CN111012341B - Artifact removal and electroencephalogram signal quality evaluation method based on wearable electroencephalogram equipment - Google Patents

Artifact removal and electroencephalogram signal quality evaluation method based on wearable electroencephalogram equipment Download PDF

Info

Publication number
CN111012341B
CN111012341B CN202010016383.7A CN202010016383A CN111012341B CN 111012341 B CN111012341 B CN 111012341B CN 202010016383 A CN202010016383 A CN 202010016383A CN 111012341 B CN111012341 B CN 111012341B
Authority
CN
China
Prior art keywords
artifact
electroencephalogram
signal
electroencephalogram signal
artifacts
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202010016383.7A
Other languages
Chinese (zh)
Other versions
CN111012341A (en
Inventor
崔兴然
高之琳
顾忠泽
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Southeast University
Original Assignee
Southeast University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Southeast University filed Critical Southeast University
Priority to CN202010016383.7A priority Critical patent/CN111012341B/en
Publication of CN111012341A publication Critical patent/CN111012341A/en
Application granted granted Critical
Publication of CN111012341B publication Critical patent/CN111012341B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/369—Electroencephalography [EEG]
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/68—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
    • A61B5/6801—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
    • A61B5/6802—Sensor mounted on worn items
    • A61B5/6803—Head-worn items, e.g. helmets, masks, headphones or goggles
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
    • A61B5/7207—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal of noise induced by motion artifacts
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/725—Details of waveform analysis using specific filters therefor, e.g. Kalman or adaptive filters
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7253—Details of waveform analysis characterised by using transforms
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Pathology (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Medical Informatics (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • Biophysics (AREA)
  • Public Health (AREA)
  • Veterinary Medicine (AREA)
  • Signal Processing (AREA)
  • Artificial Intelligence (AREA)
  • Psychiatry (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physiology (AREA)
  • Psychology (AREA)
  • Evolutionary Computation (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

The invention discloses an artifact removal and electroencephalogram signal quality evaluation method based on wearable electroencephalogram equipment, which comprises the following steps: acquiring an original electroencephalogram signal on the surface layer of a scalp through wearable electroencephalogram equipment; filtering the device noise in the original electroencephalogram signal; identifying various artifact noises in the original electroencephalogram signals, wherein the artifact noises comprise motion artifacts, burr artifacts, eye movement artifacts and myoelectricity artifacts; based on the artifact identification result, performing self-adaptive artifact cutting and signal splicing to obtain a clean electroencephalogram signal; and carrying out comprehensive quality evaluation on the clean electroencephalogram signals by adopting neural network classification and index parameters. The invention solves the problems of difficulty in artifact removal and brain electrical signal quality evaluation method loss caused by channel number limitation and real-time performance of wearable brain electrical equipment.

Description

Artifact removal and electroencephalogram signal quality evaluation method based on wearable electroencephalogram equipment
Technical Field
The invention belongs to the field of electroencephalogram signal monitoring, and particularly relates to an electroencephalogram signal quality evaluation method.
Background
The brain electrical signals are weak bioelectric signals generated by the discharge of cerebral neurons and reflecting brain activity. Because the electroencephalogram has the characteristics of easy acquisition, non-invasiveness and high time resolution, the electroencephalogram plays an increasingly important role in scientific research, disease diagnosis and other aspects. With the continuous development of electroencephalogram monitoring technology, a large number of wearable electroencephalogram devices appear in the market, and the wearable electroencephalogram devices are mainly applied to the fields of commerce, medical treatment, education and the like. However, these wearable electroencephalogram devices face many problems in practical application, two of the most important problems are: noise is difficult to remove in real time and signal quality is lacking. The electroencephalogram signal is a low-amplitude unsteady signal and is interfered by various noises in the acquisition process, such as: low frequency trend noise, high frequency noise, power frequency noise, motion artifacts, glitch artifacts, eye movement artifacts, myoelectrical artifacts, and the like. The existence of these disturbances seriously affects the basic rhythm of the brain electrical signal, and therefore algorithms capable of effectively recognizing and removing these artifacts need to be proposed. Because wearable electroencephalogram equipment is usually analyzed in real time in the actual application process, the difficulty of artifact identification and filtering is increased. Meanwhile, due to the loss of the support of manual screening and evaluation, invalid electroencephalogram signals with low signal quality cannot be removed in advance, and the accuracy of subsequent analysis of the electroencephalogram signals is seriously influenced. Therefore, a real-time artifact automatic removal and signal quality evaluation system can ensure the accuracy of the subsequent analysis result.
In the aspect of noise interference removal, the main purpose is to remove interference to the greatest extent on the basis of retaining electroencephalogram signals as much as possible. The Butterworth filter can be used for effectively filtering low-frequency trend, power frequency interference and high-frequency noise. Aiming at the artifacts such as motion, burrs, eye movement, myoelectricity and the like, because the artifacts have a plurality of overlapped parts with the basic rhythm of the electroencephalogram signal, part of useful electroencephalogram signals are inevitably removed in the artifact removing process. For multi-channel electroencephalogram, various artifacts can be extracted by generally applying an Independent Component Analysis (ICA) blind source separation method, and the artifacts can be filtered out by subtracting the artifacts from the original electroencephalogram. But for the current portable wearable electroencephalogram device, the number of channels is not enough to support the use of the ICA method. At present, the artifact removing method for the electroencephalogram signals with a small number of channels mainly comprises a fast Fourier transform method and a wavelet transform method. The fast Fourier transform method and the wavelet transform method are based on a fundamental wave to decompose electroencephalogram, the shape and the frequency of the artifacts are different, the fundamental wave and the filtering threshold value need to be set artificially, and the artifacts are difficult to be removed effectively while signals are kept. Neither of these methods is suitable for real-time applications of wearable brain electrical acquisition devices.
In general, the filtered electroencephalogram signal is regarded as a clean electroencephalogram signal and can be used for subsequent analysis, but in the practical application process, some special conditions exist, such as the signal-to-noise ratio of the original signal is too low, the signal amplitude is lower than a normal value, the energy ratio of the electroencephalogram signal wave band is not consistent with the normal value, and the like. The quality of the electroencephalogram signals under the special conditions is generally poor, and the electroencephalogram signals are not suitable for subsequent analysis and should be removed. In scientific research, all electroencephalogram data are artificially screened by researchers to remove data sections with poor signal quality, but the manual screening cannot be achieved in the practical application process, so that a technology for automatically evaluating the electroencephalogram signal quality is urgently needed.
Disclosure of Invention
In order to solve the technical problems mentioned in the background art, the invention provides an artifact removal and electroencephalogram signal quality evaluation method based on wearable electroencephalogram equipment.
In order to achieve the technical purpose, the technical scheme of the invention is as follows:
the artifact removal and electroencephalogram signal quality evaluation method based on the wearable electroencephalogram equipment comprises the following steps:
(1) acquiring an original electroencephalogram signal on the surface layer of a scalp by wearable electroencephalogram equipment, wherein the original electroencephalogram signal contains equipment noise and various artifact noises;
(2) filtering the device noise in the original electroencephalogram signal;
(3) identifying various artifact noises in the original electroencephalogram signals, wherein the artifact noises comprise motion artifacts, burr artifacts, eye movement artifacts and myoelectricity artifacts;
(4) performing self-adaptive artifact cutting and signal splicing based on the artifact identification result in the step (3) to obtain a clean electroencephalogram signal;
(5) and (4) carrying out comprehensive quality evaluation on the electroencephalogram signals obtained in the step (4) by adopting neural network classification and index parameters.
Further, in the step (2), filtering out high-frequency noise by using a Butterworth low-pass filter; filtering out low-frequency noise by using a Butterworth high-pass filter; and a Butterworth band elimination filter is used as a trapped wave to filter out power frequency interference.
Further, the cut-off frequency of the Butterworth low-pass filter is less than 100Hz, and the cut-off frequency of the Butterworth high-pass filter is less than 0.3 Hz.
Further, in step (3), the motion artifact identification method is as follows:
(1a) extracting an electroencephalogram signal between 0.3 and 2Hz by using a Butterworth band-pass filter, wherein the electroencephalogram signal contains motion artifacts;
(1b) extracting an envelope curve of the motion artifact by using a Hilbert method, wherein the envelope curve can keep the approximate shape of the motion artifact and ignore the influence of other signals;
(1c) smoothing the envelope curve obtained in the last step to further reduce the influence of other signals;
(1d) and performing standard fractional Z transformation on the smoothed signal, judging that the data exceeding a preset threshold value a is a motion artifact, and storing the motion artifact at an index position on the electroencephalogram signal, wherein a is 10 times of the amplitude of the normal clean electroencephalogram signal.
Further, in step (3), the identification method of the glitch artifact is as follows:
(2a) calculating the difference and approximate derivative of the signals, and identifying the signals with instantaneous potential change;
(2b) smoothing the difference and the absolute value of the approximate derivative using median filtering;
(2c) and (4) performing standard fraction Z transformation on the smoothed signal, judging the data exceeding a preset threshold b as a glitch artifact, and storing the glitch artifact in an index position on the electroencephalogram signal.
Further, in step (3), the method for identifying the eye movement artifact is as follows:
(3a) extracting an electroencephalogram signal between 0.3 and 10Hz by using a Butterworth band-pass filter, wherein the electroencephalogram signal contains eye movement artifacts;
(3b) extracting an envelope curve of the eye movement artifact by using a Hilbert method, wherein the envelope curve can keep the approximate shape of the eye movement artifact and neglect the influence of other signals;
(3c) smoothing the envelope curve obtained in the last step to further reduce the influence of other signals;
(3d) and (3) performing standard fractional Z transformation on the smoothed signal, judging that the data exceeding a preset threshold value c is an eye movement artifact, and storing the index position of the eye movement artifact on the electroencephalogram signal, wherein c is 5 times of the amplitude of the normal clean electroencephalogram signal.
Further, in step (3), the electromyographic artifact identification method is as follows:
(4a) extracting an electroencephalogram signal between 100 and 120Hz by using a Butterworth band-pass filter, wherein the electroencephalogram signal contains an electromyographic artifact;
(4b) extracting an envelope curve of the myoelectricity artifact by using a Hilbert method, wherein the envelope curve can keep the approximate shape of the myoelectricity artifact and neglect the influence of other signals;
(4c) smoothing the envelope curve obtained in the last step to further reduce the influence of other signals;
(4d) and performing standard fractional Z transformation on the smoothed signal, judging that the data exceeding a preset threshold value d is a myoelectric artifact, and storing the myoelectric artifact in an index position on the electroencephalogram signal.
Further, the specific process of step (4) is as follows:
(401) taking a union set of index positions of 4 kinds of artifacts identified on all channels of wearable electroencephalogram equipment;
(402) for any artifact, searching the index position of a point with the amplitude value within the range of [ -5,5] muV before and after the artifact index position, and taking the intersection on all channels of the wearable electroencephalogram equipment;
(403) finding out the nearest points before and after the artifact in the intersection as the starting position and the ending position of the artifact, and redefining the accurate artifact position;
(404) and cutting off the accurate artifact position, and splicing the residual signals after cutting to obtain a clean electroencephalogram signal.
Further, in the step (5), calculating an electroencephalogram signal quality evaluation parameter:
(A) SNRhighAnd SNRnotch：
Figure GDA0003547373210000051
Figure GDA0003547373210000052
In the above formula, S is the EEG signal for removing the noise of the device, NhighFor high frequency noise, NnotchPower frequency interference is adopted;
(B) effective signal rate E:
Figure GDA0003547373210000053
in the above formula, LleftData length, L, of post-signal-slicing and signal-splicing electroencephalograms for artifactsoriginalThe data length of the original brain electrical signal is obtained;
(C) motion artifact percentage Emove：
Figure GDA0003547373210000054
In the above formula, LmoveThe data length of the electroencephalogram signal of the motion artifact part is obtained;
(D) ratio of glitch artifact Ejump：
Figure GDA0003547373210000055
In the above formula, LjumpThe data length of the electroencephalogram signal of the burr artifact part is obtained;
(E) eye movement artifact percentage Eeog：
Figure GDA0003547373210000056
In the above formula, LeogThe data length of the EEG signal of the eye movement artifact part is obtained;
(F) myoelectric artifact ratio Eemg：
Figure GDA0003547373210000057
In the above formula, LemgThe data length of the electroencephalogram signal of the myoelectricity artifact part is obtained;
(G) the amplitude of the clean electroencephalogram signal;
(H) alpha band power spectral density fraction: extracting Alpha wave bands from clean electroencephalogram signals by utilizing wavelets, calculating the power spectral densities of the Alpha wave bands and the clean electroencephalogram signals respectively, and finally obtaining the power spectral density ratio of the Alpha wave bands in the electroencephalogram;
(I) and (4) cleaning the fuzzy entropy of the electroencephalogram signal.
Further, in the step (5), a time-frequency image of the clean electroencephalogram signal is generated by using a wavelet time-frequency analysis method, so that the one-dimensional signal is converted into a two-dimensional image and is used as an input of a deep neural network, the output of the deep neural network is the probability that the input data belongs to different categories, and the categories comprise low-quality unavailable, medium-quality available and high-quality available.
Adopt the beneficial effect that above-mentioned technical scheme brought:
(1) the method solves the problem that the artifacts are difficult to remove due to the limitation of the number of channels and real-time performance of the conventional wearable electroencephalogram equipment, realizes automatic identification, cutting and splicing of the artifacts, can completely remove the interference of equipment noise, motion artifacts, burr artifacts, eye movement artifacts and myoelectric artifacts while preserving the integrity of electroencephalogram information, and is free from the influence of the number of channels and suitable for a real-time system compared with other methods.
(2) The method solves the problem of the lack of the existing electroencephalogram signal quality evaluation method, and summarizes the signal-to-noise ratio, the effective signal rate, the occupation ratio of each artifact, the amplitude of the clean electroencephalogram signal, the occupation ratio of the power spectrum density of the Alpha wave band and the fuzzy entropy as index parameters of signal quality evaluation based on the characteristics of the real electroencephalogram signal. And meanwhile, the signal quality grade of the section of data is directly judged based on a deep neural network model, and whether the section of data is used for subsequent analysis is determined. The evaluation result can further improve the reliability, accuracy and interpretability of subsequent electroencephalogram analysis.
Drawings
FIG. 1 is an overall flow diagram of the method of the present invention;
FIG. 2 is a flow chart and process result diagram of the noise filtering of the apparatus of the present invention;
FIG. 3 is a flow chart and process result diagram of the motion artifact identification of the present invention;
FIG. 4 is a flow chart of the present invention glitch artifact identification and a process result diagram of two exemplary glitch artifacts;
FIG. 5 is a flowchart and process result diagram of eye movement artifact identification of the present invention;
FIG. 6 is a flow chart and process result diagram for electromyographic artifact identification of the present invention;
FIG. 7 is a schematic illustration of artifact clipping and stitching in accordance with the present invention;
FIG. 8 is a schematic diagram of the structure of the deep neural network of the present invention.
Detailed Description
The technical scheme of the invention is explained in detail in the following with the accompanying drawings.
As shown in fig. 1, the present invention mainly comprises 5 steps: signal acquisition, equipment noise filtering, four-artifact (motion artifact, burr artifact, eye movement artifact and myoelectric artifact) identification, artifact cutting and signal splicing and electroencephalogram signal quality evaluation. The specific operation of each step will be described in detail below.
Step 1-signal acquisition: the portable wearable electroencephalogram equipment collects original electroencephalogram signals on the surface layer of the scalp, and the original electroencephalogram signals contain equipment noise and various artifact noises besides the electroencephalogram signals. Device noise is noise interference caused by the device itself, including: low frequency noise, power frequency interference and high frequency noise. Based on these device noises, the resulting signal-to-noise ratio of the device can be used to assess the signal quality of the device itself. Various artifact noises are usually some artifact interferences caused by poor contact between equipment and a human body or human beings, and because the amplitude of the artifacts is usually far greater than that of an electroencephalogram signal, the analysis accuracy is influenced by the existence of the artifacts, and the artifacts need to be removed. The artifact noise includes: motion artifacts caused by trial motion, burr artifacts caused by unstable contact between the electrode and the trial, eye motion artifacts caused by trial blinking, and myoelectricity artifacts.
Step 2-equipment noise filtering: the pretreatment of the general brain electricity comprises the following steps: high-pass filtering, low-pass filtering and power frequency notch filtering, and aims to remove equipment noise. The brain electrical signal is a weak electrical signal of the scalp, the amplitude is usually 50 μ V, and the effective frequency range is 0.3-100 Hz. As shown in fig. 2 (a), for the original electroencephalogram signal, high-frequency noise is filtered by using a butterworth low-pass filter (the selection of the cut-off frequency depends on the electroencephalogram frequency band concerned by the subsequent analysis, but should be lower than the maximum effective frequency of scalp electroencephalogram by 100 Hz); low-frequency noise filtered by a Butterworth high-pass filter (the selection of a cut-off frequency is generally recommended to be below 0.3Hz, and the integrity of an electroencephalogram signal is influenced by an overhigh cut-off frequency); the 50Hz power frequency interference filtered by the 49-51Hz Butterworth band stop filter (the power frequency noise in different regions may be 60 Hz). The electroencephalogram signal obtained by filtering the device noise as shown in (b) in fig. 2 has no interference of the device noise with respect to the original signal, and the information of the electroencephalogram signal can be seen.
Step 3-four kinds of artifact identification:
motion artifact identification: motion artifacts typically result from subject head motion or motion such as walking, with amplitudes well above 50 μ V and frequency ranges typically [0.3,2] Hz. There are three main methods for eliminating motion artifact interference:
1. the method filters out low-frequency motion artifacts through a high-pass filter, but the method usually causes loss of most useful low-frequency electroencephalogram signals; 2. according to the acceleration sensor of the tested head, the motion trend in the brain electricity is separated from the brain electricity signal, but not all brain electricity devices are simultaneously provided with the acceleration sensor, and the algorithm of the method is not mature, so that the use of the method is greatly limited; 3. where motion artifacts are identified, the segment of the signal is subtracted directly from the brain electrical signal.
The invention can accurately identify the position of the motion artifact, and the specific identification steps shown in (a) in fig. 3 are as follows:
1. the main frequency band of the motion artifact is generally between 0.3Hz and 2Hz, and the electrical brain signals between 0.3Hz and 2Hz are extracted by using a Butterworth band-pass filter, wherein the motion artifact is contained;
2. because the amplitude of the motion artifact is usually dozens of times or even dozens of times of the electroencephalogram signal, the envelope curve of the motion artifact is extracted by using a Hilbert method, the approximate shape of the motion artifact can be reserved by the envelope curve, and the influence of other signals can be ignored;
3. smoothing the envelope curve obtained in the last step to further reduce the influence of other signals;
4. performing standard fractional Z transformation on the smoothed signal, judging data exceeding a threshold range (ten times of the amplitude of the normal clean electroencephalogram signal) as a motion artifact, and storing an index position of the motion artifact on the electroencephalogram signal;
5. the above steps are repeated for all channels.
The motion artifact recognition result is shown in (b) of fig. 3.
Recognizing the burr artifact: glitch artifacts are typically due to poor contact between the electrode and the skin or problems in the signal transmission process, causing momentary changes in the potential. In order to eliminate the interference of such artifacts, the glitch artifacts are first identified, and the following steps are taken as shown in (a) of fig. 4:
1. calculating the difference and approximate derivative of the signals, so as to identify the signals with instantaneous potential change;
2. median filtering to smooth the difference and approximate absolute values of the derivatives;
3. performing standard fraction Z transformation on the smoothed signal, judging that the data exceeding the threshold range is a burr artifact, and storing the index position of the burr artifact on the electroencephalogram signal;
4. the above steps are repeated for all channels.
The results of identifying the glitch artifacts are shown in fig. 4 (b) and (c), which respectively show two typical glitch artifacts.
Eye movement artifact identification: the eye movement artifact is an artifact commonly existing in electroencephalogram data, and has a large influence frequently. The amplitude is in the range of [50,200] μ V and the frequency range is typically [0.3,10] Hz. The current commonly used eye movement artifact removing method is a blind source separation method, but the method requires enough channels, otherwise, eye movement artifact signals cannot be effectively separated, and therefore the method is obviously not suitable for portable electroencephalogram equipment. And other methods similar to the method based on wavelet transformation or empirical mode decomposition have the defects of complex operation and easy filtering of useful frequency band signals. For portable electroencephalogram equipment which needs to be applied to actual life, the method for identifying the eye movement artifact comprises the following specific steps as shown in (a) in fig. 5:
1. extracting an electroencephalogram signal between 0.3 and 10Hz by using a Butterworth band-pass filter, wherein the electroencephalogram signal contains eye movement artifacts;
2. because the amplitude of the eye movement artifact is usually several times of the electroencephalogram signal, the envelope curve of the eye movement artifact is extracted by using a Hilbert method, the approximate shape of the eye movement artifact can be reserved by the envelope curve, and the influence of other signals can be ignored;
3. smoothing the envelope curve obtained in the last step to further reduce the influence of other signals;
4. performing standard fractional Z transformation on the smoothed signal, judging that the data exceeding a threshold range (five times of the amplitude of the normal clean electroencephalogram signal) is an eye movement artifact, and storing the index position of the eye movement artifact on the electroencephalogram signal;
5. the above steps are repeated for all channels.
The eye movement artifact recognition result is shown in (b) in fig. 5.
Myoelectric artifact identification: the myoelectric artifact exists in the process of collecting the electroencephalogram signals generally, the amplitude range and the frequency range of the myoelectric artifact are very wide, and the myoelectric artifact is difficult to separate and extract from the electroencephalogram signals. But generally the amplitude of the myoelectrical artifact is higher than 50 μ V and the frequency is higher. Although the high frequency noise has been filtered out using low pass filtering as described above, some high frequency signal remains due to the 3 rd order butterworth filter employed. For the above reasons, the present invention recognizes the myoelectric artifact as shown in (a) of fig. 6 by the following steps:
1. extracting an electroencephalogram signal between 100 and 120Hz by using a Butterworth band-pass filter, wherein the electroencephalogram signal contains high-frequency myoelectricity artifacts;
2. because the amplitude of the myoelectricity artifact is usually several times of that of the electroencephalogram signal, an envelope curve of the myoelectricity artifact is extracted by using a Hilbert method, and the envelope curve can emphasize the approximate position of the myoelectricity artifact;
3. smoothing the envelope curve obtained in the last step to further clarify the position of the myoelectric artifact;
4. performing standard fractional Z conversion on the smoothed signal, judging data exceeding a threshold range as a myoelectric artifact, and storing an index position of the data on the electroencephalogram signal;
5. the above steps are repeated for all channels.
The myoelectric artifact recognition result is shown in (b) in fig. 6.
Step 4, self-adaptive artifact cutting and electroencephalogram splicing: after the movement, the burr, the eye movement and the myoelectric artifact are identified through the steps, the electroencephalogram signals need to be cut out and spliced again. The problem of artifact residue or splicing burrs occurs in unnecessary cutting and splicing processes. The invention designs a self-adaptive cutting and splicing method, wherein some artifacts exist on all channels, while some artifacts may only exist on a certain specific channel, but in order to ensure the consistency of the channel data length, all channels need to cut artifact sections with the same length and then splice. Therefore, the specific steps shown in fig. 7 are as follows:
1. taking a union set of index positions of four artifacts identified on all channels;
2. for any artifact, searching the index position of a point with the amplitude value within the range of [ -5,5] muV before and after the artifact index position, and taking the intersection on all channels of the wearable electroencephalogram equipment; (ii) a
3. Finding out the nearest points before and after the artifact in the intersection as the starting position and the ending position of the artifact;
4. cutting out the redefined artifact positions and splicing the residual signals after cutting together;
5. and (5) repeating the steps 2-4 on all the artifacts to finish artifact cutting and electroencephalogram splicing.
Step 5, electroencephalogram signal quality evaluation: the evaluation of the electroencephalogram signal quality is mainly used for evaluating whether the electroencephalogram data can be used for analysis or not, and the evaluation method comprises index parameters and neural network classification.
1. Index parameter of signal quality
(A) SNRhighAnd SNRnotch：
Figure GDA0003547373210000111
Figure GDA0003547373210000112
In the above formula, S is the EEG signal for removing the noise of the device, NhighFor high frequency noise, NnotchPower frequency interference is adopted;
(B) effective signal rate E:
Figure GDA0003547373210000113
in the above formula, LleftData length, L, of post-signal-slicing and signal-splicing electroencephalograms for artifactsoriginalThe data length of the original brain electrical signal is obtained;
(C) motion artifact percentage Emove：
Figure GDA0003547373210000114
In the above formula, LmoveThe data length of the electroencephalogram signal of the motion artifact part is obtained;
(D) ratio of glitch artifact Ejump：
Figure GDA0003547373210000115
In the above formula, LjumpThe data length of the electroencephalogram signal of the burr artifact part is obtained;
(E) eye movement artifact percentage Eeog：
Figure GDA0003547373210000116
In the above formula, LeogThe data length of the EEG signal of the eye movement artifact part is obtained;
(F) myoelectric artifact ratio Eemg：
Figure GDA0003547373210000117
In the above formula, LemgThe data length of the electroencephalogram signal of the myoelectricity artifact part is obtained;
(G) amplitude of clean electroencephalogram signal: the amplitude of the normal brain electrical signal is about 50 muV, and the range is [20,100] muV.
(H) Alpha band power spectral density fraction: extracting Alpha wave bands from clean electroencephalogram signals by utilizing wavelets, calculating the power spectral densities of the Alpha wave bands and the clean electroencephalogram signals respectively, and finally obtaining the power spectral density ratio of the Alpha wave bands in the electroencephalogram;
(I) fuzzy entropy of clean electroencephalogram signals: and calculating the entropy of the clean electroencephalogram signal after artifact removal by using fuzzy entropy, wherein the entropy value is reduced if the signal has noise residue.
The above indices may describe the characteristics of the signal in detail from different angles.
2. Quality classification based on deep neural networks
And generating a time-frequency image of the brain electrical signal with the artifact removed by using a wavelet time-frequency analysis method, and converting the one-dimensional signal into a two-dimensional image to be used as the input of the deep neural network model. The signal quality of the brain electricity is divided into three levels: low quality unavailable, medium quality available, and high quality available, these three levels being used as three categories for the output of the model. As shown in fig. 8, connections similar to those in the residual network architecture are employed to optimize training by allowing information to propagate well in the deep neural network. The network consists of 1 basic block and 16 residual blocks. The basic block has two convolution layers (conv), and the residual block has one convolution layer and adopts max pool. After each convolutional layer, Batch Normalization (BN) and rectified linear unit (ReLU) are added, using a pre-activated block design. The first and last layers of the network are special cases due to the use of the pre-activation block structure. We also apply Dropout between convolutional layers and after non-linearity. The final fully connected layers (FC for short) and the normalized (softmax) output probability that the time-frequency graph belongs to each category. Because signal quality evaluation has no special requirements on data length, experimental environment and the like, a large amount of data can be collected in a short time to serve as a training set, and a time-frequency diagram generated by electroencephalogram of the training set can be used for training a model. And (3) taking a time-frequency graph generated by the test collection electroencephalogram as the input of the model to obtain the probability that the electroencephalogram data belong to different categories.
3. Comprehensive evaluation
The two parts respectively obtain index parameters for describing the original brain electricity and the brain electricity without artifacts, and the signal quality category of the brain electricity without artifacts. The class to which the signal quality belongs directly determines whether the piece of data is used for subsequent analysis, but the greatest disadvantage of deep neural networks is their unexplainability. When low-quality electroencephalogram exists, the reason causing low signals cannot be found quickly, and the defect can be well remedied by aid of index parameters. The combination and complementation of the two methods can quickly explain the reasons influencing the high and low signal quality and adjust in time while ensuring the high accuracy and real-time performance of the two methods. For example, when a continuous "low quality unavailable" data segment is detected, the index parameter is checked to find that the motion artifact component is large, and the user can be reminded to reduce head motion and improve data availability.
The embodiments are only for illustrating the technical idea of the present invention, and the technical idea of the present invention is not limited thereto, and any modifications made on the basis of the technical scheme according to the technical idea of the present invention fall within the scope of the present invention.

Claims (9)

1. The artifact removal and electroencephalogram signal quality evaluation method based on the wearable electroencephalogram equipment is characterized by comprising the following steps:
(1) acquiring an original electroencephalogram signal on the surface layer of a scalp by wearable electroencephalogram equipment, wherein the original electroencephalogram signal contains equipment noise and various artifact noises;
(2) filtering the device noise in the original electroencephalogram signal;
(3) identifying various artifact noises in the original electroencephalogram signals, wherein the artifact noises comprise motion artifacts, burr artifacts, eye movement artifacts and myoelectricity artifacts;
(4) performing self-adaptive artifact cutting and signal splicing based on the artifact identification result in the step (3) to obtain a clean electroencephalogram signal; the specific process is as follows:
(401) taking a union set of index positions of 4 kinds of artifacts identified on all channels of wearable electroencephalogram equipment;
(402) for any artifact, searching the index position of a point with the amplitude value within the range of [ -5,5] muV before and after the artifact index position, and taking the intersection on all channels of the wearable electroencephalogram equipment;
(403) finding out the nearest points before and after the artifact in the intersection as the starting position and the ending position of the artifact, and redefining the accurate artifact position;
(404) cutting off an accurate artifact position, and splicing the residual signals after cutting to obtain a clean electroencephalogram signal;
(5) and (4) carrying out comprehensive quality evaluation on the electroencephalogram signals obtained in the step (4) by adopting neural network classification and index parameters.
2. The wearable electroencephalogram device-based artifact removal and electroencephalogram signal quality evaluation method according to claim 1, wherein in the step (2), high-frequency noise is filtered out by using a Butterworth low-pass filter; filtering out low-frequency noise by using a Butterworth high-pass filter; and a Butterworth band elimination filter is used as a trapped wave to filter out power frequency interference.
3. The wearable electroencephalogram device-based artifact removal and electroencephalogram signal quality evaluation method of claim 2, wherein the cutoff frequency of the butterworth low-pass filter is less than 100Hz, and the cutoff frequency of the butterworth high-pass filter is less than 0.3 Hz.
4. The wearable electroencephalogram equipment-based artifact removal and electroencephalogram signal quality evaluation method according to claim 1, wherein in the step (3), the motion artifact identification method is as follows:
(1a) extracting an electroencephalogram signal between 0.3 and 2Hz by using a Butterworth band-pass filter, wherein the electroencephalogram signal contains motion artifacts;
(1b) extracting an envelope curve of the motion artifact by using a Hilbert method, wherein the envelope curve can keep the approximate shape of the motion artifact and ignore the influence of other signals;
(1c) smoothing the envelope curve obtained in the last step to further reduce the influence of other signals;
(1d) and performing standard fractional Z transformation on the smoothed signal, judging that the data exceeding a preset threshold value a is a motion artifact, and storing the motion artifact at an index position on the electroencephalogram signal, wherein a is 10 times of the amplitude of the normal clean electroencephalogram signal.
5. The wearable electroencephalogram equipment-based artifact removal and electroencephalogram signal quality evaluation method according to claim 1, wherein in the step (3), the identification method of the burr artifacts is as follows:
(2a) calculating the difference and approximate derivative of the signals, and identifying the signals with instantaneous potential change;
(2b) smoothing the difference and the absolute value of the approximate derivative using median filtering;
(2c) and (4) performing standard fraction Z transformation on the smoothed signal, judging the data exceeding a preset threshold b as a glitch artifact, and storing the glitch artifact in an index position on the electroencephalogram signal.
6. The wearable electroencephalogram equipment-based artifact removal and electroencephalogram signal quality evaluation method according to claim 1, wherein in the step (3), the eye movement artifact identification method is as follows:
(3a) extracting an electroencephalogram signal between 0.3 and 10Hz by using a Butterworth band-pass filter, wherein the electroencephalogram signal contains eye movement artifacts;
(3b) extracting an envelope curve of the eye movement artifact by using a Hilbert method, wherein the envelope curve can keep the approximate shape of the eye movement artifact and neglect the influence of other signals;
(3c) smoothing the envelope curve obtained in the last step to further reduce the influence of other signals;
(3d) and (3) performing standard fractional Z transformation on the smoothed signal, judging that the data exceeding a preset threshold value c is an eye movement artifact, and storing the index position of the eye movement artifact on the electroencephalogram signal, wherein c is 5 times of the amplitude of the normal clean electroencephalogram signal.
7. The wearable electroencephalogram equipment-based artifact removal and electroencephalogram signal quality evaluation method according to claim 1, wherein in the step (3), the electromyographic artifact identification method is as follows:
(4a) extracting an electroencephalogram signal between 100 and 120Hz by using a Butterworth band-pass filter, wherein the electroencephalogram signal contains an electromyographic artifact;
(4b) extracting an envelope curve of the myoelectricity artifact by using a Hilbert method, wherein the envelope curve can keep the approximate shape of the myoelectricity artifact and neglect the influence of other signals;
(4c) smoothing the envelope curve obtained in the last step to further reduce the influence of other signals;
(4d) and performing standard fractional Z transformation on the smoothed signal, judging that the data exceeding a preset threshold value d is a myoelectric artifact, and storing the myoelectric artifact in an index position on the electroencephalogram signal.
8. The wearable electroencephalogram equipment-based artifact removal and electroencephalogram signal quality evaluation method according to claim 1, wherein in the step (5), electroencephalogram signal quality evaluation parameters are calculated:
(A) SNRhighAnd SNRnotch：
Figure FDA0003547373200000031
Figure FDA0003547373200000032
In the above formula, S is the EEG signal for removing the noise of the device, NhighFor high frequency noise, NnotchPower frequency interference is adopted;
(B) effective signal rate E:
Figure FDA0003547373200000033
in the above formula, LleftData length, L, of post-signal-slicing and signal-splicing electroencephalograms for artifactsoriginalThe data length of the original brain electrical signal is obtained;
(C) motion artifact percentage Emove：
Figure FDA0003547373200000034
In the above formula，LmoveThe data length of the electroencephalogram signal of the motion artifact part is obtained;
(D) ratio of glitch artifact Ejump：
Figure FDA0003547373200000041
In the above formula, LjumpThe data length of the electroencephalogram signal of the burr artifact part is obtained;
(E) eye movement artifact percentage Eeog：
Figure FDA0003547373200000042
In the above formula, LeogThe data length of the EEG signal of the eye movement artifact part is obtained;
(F) myoelectric artifact ratio Eemg：
Figure FDA0003547373200000043
In the above formula, LemgThe data length of the electroencephalogram signal of the myoelectricity artifact part is obtained;
(G) the amplitude of the clean electroencephalogram signal;
(H) alpha band power spectral density fraction: extracting Alpha wave bands from clean electroencephalogram signals by utilizing wavelets, calculating the power spectral densities of the Alpha wave bands and the clean electroencephalogram signals respectively, and finally obtaining the power spectral density ratio of the Alpha wave bands in the electroencephalogram;
(I) and (4) cleaning the fuzzy entropy of the electroencephalogram signal.
9. The method for artifact removal and electroencephalogram signal quality evaluation based on wearable electroencephalogram equipment as claimed in claim 1, wherein in step (5), a wavelet time-frequency analysis method is utilized to generate a time-frequency graph of a clean electroencephalogram signal, so that a one-dimensional signal is converted into a two-dimensional image and is used as an input of a deep neural network, the output of the deep neural network is the probability that input data belongs to different categories, and the categories comprise low-quality unavailable, medium-quality available and high-quality available.
CN202010016383.7A 2020-01-08 2020-01-08 Artifact removal and electroencephalogram signal quality evaluation method based on wearable electroencephalogram equipment Active CN111012341B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010016383.7A CN111012341B (en) 2020-01-08 2020-01-08 Artifact removal and electroencephalogram signal quality evaluation method based on wearable electroencephalogram equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010016383.7A CN111012341B (en) 2020-01-08 2020-01-08 Artifact removal and electroencephalogram signal quality evaluation method based on wearable electroencephalogram equipment

Publications (2)

Publication Number Publication Date
CN111012341A CN111012341A (en) 2020-04-17
CN111012341B true CN111012341B (en) 2022-04-22

Family

ID=70198537

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010016383.7A Active CN111012341B (en) 2020-01-08 2020-01-08 Artifact removal and electroencephalogram signal quality evaluation method based on wearable electroencephalogram equipment

Country Status (1)

Country Link
CN (1) CN111012341B (en)

Families Citing this family (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112115856B (en) * 2020-09-17 2024-06-14 浙江迈联医疗科技有限公司 Electroencephalogram quality assessment method, storage medium and system
CN112043262B (en) * 2020-09-28 2023-01-31 南京伟思医疗科技股份有限公司 Identification system and method for neonatal brain electrical signal artifact
CN112826510B (en) * 2021-03-11 2023-02-07 清华大学 Electrophysiological signal acquisition system
CN115429292A (en) * 2021-06-02 2022-12-06 博睿康科技（常州）股份有限公司 Electroencephalogram signal quality detection device and system based on spectrum analysis
CN113040788B (en) * 2021-06-02 2021-09-21 博睿康科技（常州）股份有限公司 Electroencephalogram signal quality detection method based on spectrum analysis
CN113812965B (en) * 2021-08-19 2024-04-09 杭州回车电子科技有限公司 Sleep state identification method, sleep state identification device, electronic device and storage medium
CN114041803B (en) * 2021-11-16 2022-10-14 电子科技大学 TMS pulse artifact removing device based on interval median filtering
CN114189795B (en) * 2022-02-15 2022-05-17 武汉左点科技有限公司 Hearing aid peak noise self-adaptive adjusting method and device
CN114886388B (en) * 2022-07-12 2022-11-22 浙江普可医疗科技有限公司 Evaluation method and device for quality of electroencephalogram signal in anesthesia depth monitoring process
CN117064409B (en) * 2023-10-12 2024-01-30 深圳般意科技有限公司 Method, device and terminal for evaluating transcranial direct current intervention stimulation effect in real time
CN117257327A (en) * 2023-10-24 2023-12-22 沈阳工业大学 Electroencephalogram signal quality assessment method and system
CN117643474B (en) * 2023-12-24 2024-05-28 沈阳工业大学 Electroencephalogram signal processing method and system based on deep neural network
CN118576216B (en) * 2024-08-06 2024-10-15 江西杰联医疗设备有限公司 Artifact suppression capability evaluation method, device, electronic equipment and storage medium

Citations (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5603088A (en) * 1995-02-28 1997-02-11 Motorola, Inc. Method and apparatus for determining a quality level of an analog signal in a radio communication system
CA2369163A1 (en) * 2001-01-31 2002-07-31 Markus Naf Automatic colour defect correction
EP1237621A1 (en) * 1999-12-07 2002-09-11 Krasnow Institute Adaptive electric field modulation of neural systems
CN105249961A (en) * 2015-11-02 2016-01-20 东南大学 Real-time driving fatigue detection system and detection method based on Bluetooth electroencephalogram headset
CN105361880A (en) * 2015-11-30 2016-03-02 上海乃欣电子科技有限公司 Muscle movement event recognition system and method
WO2016110804A1 (en) * 2015-01-06 2016-07-14 David Burton Mobile wearable monitoring systems
CN106537901A (en) * 2014-03-26 2017-03-22 马克·W·帕布利科弗 Computerized method and system for providing customized entertainment content
CN106568589A (en) * 2016-11-04 2017-04-19 东南大学 Rubbing acoustic emission denoise method based on empirical wavelet transform
CN107095672A (en) * 2017-06-10 2017-08-29 深圳创达云睿智能科技有限公司 A kind of method of artefact signal in removal EEG signals
CN107468243A (en) * 2017-09-01 2017-12-15 北京大智商医疗器械有限公司 A kind of method and device for assessing stupor degree
EP3323742A1 (en) * 2016-11-22 2018-05-23 Tetra Laval Holdings & Finance S.A. A method of operating an apparatus for applying components to packaging containers
CN108175406A (en) * 2017-12-14 2018-06-19 中山大学附属第医院 Method and device for treating central sleep apnea
CN108478218A (en) * 2018-04-24 2018-09-04 武汉理工大学 Eeg signal acquisition system based on high performance circuit and virtual instrument
CN108960182A (en) * 2018-07-19 2018-12-07 大连理工大学 A kind of P300 event related potential classifying identification method based on deep learning
CN109009092A (en) * 2018-06-15 2018-12-18 东华大学 A method of removal EEG signals noise artefact
CN109318207A (en) * 2018-11-07 2019-02-12 西安交通大学 A kind of lower extremity movement readiness potential detection system and method using myoelectricity timing
CN109497998A (en) * 2018-12-14 2019-03-22 深圳和而泰数据资源与云技术有限公司 EEG signals method for evaluating quality, eeg signal acquisition device and helmet
CN110269608A (en) * 2019-06-10 2019-09-24 浙江普可医疗科技有限公司 A kind of method, apparatus and readable storage medium storing program for executing removing signal interference
CN110353672A (en) * 2019-07-15 2019-10-22 西安邮电大学 Eye artefact removal system and minimizing technology in a kind of EEG signals
CN110568483A (en) * 2019-07-22 2019-12-13 中国石油化工股份有限公司 Automatic evaluation method for seismic linear noise suppression effect based on convolutional neural network
WO2019239143A1 (en) * 2018-06-15 2019-12-19 CLARK, David Julian An eeg signal processor, and associated methods

Patent Citations (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5603088A (en) * 1995-02-28 1997-02-11 Motorola, Inc. Method and apparatus for determining a quality level of an analog signal in a radio communication system
EP1237621A1 (en) * 1999-12-07 2002-09-11 Krasnow Institute Adaptive electric field modulation of neural systems
CA2369163A1 (en) * 2001-01-31 2002-07-31 Markus Naf Automatic colour defect correction
CN106537901A (en) * 2014-03-26 2017-03-22 马克·W·帕布利科弗 Computerized method and system for providing customized entertainment content
WO2016110804A1 (en) * 2015-01-06 2016-07-14 David Burton Mobile wearable monitoring systems
CN105249961A (en) * 2015-11-02 2016-01-20 东南大学 Real-time driving fatigue detection system and detection method based on Bluetooth electroencephalogram headset
CN105361880A (en) * 2015-11-30 2016-03-02 上海乃欣电子科技有限公司 Muscle movement event recognition system and method
CN106568589A (en) * 2016-11-04 2017-04-19 东南大学 Rubbing acoustic emission denoise method based on empirical wavelet transform
EP3323742A1 (en) * 2016-11-22 2018-05-23 Tetra Laval Holdings & Finance S.A. A method of operating an apparatus for applying components to packaging containers
CN107095672A (en) * 2017-06-10 2017-08-29 深圳创达云睿智能科技有限公司 A kind of method of artefact signal in removal EEG signals
CN107468243A (en) * 2017-09-01 2017-12-15 北京大智商医疗器械有限公司 A kind of method and device for assessing stupor degree
CN108175406A (en) * 2017-12-14 2018-06-19 中山大学附属第医院 Method and device for treating central sleep apnea
CN108478218A (en) * 2018-04-24 2018-09-04 武汉理工大学 Eeg signal acquisition system based on high performance circuit and virtual instrument
CN109009092A (en) * 2018-06-15 2018-12-18 东华大学 A method of removal EEG signals noise artefact
WO2019239143A1 (en) * 2018-06-15 2019-12-19 CLARK, David Julian An eeg signal processor, and associated methods
CN108960182A (en) * 2018-07-19 2018-12-07 大连理工大学 A kind of P300 event related potential classifying identification method based on deep learning
CN109318207A (en) * 2018-11-07 2019-02-12 西安交通大学 A kind of lower extremity movement readiness potential detection system and method using myoelectricity timing
CN109497998A (en) * 2018-12-14 2019-03-22 深圳和而泰数据资源与云技术有限公司 EEG signals method for evaluating quality, eeg signal acquisition device and helmet
CN110269608A (en) * 2019-06-10 2019-09-24 浙江普可医疗科技有限公司 A kind of method, apparatus and readable storage medium storing program for executing removing signal interference
CN110353672A (en) * 2019-07-15 2019-10-22 西安邮电大学 Eye artefact removal system and minimizing technology in a kind of EEG signals
CN110568483A (en) * 2019-07-22 2019-12-13 中国石油化工股份有限公司 Automatic evaluation method for seismic linear noise suppression effect based on convolutional neural network

Non-Patent Citations (10)

* Cited by examiner, † Cited by third party
Title
Recognition of Emotional States Using Multiscale Information Analysis of High Frequency EEG Oscillations;Gao, Zhilin；Cui, Xingran；Wan, Wang;《ENTROPY》;20190630;第21卷(第6期);全文 *
便携式BCI设备快速自动去眼电伪迹算法的研究;牛群峰,周季冬,王莉,惠延波;《自动化与仪表》;20191231;第34卷(第2期);全文 *
儿童早期脑的发育与医教融合大健康;杨元魁,崔兴然.;《教育家》;20191231(第28期);全文 *
基于伪迹去除的单通道脑机接口技术研究;张璐;《中国优秀硕士学位论文全文数据库 信息科技辑》;20190115;正文第10-11页 *
柔性通信对抗模拟训练系统研究与实现;常宏斌.;《万方》;20180830;全文 *
海洛因成瘾戒断群体在MID任务中的奖赏期待与加工机制研究;周昱;《万方》;20191125;全文 *
经颅磁刺激同步干预的头皮脑电信号伪迹离线去除方法综述;银珊,李颖洁;《生物医学工程学杂志》;20191231;第36卷(第1期);全文 *
脑血管病后的康复训练现况和临床应用;崔宝娟;《万方》;20160623;全文 *
针刺对脑缺血后神经元损伤保护作用的研究;许能贵;《万方》;20011231;全文 *
面向转化医学的非线性信号分析及其应用研究;崔兴然.;《万方》;20131129;全文 *

Also Published As

Publication number Publication date
CN111012341A (en) 2020-04-17

Similar Documents

Publication Publication Date Title
CN111012341B (en) Artifact removal and electroencephalogram signal quality evaluation method based on wearable electroencephalogram equipment
Hamad et al. Feature extraction of epilepsy EEG using discrete wavelet transform
Islam et al. A wavelet-based artifact reduction from scalp EEG for epileptic seizure detection
Fathima et al. Wavelet based features for epileptic seizure detection
CN105496363A (en) Method for classifying sleep stages on basis of sleep EGG (electroencephalogram) signal detection
CN108577834A (en) A method of it is detected automatically for phase spike between epilepsy
CN114190944B (en) Robust emotion recognition method based on electroencephalogram signals
Wang et al. Wavelet transform based ECG denoising using adaptive thresholding
CN116784860B (en) Electrocardiosignal characteristic extraction system based on morphological heart beat template clustering
CN114757236B (en) Electroencephalogram signal denoising optimization method and system based on TQWT and SVMD
Tibdewal et al. Detection and classification of eye blink artifact in electroencephalogram through discrete wavelet transform and neural network
CN114533089A (en) Lower limb action recognition and classification method based on surface electromyographic signals
CN113208633A (en) Emotion recognition method and system based on EEG brain waves
Alturki et al. A DWT-band power-SVM based architecture for neurological brain disorders diagnosis using EEG signals
CN116616714A (en) Sleep detection method and system based on brain myooculomotor deep learning fusion model
CN115470821A (en) Electroencephalogram response deep learning classification identification method based on underwater acoustic signal stimulation
Li et al. Ocular Artifact Removal Algorithm of Single Channel EEG Based on CEEMDAN-ICA-WTD
CN113855047A (en) FVEP signal enhancement method based on abnormal segment recognition
Wei et al. Automatic recognition of chewing noises in epileptic EEG based on period segmentation
CN113180663A (en) Emotion recognition method and system based on convolutional neural network
CN112971786A (en) Apoplexy rehabilitation evaluation method based on brain electromyographic signal wavelet coherence coefficient
CN118078305B (en) Dynamic electrocardiograph data processing method for wearable blood flow dynamic monitor
CN115067943B (en) Concentration extraction method and system based on brain wave signals
CN116035594B (en) Electroencephalogram artifact removing method based on segmentation-noise reduction network
Sundaram et al. Discrete wavelet transform based on eeg signal analysis for diagnosing neurological disorder

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant