset functions until the characteristic spectra at all different scales are obtained, and then integrate the characteristic spectra at different scales to obtain a normalized total characteristic spectrum;

[0199] The noise signal interval acquisition unit 36 is used to set a threshold value and extract the noise signal interval based on the normalized total characteristic spectrum and the threshold value;

[0200] The third acquisition unit 37 is configured to obtain a noise signal according to the noise signal interval, perform single-channel blind source separation on the noise signal and the first filtered signal, and adjust the amplitude to obtain a de-noised electromyographic signal.

[0201] Specifically, the step of constructing the Mexican hat wavelet basis in the construction unit 31 is implemented by the following formula:

[0202]

[0203] Where C represents the normalization constant during reconstruction.

[0204] The step of convolving the first filtered signal with the fitted wavelet basis by the convolution unit 33 is implemented by the following formula:

[0205]

[0206] Where ψ′(t) represents the fitted wavelet basis, x(t) represents the first filtered signal, and x′(t) represents the characteristic signal obtained by convolution.

[0207] The step of obtaining a single characteristic spectrum at a scale by the characteristic spectrum acquisition unit 34 is implemented by the following formula:

[0208] X(t)=|x′(t)| 2

[0209] Among them, X(t) represents the characteristic spectrum at one scale.

[0210] The integration unit 35 obtains the total characteristic spectrum by summing the characteristic spectra at different scales. The step of "obtaining the normalized total characteristic spectrum" performed by the integration unit 35 is implemented by the following formula:

[0211]

[0212] Among them, X N (t) represents the normalized total characteristic spectrum, X all (t) represents the total characteristic spectrum.

[0213] The step of extracting the noise signal interval by the noise signal interval acquisition unit 36 is implemented by the following formula:

[0214]

[0215] Where ND(t) represents the noise signal interval and δ represents the threshold.

[0216] The step of “obtaining the noise signal according to the noise signal interval” performed by the third acquisition unit 37 is implemented by the following formula:

[0217] Noise(t)=ND(t)·x(t)

[0218] Wherein, Noise(t) represents the noise signal.

[0219] In summary, the present invention provides a method and device for single-channel EEG blind source separation. The present invention innovatively proposes a method for separating eye movement signals from single-channel EEG signals, separating the eye movement signals and myoelectric signals that were originally noise interference to assist in the analysis of single-channel EEG signals. Based on the difference in the appearance time of eye movement signals and slow wave signals in the sleep stages, the EEG signals and eye movement signals are effectively distinguished, greatly reducing the interference between the signals.

[0220] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for single-channel EEG blind source separation, characterized in that: The method comprises: EEG data were collected through a single-channel device; Filtering the frequency band of the electromyographic signal in the electroencephalogram data to obtain a first filtered signal; Performing secondary filtering on the first filtered signal to obtain the electromyographic signal; removing drift interference from the electromyographic signal, filtering the frequency band of the electroencephalogram signal and the electrooculogram signal in the electromyographic signal to obtain a second filtered signal; performing a slow wave criterion on the second filtered signal to distinguish a slow wave part from a non-slow wave part, performing empirical mode decomposition and high-frequency signal superposition on the second filtered signal of the slow wave part, and performing empirical mode decomposition and low-frequency signal superposition on the second filtered signal of the non-slow wave part to obtain a final EEG signal; The “performing secondary filtering on the first filtered signal to obtain the electromyographic signal” includes: Step A, constructing wavelet basis; Step B: performing fitting using the wavelet basis based on a burr noise case in the first filtered signal to obtain the fitted wavelet basis; Step C, convolving the first filtered signal with the fitted wavelet basis to obtain a characteristic signal; Step D: obtaining a characteristic spectrum at a scale based on the characteristic signal; Step E, repeating steps B to D until the characteristic spectra at all different scales are obtained, integrating the characteristic spectra at different scales to obtain a normalized total characteristic spectrum; Step F: setting a threshold, and extracting a noise signal interval based on the normalized total characteristic spectrum and the threshold; Step G: obtaining a noise signal according to the noise signal interval, performing single-channel blind source separation on the noise signal and the first filtered signal, and adjusting the amplitude to obtain the denoised electromyographic signal.

2. The method for single-channel EEG blind source separation according to claim 1, characterized in that: The wavelet basis is constructed by the following formula: Wherein, ψ(t) represents the wavelet basis, and C represents the normalization constant during reconstruction; The step C is implemented by the following formula: Wherein, ψ′(t) represents the wavelet basis after fitting, x(t) represents the first filtered signal, and x′(t) represents the characteristic signal obtained by convolution; The step D is implemented by the following formula: X(t)=|x′(t)| 2 Wherein, X(t) represents the characteristic spectrum at a scale; The total characteristic spectrum in step E is obtained by summing the characteristic spectra at different scales, and the "obtaining a normalized total characteristic spectrum" is achieved by the following formula: Among them, X N (t) represents the normalized total characteristic spectrum, X all (t) represents the total characteristic spectrum; The step F is implemented by the following formula: Wherein, ND(t) represents the noise signal interval, and δ represents the threshold; The step of "obtaining a noise signal according to the noise signal interval" in step G is achieved by the following formula: Noise(t)=ND(t)·x(t) Wherein, Noise(t) represents the noise signal.

3. The method for single-channel EEG blind source separation according to claim 1, characterized in that: The final EEG signal includes a first EEG signal and a second EEG signal, wherein the first EEG signal is obtained by performing empirical mode decomposition on the second filtered signal of the slow wave part and superimposing high-frequency signals, and the “performing empirical mode decomposition on the second filtered signal of the slow wave part and superimposing high-frequency signals” includes: Performing K=5 empirical mode decomposition on the second filtered signal of the slow wave part to obtain five layers of modes and residual terms; The first and second modal layers obtained by decomposition and the residual term are superimposed to obtain EOG; the third, fourth and fifth modal layers obtained by decomposition are superimposed to obtain the first EEG signal.

4. The method for single-channel EEG blind source separation according to claim 3, characterized in that: The five-layer mode and the residual term are obtained by the following formula: Among them, IMF k (t) represents the kth mode, r(t) represents the residual term; EOG is obtained by the following formula: The first EEG signal is obtained by the following formula: Wherein, EEG(t) represents the first electroencephalogram signal.

5. The method for single-channel EEG blind source separation according to claim 3, characterized in that: The second EEG signal is obtained by performing empirical mode decomposition on the second filtered signal of the non-slow wave part and superimposing low-frequency signals, and the “performing empirical mode decomposition on the second filtered signal of the non-slow wave part and superimposing low-frequency signals” includes: Performing K=5 empirical mode decomposition on the second filtered signal of the non-slow wave part to obtain five layers of modes and residual terms; The third, fourth and fifth layer modalities obtained by decomposition and the residual term are superimposed to obtain EOG; the first layer and second layer modalities obtained by decomposition are superimposed to obtain the second EEG signal.

6. The method for single-channel EEG blind source separation according to claim 5, characterized in that: The five-layer mode and the residual term are obtained by the following formula: Among them, IMF k (t) represents the kth mode, r(t) represents the residual term; EOG is obtained by the following formula: The second EEG signal is obtained by the following formula: Wherein, EEG(t) represents the second EEG signal.

7. A single-channel EEG blind source separation device, characterized in that: The device is used to implement the method according to any one of claims 1 to 6, and the device includes: Acquisition module, used to collect EEG data through a single-channel device; A first filtering module, configured to filter the frequency band of the electromyographic signal in the electroencephalogram data to obtain a first filtered signal; A first acquisition module, configured to perform secondary filtering on the first filtered signal to obtain the electromyographic signal; A second filtering module is used to remove drift interference in the electromyographic signal, filter the frequency band of the electroencephalogram signal and the electrooculogram signal in the electromyographic signal, and obtain a second filtered signal; a second acquisition module, configured to perform a slow wave criterion on the second filtered signal to distinguish a slow wave part from a non-slow wave part, perform empirical mode decomposition and high-frequency signal superposition on the second filtered signal of the slow wave part, and perform empirical mode decomposition and low-frequency signal superposition on the second filtered signal of the non-slow wave part, so as to obtain a final EEG signal; The first acquisition module includes: Construction unit, used to construct wavelet basis; a fitting unit, configured to perform fitting using the wavelet basis based on a burr noise case in the first filtered signal to obtain the fitted wavelet basis; a convolution unit, configured to convolve the first filtered signal with the fitted wavelet basis to obtain a characteristic signal; A characteristic spectrum acquisition unit, configured to obtain a characteristic spectrum at a certain scale based on the characteristic signal; an integration unit, configured to cause the fitting unit, the convolution unit, and the characteristic spectrum acquisition unit to repeatedly execute preset functions until the characteristic spectra at all different scales are obtained, and then integrate the characteristic spectra at different scales to obtain a normalized total characteristic spectrum; a noise signal interval acquisition unit, configured to set a threshold value and extract a noise signal interval based on the normalized total characteristic spectrum and the threshold value; A third acquisition unit is configured to obtain a noise signal according to the noise signal interval, perform single-channel blind source separation on the noise signal and the first filtered signal, and adjust the amplitude to obtain the denoised electromyographic signal.

8. The single-channel EEG blind source separation device according to claim 7, characterized in that: The final EEG signal includes a first EEG signal and a second EEG signal, and the second acquisition module includes: a first acquisition unit, configured to perform K=5 empirical mode decomposition on the second filtered signal of the slow wave portion to obtain five layers of modes and residual terms; superimpose the first and second layers of modes obtained by the decomposition and the residual terms to obtain EOG; and superimpose the third, fourth, and fifth layers of modes obtained by the decomposition to obtain the first electroencephalogram signal; The second acquisition unit is used to perform K=5 empirical mode decomposition on the second filtered signal of the non-slow wave part to obtain five layers of modes and residual terms; superimpose the third, fourth, and fifth layers of modes and the residual terms obtained by the decomposition to obtain EOG; and superimpose the first and second layers of modes obtained by the decomposition to obtain the second EEG signal.

Images