WO2022166401A1 - 基于eemd-pca去除eeg信号中运动伪迹的方法及装置 - Google Patents

基于eemd-pca去除eeg信号中运动伪迹的方法及装置 Download PDF

Info

Publication number
WO2022166401A1
WO2022166401A1 PCT/CN2021/137312 CN2021137312W WO2022166401A1 WO 2022166401 A1 WO2022166401 A1 WO 2022166401A1 CN 2021137312 W CN2021137312 W CN 2021137312W WO 2022166401 A1 WO2022166401 A1 WO 2022166401A1
Authority
WO
WIPO (PCT)
Prior art keywords
eemd
pca
eeg signal
artifact
principal component
Prior art date
Application number
PCT/CN2021/137312
Other languages
English (en)
French (fr)
Inventor
李慧慧
马俊嵩
王磊
王博
谯小豪
颜延
Original Assignee
中国科学院深圳先进技术研究院
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 中国科学院深圳先进技术研究院 filed Critical 中国科学院深圳先进技术研究院
Publication of WO2022166401A1 publication Critical patent/WO2022166401A1/zh

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
    • A61B5/72—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

Definitions

  • the present invention relates to the field of planning, in particular, to a method and device for removing motion artifacts in EEG signals based on EEMD-PCA.
  • BCIs brain-computer interfaces
  • EEG electroencephalogram
  • Figure Figure
  • manufacturers have developed and produced various types of portable EEG acquisition devices on the market; in addition, emerging sensor technologies can realize the use of gel-free EEG acquisition electrodes, and users can quickly and easily use the EEG acquisition equipment themselves.
  • the scene is greatly expanded.
  • EEG signals are very weak and are easily affected by various noises and artifacts.
  • the changes in usage scenarios and sensors make the EEG signals collected by these portable devices more susceptible to interference, especially motion artifacts due to human motion.
  • Motion artifacts have a broad spectral distribution and therefore interfere with all EEG bands. In particular, their corresponding spectrums have a large overlap with the Beta band in the 15-30Hz range. Second, the magnitude of motion artifacts can be one to two orders of magnitude larger than EEG signals, and finally, motion artifacts are associated with a tendency to be less repeatable than other EEG artifacts.
  • EEG motion artifacts There are currently two main types of removal of EEG motion artifacts. One is that most of these studies are either limited to highly controlled laboratory environments, such as walking on a treadmill, or require additional reference information, such as using inertial sensors as a reference . The second is to use blind source separation techniques, such as ICA and CCA. ICA uses high-order statistics to obtain independent sources in statistical sense, and CCA uses second-order statistics to obtain independent sources in statistical sense. The two methods have limited effect on completely separating motion artifact components and removing artifacts.
  • Embodiments of the present invention provide a method and device for removing motion artifacts in an EEG signal based on EEMD-PCA, and the effect of removing artifacts is significantly improved compared to the current technology.
  • a method for removing motion artifacts in an EEG signal based on EEMD-PCA comprising the following steps:
  • the principal components are separated from the eigenmode functions of each order based on PCA;
  • a principal component whose autocorrelation is greater than a preset threshold is determined as an artifact principal component
  • the remaining principal components are inversely transformed by PCA and then inversely transformed by EEMD to obtain the EEG signal after noise removal.
  • decomposing the single-channel EEG signal based on EEMD to obtain the eigenmode components of each order includes:
  • the ensemble empirical mode decomposition is performed on the signal X(t), and the eigenmode functions of each order are obtained.
  • the separation of the eigenmode functions of each order into the principal components based on PCA includes:
  • calculating the autocorrelation of each principal component includes:
  • the method further includes: a preset threshold for the autocorrelation.
  • the source component S identified as the artifact component is removed, and the source component S' is obtained after removal.
  • the EEMD inverse change processing is performed, and the obtained EEG signal after noise removal includes:
  • the source component S' is subjected to PCA inverse transformation post-processing and then EEMD inverse transformation processing to obtain the EEG signal after noise removal.
  • the amplitude standard deviation of the white noise added by the ensemble empirical mode decomposition is set to be between 0.05 and 0.15, and the number of noise additions is set to be between 80 and 120 times.
  • a device for removing motion artifacts in an EEG signal based on EEMD-PCA comprising:
  • the signal decomposition module is used to decompose the single-channel EEG signal based on EEMD to obtain the eigenmode functions of each order;
  • the principal component separation module is used to separate the principal components from the eigenmode functions of each order based on PCA;
  • the correlation calculation module is used to calculate the autocorrelation of each principal component
  • an artifact component determination module configured to determine a principal component whose autocorrelation is greater than a preset threshold as an artifact principal component
  • the artifact component removal module is used to remove the principal component determined to be an artifact
  • the inverse change processing module is used to inversely change the remaining principal components through PCA and then through EEMD to obtain the EEG signal after noise removal.
  • the method and device for removing motion artifacts in an EEG signal based on EEMD-PCA in the embodiments of the present invention decompose a single-channel EEG signal based on EEMD to obtain eigenmode functions of each order; separate eigenmode functions of each order based on PCA Calculate the autocorrelation of each principal component; determine the principal component whose autocorrelation is greater than the preset threshold as the artifact principal component; remove the principal component determined as the artifact; remove the remaining principal components
  • the components are inversely changed by PCA and then inversely changed by EEMD to obtain the EEG signal after denoising.
  • the invention uses the PCA method to separate the artifact components, then automatically selects the artifact components according to the autocorrelation, and removes the artifact components to retain useful EEG information; Compared with the current technology, the anti-aliasing effect has been significantly improved.
  • Fig. 1 is the flow chart of the method for removing motion artifact in EEG signal based on EEMD-PCA of the present invention
  • Fig. 2 is that the present invention contains motion artifact EEG signal and pure EEG signal diagram
  • Fig. 3 is the eigenmode component that contains the motion artifact EEG decomposition of the present invention
  • Fig. 4 is the principal component diagram obtained after the eigenmode component of the present invention is decomposed by PCA;
  • Fig. 5 is the identification above the autocorrelation coefficient line corresponding to the principal component of the present invention as an artifact component diagram
  • FIG. 6 is a comparison diagram of the EEG signal after the artifact is removed by the present invention and the pure EEG signal;
  • Fig. 7 is the effect comparison diagram of the technology of the present invention and the prior art
  • FIG. 8 is a schematic diagram of an apparatus for removing motion artifacts in an EEG signal based on EEMD-PCA according to the present invention.
  • Reference numerals 201-signal decomposition module, 202-principal component separation module, 203-correlation calculation module, 204-artifact component determination module, 206-artifact component removal module.
  • a method for removing motion artifacts in an EEG signal based on EEMD-PCA is provided, referring to FIG. 1 , including the following steps:
  • S101 Decompose a single-channel EEG signal based on EEMD to obtain eigenmode functions of each order;
  • S102 separate the principal components from the eigenmode functions of each order based on PCA;
  • S104 Determine the principal component whose autocorrelation is greater than the preset threshold as the artifact principal component
  • S106 Perform PCA inverse change processing on the remaining principal components, and then perform EEMD inverse change processing to obtain an EEG signal after noise removal.
  • the method and device for removing motion artifacts in an EEG signal based on EEMD-PCA in the embodiments of the present invention decompose a single-channel EEG signal based on EEMD to obtain eigenmode functions of each order; separate eigenmode functions of each order based on PCA Calculate the autocorrelation of each principal component; determine the principal component whose autocorrelation is greater than the preset threshold as the artifact principal component; remove the principal component determined as the artifact; remove the remaining principal components
  • the components are inversely changed by PCA and then inversely changed by EEMD to obtain the EEG signal after denoising.
  • the invention uses the PCA method to separate the artifact components, then automatically selects the artifact components according to the autocorrelation, and removes the artifact components to retain useful EEG information; Compared with the current technology, the anti-aliasing effect has been significantly improved.
  • the eigenmode components of each order are obtained by decomposing a single-channel EEG signal based on EEMD, including:
  • the ensemble empirical mode decomposition is performed on the signal X(t), and the eigenmode functions of each order are obtained.
  • Step 1 Add the signal X(t) to the noise signal W(t) to obtain the signal X'(t);
  • Step 2 Perform empirical mode decomposition on the signal X'(t) to obtain eigenmode function components of each order, wherein the remaining components after decomposition are r n (t);
  • Step 3 Repeat steps 1 and 2, adding white noise with the same intensity and different sequences each time;
  • separating the principal components of the eigenmode functions of each order based on PCA includes:
  • Step 1 Calculate the l(t) covariance matrix obtained in Step 4;
  • Step 2 Calculate the eigenvalues and eigenvectors of the l(t) covariance matrix
  • Step 3 Sort the eigenvalues from large to small, and use the eigenvectors corresponding to the eigenvalues as column vectors to form an eigenvector matrix;
  • the method before the principal component whose autocorrelation is greater than the preset threshold is determined as the artifact principal component, the method further includes: a preset threshold for the autocorrelation.
  • the preset threshold is set to 0.97, and when the calculated autocorrelation is greater than the preset threshold of 0.97, the principal component whose autocorrelation is greater than the preset threshold is determined as an artifact principal component.
  • the straight line between 0.9-1 in the figure is the autocorrelation coefficient line, and the components above the autocorrelation coefficient line are determined as artifact components.
  • the removal of the principal components determined to be artifacts includes:
  • the source component S identified as the artifact component is removed, and the source component S' is obtained after removal.
  • EEG signal after noise removal includes:
  • the source component S' is subjected to PCA inverse transformation processing and then EEMD inverse transformation processing to obtain the EEG signal after noise removal.
  • the method before performing collective empirical mode decomposition on the signal X(t) to obtain N eigenmode functions, the method further includes:
  • the amplitude standard deviation of the white noise added by the collective empirical mode decomposition is set to be between 0.05 and 0.15, and the number of noise additions is set to be between 80 and 120 times.
  • the amplitude standard deviation of white noise added by ensemble empirical mode decomposition is set to 0.1, and the number of noise additions is set to 100 times.
  • the standard deviation of the amplitude of the white noise added by EEMD is set to 0.1, and the number of noise additions is set to 100 times.
  • the ground truth signal represents the pure signal in the EEG signal, which is basically a flat straight line in the figure; the noise signal (Noisy signal), that is, the motion artifact in the EEG signal, in the figure
  • the amplitude fluctuates greatly; it can be seen that in the EEG signal not processed by the present invention, the motion artifact component is obvious and deviates greatly from the pure signal.
  • Fig. 3 is the eigenmode function obtained after decomposing the EEG signal.
  • the separated N principal components namely N source components S
  • the autocorrelation of each source component is calculated. If the autocorrelation is greater than 0.97, That is, the source components above the autocorrelation coefficient line are regarded as artifact components.
  • the clean signal is the signal after the artifact is removed by the method of the present invention, and it can be seen in FIG. 6 that the line represented by the Cleaned signal after removing the artifact and the line represented by the Ground truth signal overlap to a large extent . That is, the EEG signal after removing the artifact basically coincides with the pure EEG signal, and the EEG signal after removing the noise is obtained by removing the artifact.
  • the present invention has been tested on the disclosed two-channel EEG signal removing motion artifact verification data set, and the effect is significantly improved compared with the existing EEMD-ICA and EEMD-CCA techniques;
  • the root mean square error from the original signal after the artifact component is reduced by 24.2, the signal-to-noise ratio is improved by 9.2 dB, and the similarity is improved by 0.24.
  • the root mean square error is the differential evaluation index.
  • an apparatus for removing motion artifacts in an EEG signal based on EEMD-PCA is provided, see FIG. 8 , including:
  • the signal decomposition module 201 is used to decompose the single-channel EEG signal based on EEMD to obtain the eigenmode function of each order;
  • a principal component separation module 202 configured to separate the principal components from the eigenmode functions of each order based on PCA;
  • Correlation calculation module 203 used to calculate the autocorrelation of each principal component
  • an artifact component determination module 204 configured to determine a principal component whose autocorrelation is greater than a preset threshold as an artifact principal component
  • Artifact component removal module 205 configured to remove the principal component determined to be an artifact
  • the inverse change processing module 206 is configured to perform PCA inverse change processing on the remaining principal components and then perform EEMD inverse change processing to obtain an EEG signal after noise removal.
  • the method and device for removing motion artifacts in an EEG signal based on EEMD-PCA in the embodiments of the present invention decompose a single-channel EEG signal based on EEMD to obtain eigenmode functions of each order; separate eigenmode functions of each order based on PCA Calculate the autocorrelation of each principal component; determine the principal component whose autocorrelation is greater than the preset threshold as the artifact principal component; remove the principal component determined as the artifact; remove the remaining principal components
  • the components are inversely changed by PCA and then inversely changed by EEMD to obtain the EEG signal after denoising.
  • the invention uses the PCA method to separate the artifact components, then automatically selects the artifact components according to the autocorrelation, and removes the artifact components to retain useful EEG information; Compared with the current technology, the anti-aliasing effect has been significantly improved.
  • the eigenmode components of each order are obtained by decomposing a single-channel EEG signal based on EEMD, including:
  • the ensemble empirical mode decomposition is performed on the signal X(t), and the eigenmode functions of each order are obtained.
  • Step 1 Add the signal X(t) to the noise signal W(t) to obtain the signal X'(t);
  • Step 2 Perform empirical mode decomposition on the signal X'(t) to obtain eigenmode function components of each order, wherein the remaining components after decomposition are r n (t);
  • Step 3 Repeat steps 1 and 2, adding white noise with the same intensity and different sequences each time;
  • separating the principal components of the eigenmode functions of each order based on PCA includes:
  • Step 1 Calculate the l(t) covariance matrix obtained in Step 4;
  • Step 2 Calculate the eigenvalues and eigenvectors of the l(t) covariance matrix
  • Step 3 Sort the eigenvalues from large to small, and use the eigenvectors corresponding to the eigenvalues as column vectors to form an eigenvector matrix;
  • calculating the autocorrelation of each principal component includes:
  • the method before the principal component whose autocorrelation is greater than the preset threshold is determined as the artifact principal component, the method further includes: a preset threshold for the autocorrelation.
  • the preset threshold is set to 0.97, and when the calculated autocorrelation is greater than the preset threshold of 0.97, the principal component whose autocorrelation is greater than the preset threshold is determined as an artifact principal component.
  • the straight line between 0.9-1 in the figure is the autocorrelation coefficient line, and the components above the autocorrelation coefficient line are determined as artifact components.
  • the removal of the principal components determined to be artifacts includes:
  • the source component S identified as the artifact component is removed, and the source component S' is obtained after removal.
  • EEG signal after noise removal includes:
  • the source component S' undergoes an inverse change of PCA and then undergoes an inverse change of EEMD to obtain an EEG signal after noise removal.
  • the method before the collective empirical mode decomposition is performed on the signal X(t) to obtain N eigenmode functions, the method further includes:
  • the amplitude standard deviation of the white noise added by the collective empirical mode decomposition is set to be between 0.05 and 0.15, and the number of noise additions is set to be between 80 and 120 times.
  • the amplitude standard deviation of white noise added by ensemble empirical mode decomposition is set to 0.1, and the number of noise additions is set to 100 times.
  • the standard deviation of the amplitude of the white noise added by EEMD is set to 0.1, and the number of noise additions is set to 100 times.
  • the ground truth signal represents the pure signal in the EEG signal, which is basically a flat straight line in the figure; the noise signal (Noisy signal), that is, the motion artifact in the EEG signal, in the figure
  • the amplitude fluctuates greatly; it can be seen that in the EEG signal not processed by the present invention, the motion artifact component is obvious and deviates greatly from the pure signal.
  • Fig. 3 is the eigenmode function obtained after decomposing the EEG signal.
  • the separated N principal components namely N source components S
  • the autocorrelation of each source component is calculated. If the autocorrelation is greater than 0.97, That is, the source components above the autocorrelation coefficient line are regarded as artifact components.
  • the clean signal is the signal after the artifact is removed by the method of the present invention, and it can be seen in Fig. 6 that the line represented by the Cleaned signal after removing the artifact and the line represented by the Ground truth signal overlap to a large extent . That is, the EEG signal after removing the artifact basically coincides with the pure EEG signal, and the EEG signal after removing the noise is obtained by removing the artifact.
  • the present invention has been tested on the disclosed two-channel EEG signal removing motion artifact verification data set, and the effect is significantly improved compared with the existing EEMD-ICA and EEMD-CCA techniques;
  • the root mean square error from the original signal after the artifact component is reduced by 24.2, the signal-to-noise ratio is improved by 9.2 dB, and the similarity is improved by 0.24.
  • the root mean square error is the difference evaluation index.

Landscapes

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

Abstract

一种基于EEMD-PCA去除EEG信号中运动伪迹的方法及装置，涉及EEG移动脑电图处理领域，该方法及装置基于EEMD分解单通道EEG信号，得到各阶的本征模态函数（S101）；基于PCA将各阶本征模态函数分离出主成分（S102）；计算出每个主成分的自相关性（S103）；将自相关性大于预设阈值的主成分被判定为伪迹主成分（S104）；将被判定为伪迹的主成分进行去除（S105）；对剩余的主成分进行PCA逆变化处理后再进行EEMD逆变化处理，得到去除噪声后的EEG信号（S106）。该伪迹去除方法，相比于目前技术去伪效果得到显著提升。

Description

基于EEMD-PCA去除EEG信号中运动伪迹的方法及装置 技术领域
本发明涉及规划领域，具体而言，涉及一种基于EEMD-PCA去除EEG信号中运动伪迹的方法及装置。
背景技术
在过去的几年中，许多研究都集中在研究和开发基于移动脑电图的脑机接口(brain-computer interfaces，BCI)，这些接口技术能够在日常生活中采集人们的EEG(electroencephalogram，脑电图)信号。目前市面上，制造商已经开发生产出多种型号的便携式EEG采集设备；另外新兴的传感器技术可实现使用无凝胶的EEG采集电极，并可由用户自己快速简便地使用EEG采集设备，用户的使用场景得到很大的扩展。EEG信号非常微弱，容易受到各种噪声和伪迹的影响，便携式采集设备与传统的脑电图采集设备相比，使用场景、传感器的变化，使这些便携式设备采集的脑电图信号更容易受到干扰，尤其是由于人体运动引起的运动伪迹。运动伪迹具有较宽的频谱分布，因此会干扰所有EEG频段。特别是它们对应的频谱与15-30Hz范围的Beta频带有很大的重叠。其次运动伪迹的幅值相比与EEG信号可大一到两个数量级，最后，与其他脑电伪迹相比，运动伪迹与重复性较低的趋势相关。
2012年研究人员使用自适应滤波、卡尔曼滤波器和集合经验模态分解(ensemble empirical mode decomposition，EEMD)-独立成分分析(Independent component analysis,ICA)方法对比了单通道EEG运动伪迹的去除效果，发现EEMD-ICA相比与另外两种方法有较好的效果。接下来研究人员又提出了集合经验模态分解(EEMD)-典型相关成分分析(Canonical correlation analysis，CCA)方法，运动伪迹去除的效果得到了提升，但是这个方法仍然不能去除完全去除运动伪迹成分。主成分分析(Principal  component analysis，PCA)最早被引入脑电图分析中，作者使用它来凭经验确定眼睛活动的空间分布。目前PCA通常作为多通道EEG信号的降维步骤。
目前去除EEG运动伪迹主要分为两种，一是这些研究要么大多数都局限于高度受控的实验室环境，例如在跑步机上行走，要么就是需要额外的参考信息，例如使用惯性传感器作为参考。第二种是使用盲源分离技术，列如ICA和CCA。ICA使用高阶统计量获得统计意义上的独立源，CCA使用二阶统计量获得统计意义上的独立源，两种方法对于完全分离出运动伪迹成分，去除伪迹效果有限。
发明内容
本发明实施例提供了一种基于EEMD-PCA去除EEG信号中运动伪迹的方法及装置，相比于目前技术去伪效果得到显著提升。
根据本发明的一实施例，提供了一种基于EEMD-PCA去除EEG信号中运动伪迹的方法，包括以下步骤：
基于EEMD分解单通道EEG信号，得到各阶的本征模态函数；
基于PCA将各阶本征模态函数分离出主成分；
计算出每个主成分的自相关性；
将自相关性大于预设阈值的主成分被判定为伪迹主成分；
将被判定为伪迹的主成分进行去除；
将剩余的主成分经过PCA逆变化再经过EEMD逆变化，得到去除噪声后的EEG信号。
进一步地，在基于EEMD分解单通道EEG信号，得到各阶的本征模态分量中包括：
对含有运动伪迹的EEG信号进行减去均值处理，得到去除直流信号的EEG信号X(t)；
对信号X(t)进行集合经验模态分解，得到各阶本征模态函数。
进一步地，在基于PCA将各阶本征模态函数分离出主成分中包括：
对各阶本征模态函数进行主成分分离，得到N个源成分S。
进一步地，在计算出每个主成分的自相关性中包括：
计算每个源成分S的自相关性。
进一步地，在自相关性大于预设阈值的主成分被判定为伪迹主成分之前还包括：预设自相关性的阙值。
进一步地，在将被判定为伪迹的主成分进行去除中包括：
通过进行置零处理，将被识别为伪迹成分的源成分S进行去除，去除后得到源成分S'。
进一步地，在对剩余的主成分进行PCA逆变化处理后再进行EEMD逆变化处理，得到去除噪声后的EEG信号包括：
将源成分S'经过PCA逆变化后处理再经过EEMD逆变化处理，得到去除噪声后的EEG信号。
进一步地，在对信号X(t)进行集合经验模态分解，得到N个本征模态函数之前还包括：
设置集合经验模态分解所添加的白噪声的幅值标准差值和噪声加入的次数。
进一步地，将集合经验模态分解所添加的白噪声的幅值标准差设置为0.05-0.15之间、噪声加入次数设置为80-120次之间。
一种基于EEMD-PCA去除EEG信号中运动伪迹的装置，该装置包括：
信号分解模块，用于基于EEMD分解单通道EEG信号，得到各阶的本征模态函数；
主成分分离模块，用于基于PCA将各阶本征模态函数分离出主成分；
相关性计算模块，用于计算出每个主成分的自相关性；
伪迹成分判定模块，用于将自相关性大于预设阈值的主成分被判定为伪迹主成分；
伪迹成分去除模块，用于将被判定为伪迹的主成分进行去除；
逆变化处理模块，用于将剩余的主成分经过PCA逆变化再经过EEMD逆变化，得到去除噪声后的EEG信号。
本发明实施例中的基于EEMD-PCA去除EEG信号中运动伪迹的方法及装置基于EEMD分解单通道EEG信号，得到各阶的本征模态函数；基于PCA将各阶本征模态函数分离出主成分；计算出每个主成分的自相关性；将自相关性大于预设阈值的主成分被判定为伪迹主成分；将被判定为伪迹的主成分进行去除；将剩余的主成分经过PCA逆变化再经过EEMD逆变化，得到去除噪声后的EEG信号。本发明根据运动伪迹与EEG信号相关性不大的特点使用PCA方法分离出伪迹成分，然后根据自相关性自动选择伪迹成分，并去除伪迹成分保留有用的脑电信息；本发明伪迹去除方法，相比于目前技术去伪效果得到显著提升。
附图说明
此处所说明的附图用来提供对本发明的进一步理解，构成本申请的一部分，本发明的示意性实施例及其说明用于解释本发明，并不构成对本发明的不当限定。在附图中：
图1为本发明基于EEMD-PCA去除EEG信号中运动伪迹的方法的流程图；
图2为本发明含有运动伪迹EEG信号和纯净的EEG信号图；
图3为本发明含有运动伪迹EEG分解的本征模态分量；
图4为本发明本征模态分量经过PCA分解后得到的主成分图；
图5为本发明主成分对应的自相关系数线以上的识别为伪迹成分图；
图6为本发明去除伪迹后的EEG信号与纯净的EEG信号对比图；
图7为本发明技术与现有技术的效果对比图；
图8为本发明基于EEMD-PCA去除EEG信号中运动伪迹的装置的原理图。
附图标记：201-信号分解模块、202-主成分分离模块、203-相关性计算模块、204-伪迹成分判定模块、206-伪迹成分去除模块。
具体实施方式
为了使本技术领域的人员更好地理解本发明方案，下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分的实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都应当属于本发明保护的范围。
需要说明的是，本发明的说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的本发明的实施例能够以除了在这里图示或描述的那些以外的顺序实施。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或单元的过程、方法、系统、产品或设备不必限于清楚地列出的那些步骤或单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。
实施例1
根据本发明一实施例，提供了一种基于EEMD-PCA去除EEG信号中运动伪迹的方法，参见图1，包括以下步骤：
S101：基于EEMD分解单通道EEG信号，得到各阶的本征模态函数；
S102：基于PCA将各阶本征模态函数分离出主成分；
S103：计算出每个主成分的自相关性；
S104：将自相关性大于预设阈值的主成分被判定为伪迹主成分；
S105：将被判定为伪迹的主成分进行去除；
S106：对剩余的主成分进行PCA逆变化处理后再进行EEMD逆变化处理，得到去除噪声后的EEG信号。
本发明实施例中的基于EEMD-PCA去除EEG信号中运动伪迹的方法及装置基于EEMD分解单通道EEG信号，得到各阶的本征模态函数；基于PCA将各阶本征模态函数分离出主成分；计算出每个主成分的自相关性；将自相关性大于预设阈值的主成分被判定为伪迹主成分；将被判定为伪迹的主成分进行去除；将剩余的主成分经过PCA逆变化再经过EEMD逆变化，得到去除噪声后的EEG信号。本发明根据运动伪迹与EEG信号相关性不大的特点使用PCA方法分离出伪迹成分，然后根据自相关性自动选择伪迹成分，并去除伪迹成分保留有用的脑电信息；本发明伪迹去除方法，相比于目前技术去伪效果得到显著提升。
本实施例中，在基于EEMD分解单通道EEG信号，得到各阶的本征模态分量中包括：
对含有运动伪迹的EEG信号进行减去均值处理，得到去除直流信号的EEG信号X(t)；
对信号X(t)进行集合经验模态分解，得到各阶本征模态函数。
具体的，该对信号X(t)进行集合经验模态分解，得到各阶本征模态函数的步骤如下：
步骤一：在噪声信号W(t)中加入信号X(t)得到信号X'(t)；
步骤二：将信号X'(t)进行经验模态分解,得到各阶本征模态函数分量，其中分解后的剩余分量为r n(t)；
步骤三：重复步骤一和步骤二，每次加入强度相同序列不同的白噪声；
步骤四：利用白噪声频谱的均值为零，将各阶本征模态函数求平均值，得到最终的本征模态函数分量，即得到N个本征模态函数l(t)＝[l-1.(t),l-2.(t),...,l-n.(t)] -T并构成l(t)协方差矩阵。
本实施例中，在基于PCA将各阶本征模态函数分离出主成分中包括：
对各阶本征模态函数进行主成分分离，得到N个源成分S；
具体地，该对各阶本征模态函数进行主成分分离，得到N个源成分S的步骤如下：
步骤一：计算步骤四中得到的l(t)协方差矩阵；
步骤二：计算l(t)协方差矩阵的特征值与特征向量；
步骤三：对特征值按照从大到小进行排序，将特征值对应的特征向量分别作为列向量组成特征向量矩阵；
步骤四：将数据转换到n个特征向量构建的新空间中，得到N个源成分S(t)＝[S-1.(t),S-2.(t),...,S-n.(t)] -T。
本实施例中，4.根据权利要求3的一种基于EEMD-PCA去除EEG信号中运动伪迹的方法，其特征在于，在计算出每个主成分的自相关性中包括：
计算每个源成分S的自相关性；
本实施例中，在自相关性大于预设阈值的主成分被判定为伪迹主成分之前还包括：预设自相关性的阙值。
其中预设阙值设定为0.97，当计算出的自相关性大于预设阙值0.97时，则将自相关性大于预设阈值的主成分被判定为伪迹主成分。
参见图5所示，在图中0.9-1之间的直线为自相关系数线，在该自相关系数线以上的成分则被判定为伪迹成分。
本实施例中，在将被判定为伪迹的主成分进行去除中包括：
通过进行置零处理，将被识别为伪迹成分的源成分S进行去除，去除后得到源成分S'。
本实施例中，在对剩余的主成分进行PCA逆变化处理后再进行EEMD逆变化处理，得到去除噪声后的EEG信号中包括：
将源成分S'经过PCA逆变化处理后再经过EEMD逆变化处理，得到去除噪声后的EEG信号。
本实施例中，在对信号X(t)进行集合经验模态分解，得到N个本征模态函数之前还包括：
设置集合经验模态分解所添加的白噪声的幅值标准差值和噪声加入的次数。
本实施例中，将集合经验模态分解所添加的白噪声的幅值标准差设置为0.05-0.15之间、噪声加入次数设置为80-120次之间。
具体地，将集合经验模态分解(EEMD)所添加的白噪声的幅值标准差设置为0.1、噪声加入次数设置为100次。
设置EEMD所添加的白噪声的幅值标准差为0.1、噪声加入次数设置为100次。
参见图2，地面真实信号(Ground truth signal)代表EEG信号中的纯净的信号，在图中显示基本呈平缓的直线；噪声信号(Noisy signal)，即EEG信号中运动伪迹，在图中的幅度波动很大；可以看到在未经过本发明处理的EEG信号中，运动伪迹成分明显且与纯净信号偏离很大。
参见图3，图3为对EEG信号分解后得到的本征模态函数。
参见图4和5，在对本征模态函数经过PCA分解后得到分离出的N个主成分，即N个源成分S，然后计算每个源成分的自相关性，自相关性大于0.97的， 即源成分位于自相关系数线以上的就被视为伪迹成分。
参见图6，干净信号(Cleaned signal)为经过本发明方法去除伪迹后的信号，图6中可以看到去除伪迹后的Cleaned signal代表的线条与Ground truth signal代表的线条很大程度的重合。即去除伪迹后的EEG信号与纯净的EEG信号基本重合，消除伪迹得到了去除噪声后的EEG信号。
参见图7，本发明在公开的两通道EEG信号中去除运动伪迹验证数据集上进行了测试，相比现有的EEMD-ICA和EEMD-CCA技术效果得到显著提升；使用本发明EEG信号去除伪迹成分后与原始信号的均方根误差降低了24.2、信噪比提高了9.2分贝以及相似性提高了0.24。
其中，均方根误差是差分评价指标，数值越大表示信号间数值差越大；信噪比的值越高，意味着包含的噪声越少，去噪效果越好；相似度的值越接近1，则代表去噪后信号与真实信号越接近。
实施例2
根据本发明的另一实施例，提供了一种基于EEMD-PCA去除EEG信号中运动伪迹的装置，参见图8，包括：
信号分解模块201，用于基于EEMD分解单通道EEG信号，得到各阶的本征模态函数；
主成分分离模块202，用于基于PCA将各阶本征模态函数分离出主成分；
相关性计算模块203，用于计算出每个主成分的自相关性；
伪迹成分判定模块204，用于将自相关性大于预设阈值的主成分被判定为伪迹主成分；
伪迹成分去除模块205，用于将被判定为伪迹的主成分进行去除；
逆变化处理模块206，用于对剩余的主成分进行PCA逆变化处理后再进行EEMD逆变化处理，得到去除噪声后的EEG信号。
本发明实施例中的基于EEMD-PCA去除EEG信号中运动伪迹的方法及装置基于EEMD分解单通道EEG信号，得到各阶的本征模态函数；基于PCA将各阶本征模态函数分离出主成分；计算出每个主成分的自相关性；将自相关性大于预设阈值的主成分被判定为伪迹主成分；将被判定为伪迹的主成分进行去除；将剩余的主成分经过PCA逆变化再经过EEMD逆变化，得到去除噪声后的EEG信号。本发明根据运动伪迹与EEG信号相关性不大的特点使用PCA方法分离出伪迹成分，然后根据自相关性自动选择伪迹成分，并去除伪迹成分保留有用的脑电信息；本发明伪迹去除方法，相比于目前技术去伪效果得到显著提升。
本实施例中，在基于EEMD分解单通道EEG信号，得到各阶的本征模态分量中包括：
对含有运动伪迹的EEG信号进行减去均值处理，得到去除直流信号的EEG信号X(t)；
对信号X(t)进行集合经验模态分解，得到各阶本征模态函数。
具体的，该对信号X(t)进行集合经验模态分解，得到各阶本征模态函数的步骤如下：
步骤一：在噪声信号W(t)中加入信号X(t)得到信号X'(t)；
步骤二：将信号X'(t)进行经验模态分解,得到各阶本征模态函数分量，其中分解后的剩余分量为r n(t)；
步骤三：重复步骤一和步骤二，每次加入强度相同序列不同的白噪声；
步骤四：利用白噪声频谱的均值为零，将各阶本征模态函数求平均值，得到最终的本征模态函数分量，即得到N个本征模态函数l(t)＝[l-1.(t),l-2.(t),...,l-n.(t)] -T并构成l(t)协方差矩阵。
本实施例中，在基于PCA将各阶本征模态函数分离出主成分中包括：
对各阶本征模态函数进行主成分分离，得到N个源成分S；
具体地，该对各阶本征模态函数进行主成分分离，得到N个源成分S的步骤如下：
步骤一：计算步骤四中得到的l(t)协方差矩阵；
步骤二：计算l(t)协方差矩阵的特征值与特征向量；
步骤三：对特征值按照从大到小进行排序，将特征值对应的特征向量分别作为列向量组成特征向量矩阵；
步骤四：将数据转换到n个特征向量构建的新空间中，得到N个源成分S(t)＝[S-1.(t),S-2.(t),...,S-n.(t)] -T。
本实施例中，在计算出每个主成分的自相关性中包括：
计算每个源成分S的自相关性；
本实施例中，在自相关性大于预设阈值的主成分被判定为伪迹主成分之前还包括：预设自相关性的阙值。
其中预设阙值设定为0.97，当计算出的自相关性大于预设阙值0.97时，则将自相关性大于预设阈值的主成分被判定为伪迹主成分。
参见图5所示，在图中0.9-1之间的直线为自相关系数线，在该自相关系数线以上的成分则被判定为伪迹成分。
本实施例中，在将被判定为伪迹的主成分进行去除中包括：
通过进行置零处理，将被识别为伪迹成分的源成分S进行去除，去除后得到源成分S'。
本实施例中，在对剩余的主成分进行PCA逆变化处理后再进行EEMD逆变化处理，得到去除噪声后的EEG信号中包括：
源成分S'经过PCA逆变化后再经过EEMD逆变化，得到去除噪声后的EEG 信号。
本实施例中，在对信号X(t)进行集合经验模态分解，得到N个本征模态函数之前还包括：
设置集合经验模态分解所添加的白噪声的幅值标准差值和噪声加入的次数。
本实施例中，将集合经验模态分解所添加的白噪声的幅值标准差设置为0.05-0.15之间、噪声加入次数设置为80-120次之间。
具体地，将集合经验模态分解(EEMD)所添加的白噪声的幅值标准差设置为0.1、噪声加入次数设置为100次。
设置EEMD所添加的白噪声的幅值标准差为0.1、噪声加入次数设置为100次。
参见图2，地面真实信号(Ground truth signal)，代表EEG信号中的纯净的信号，在图中显示基本呈平缓的直线；噪声信号(Noisy signal)，即EEG信号中运动伪迹，在图中的幅度波动很大；可以看到在未经过本发明处理的EEG信号中，运动伪迹成分明显且与纯净信号偏离很大。
参见图3，图3为对EEG信号分解后得到的本征模态函数。
参见图4和5，在对本征模态函数经过PCA分解后得到分离出的N个主成分，即N个源成分S，然后计算每个源成分的自相关性，自相关性大于0.97的，即源成分位于自相关系数线以上的就被视为伪迹成分。
参见图6，干净信号(Cleaned signal)为经过本发明方法去除伪迹后的信号，图6中可以看到去除伪迹后的Cleaned signal代表的线条与Ground truth signal代表的线条很大程度的重合。即去除伪迹后的EEG信号与纯净的EEG信号基本重合，消除伪迹得到了去除噪声后的EEG信号。
参见图7，本发明在公开的两通道EEG信号中去除运动伪迹验证数据集上进行了测试，相比现有的EEMD-ICA和EEMD-CCA技术效果得到显著提升；使用 本发明EEG信号去除伪迹成分后与原始信号的均方根误差降低了24.2、信噪比提高了9.2分贝以及相似性提高了0.24。
其中，均方根误差是差分评价指标，数值越大表示信号间数值差越大；信噪比的值越高，意味着包含的噪声越少，去噪效果越好；相似度的值越接近1，则代表去噪后信号与真实信号越接近。
以上所述仅是本发明的优选实施方式，应当指出，对于本技术领域的普通技术人员来说，在不脱离本发明原理的前提下，还可以做出若干改进和润饰，这些改进和润饰也应视为本发明的保护范围。

Claims (10)

  1. 一种基于EEMD-PCA去除EEG信号中运动伪迹的方法，其特征在于，包括以下步骤：
    基于EEMD分解单通道EEG信号，得到各阶的本征模态函数；
    基于PCA将各阶所述本征模态函数分离出主成分；
    计算出每个主成分的自相关性；
    将自相关性大于预设阈值的主成分被判定为伪迹主成分；
    将被判定为伪迹的主成分进行去除；
    对剩余的主成分进行PCA逆变化处理后再进行EEMD逆变化处理，得到去除噪声后的EEG信号。
  2. 根据权利要求1所述的一种基于EEMD-PCA去除EEG信号中运动伪迹的方法，其特征在于，在所述基于EEMD分解单通道EEG信号，得到各阶的本征模态分量中包括：
    对含有运动伪迹的EEG信号进行减去均值处理，得到去除直流信号的EEG信号X(t)；
    对信号X(t)进行集合经验模态分解，得到各阶本征模态函数。
  3. 根据权利要求2所述的一种基于EEMD-PCA去除EEG信号中运动伪迹的方法，其特征在于，在所述基于PCA将各阶所述本征模态函数分离出主成分中包括：
    对各阶本征模态函数进行主成分分离，得到N个源成分S。
  4. 根据权利要求3所述的一种基于EEMD-PCA去除EEG信号中运动伪迹的方法，其特征在于，在所述计算出每个主成分的自相关性中包括：
    计算每个源成分S的自相关性。
  5. 根据权利要求4所述的一种基于EEMD-PCA去除EEG信号中运动伪迹的方 法，其特征在于，在所述自相关性大于预设阈值的主成分被判定为伪迹主成分之前还包括：预设自相关性的阙值。
  6. 根据权利要求5所述的一种基于EEMD-PCA去除EEG信号中运动伪迹的方法，其特征在于，在所述将被判定为伪迹的主成分进行去除中包括：
    通过进行置零处理，将被识别为伪迹成分的源成分S进行去除，去除后得到源成分S'。
  7. 根据权利要求1所述的一种基于EEMD-PCA去除EEG信号中运动伪迹的方法，其特征在于，在所述对剩余的主成分进行PCA逆变化处理后再进行EEMD逆变化处理，得到去除噪声后的EEG信号包括：
    将所述源成分S'经过PCA逆变化处理后再经过EEMD逆变化处理，得到去除噪声后的EEG信号。
  8. 根据权利要求2所述的一种基于EEMD-PCA去除EEG信号中运动伪迹的方法，其特征在于，在所述对信号X(t)进行集合经验模态分解，得到N个本征模态函数之前还包括：
    设置集合经验模态分解所添加的白噪声的幅值标准差值和噪声加入的次数。
  9. 根据权利要求8所述的一种基于EEMD-PCA去除EEG信号中运动伪迹的方法，其特征在于，将集合经验模态分解所添加的白噪声的幅值标准差设置为0.05-0.15之间、噪声加入次数设置为80-120次之间。
  10. 一种基于EEMD-PCA去除EEG信号中运动伪迹的装置，其特征在于，包括：
    信号分解模块，用于基于EEMD分解单通道EEG信号，得到各阶的本征模态函数；
    主成分分离模块，用于基于PCA将各阶所述本征模态函数分离出主成分；
    相关性计算模块，用于计算出每个主成分的自相关性；
    伪迹成分判定模块，用于将自相关性大于预设阈值的主成分被判定为伪迹主成分；
    伪迹成分去除模块，用于将被判定为伪迹的主成分进行去除；
    逆变化处理模块，用于对剩余的主成分进行PCA逆变化处理后再进行EEMD逆变化处理，得到去除噪声后的EEG信号。
PCT/CN2021/137312 2021-02-05 2021-12-12 基于eemd-pca去除eeg信号中运动伪迹的方法及装置 WO2022166401A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110161627.5A CN112957055B (zh) 2021-02-05 2021-02-05 基于eemd-pca去除eeg信号中运动伪迹的方法及装置
CN202110161627.5 2021-02-05

Publications (1)

Publication Number Publication Date
WO2022166401A1 true WO2022166401A1 (zh) 2022-08-11

Family

ID=76274453

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/137312 WO2022166401A1 (zh) 2021-02-05 2021-12-12 基于eemd-pca去除eeg信号中运动伪迹的方法及装置

Country Status (2)

Country Link
CN (1) CN112957055B (zh)
WO (1) WO2022166401A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116909408A (zh) * 2023-09-13 2023-10-20 中物联讯(北京)科技有限公司 一种基于mr智能眼镜的内容交互方法
CN117349661A (zh) * 2023-12-04 2024-01-05 浙江大学高端装备研究院 柱塞泵振动信号特征提取方法、装置、设备和存储介质
CN118806298A (zh) * 2024-06-19 2024-10-22 天津大学 一种普适环境中脑电信号的自适应去噪方法

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112957055B (zh) * 2021-02-05 2023-12-29 中国科学院深圳先进技术研究院 基于eemd-pca去除eeg信号中运动伪迹的方法及装置
CN117828274B (zh) * 2024-01-05 2024-07-19 天津大学 一种从脑电信号中去除人工耳蜗伪迹的方法

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102631195A (zh) * 2012-04-18 2012-08-15 太原科技大学 人体表面肌电信号单通道盲源分离法
CN106805945A (zh) * 2017-01-22 2017-06-09 合肥工业大学 一种少数通道的脑电信号中肌电伪迹的消除方法
CN108231087A (zh) * 2017-12-14 2018-06-29 宁波升维信息技术有限公司 一种单通道盲源分离法
CN108309290A (zh) * 2018-02-24 2018-07-24 华南理工大学 单通道脑电信号中肌电伪迹的自动去除方法
US20180279960A1 (en) * 2017-03-31 2018-10-04 Agency For Science, Technology And Research Method and apparatus for real-time discriminative ocular artefact removal from eeg signals
CN109480832A (zh) * 2018-10-31 2019-03-19 合肥工业大学 一种单通道的脑电信号中肌电伪迹的消除方法
CN111904414A (zh) * 2020-07-30 2020-11-10 太原科技大学 一种多入单出脑电信号采集器
CN112957055A (zh) * 2021-02-05 2021-06-15 中国科学院深圳先进技术研究院 基于eemd-pca去除eeg信号中运动伪迹的方法及装置

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112294340A (zh) * 2020-10-27 2021-02-02 中国科学院苏州生物医学工程技术研究所 快速自动去除肌电伪迹的方法、系统、存储介质及计算机设备

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102631195A (zh) * 2012-04-18 2012-08-15 太原科技大学 人体表面肌电信号单通道盲源分离法
CN106805945A (zh) * 2017-01-22 2017-06-09 合肥工业大学 一种少数通道的脑电信号中肌电伪迹的消除方法
US20180279960A1 (en) * 2017-03-31 2018-10-04 Agency For Science, Technology And Research Method and apparatus for real-time discriminative ocular artefact removal from eeg signals
CN108231087A (zh) * 2017-12-14 2018-06-29 宁波升维信息技术有限公司 一种单通道盲源分离法
CN108309290A (zh) * 2018-02-24 2018-07-24 华南理工大学 单通道脑电信号中肌电伪迹的自动去除方法
CN109480832A (zh) * 2018-10-31 2019-03-19 合肥工业大学 一种单通道的脑电信号中肌电伪迹的消除方法
CN111904414A (zh) * 2020-07-30 2020-11-10 太原科技大学 一种多入单出脑电信号采集器
CN112957055A (zh) * 2021-02-05 2021-06-15 中国科学院深圳先进技术研究院 基于eemd-pca去除eeg信号中运动伪迹的方法及装置

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116909408A (zh) * 2023-09-13 2023-10-20 中物联讯(北京)科技有限公司 一种基于mr智能眼镜的内容交互方法
CN116909408B (zh) * 2023-09-13 2024-02-09 中物联讯(北京)科技有限公司 一种基于mr智能眼镜的内容交互方法
CN117349661A (zh) * 2023-12-04 2024-01-05 浙江大学高端装备研究院 柱塞泵振动信号特征提取方法、装置、设备和存储介质
CN117349661B (zh) * 2023-12-04 2024-02-20 浙江大学高端装备研究院 柱塞泵振动信号特征提取方法、装置、设备和存储介质
CN118806298A (zh) * 2024-06-19 2024-10-22 天津大学 一种普适环境中脑电信号的自适应去噪方法

Also Published As

Publication number Publication date
CN112957055B (zh) 2023-12-29
CN112957055A (zh) 2021-06-15

Similar Documents

Publication Publication Date Title
WO2022166401A1 (zh) 基于eemd-pca去除eeg信号中运动伪迹的方法及装置
Bono et al. Hybrid wavelet and EMD/ICA approach for artifact suppression in pervasive EEG
CN113349752A (zh) 一种基于传感融合的可穿戴设备实时心率监测方法
KR101987225B1 (ko) Vr 멀미 검출 장치 및 방법
Sameni et al. An iterative subspace denoising algorithm for removing electroencephalogram ocular artifacts
Maddirala et al. ICA with CWT and k-means for eye-blink artifact removal from fewer channel EEG
CN113827253B (zh) 计算设备和从脑电信号中去除噪声的方法
CN116369950A (zh) 一种基于脑电溯源和多特征提取的目标检测方法
CN109657646A (zh) 生理时间序列的特征表示与提取方法、装置及存储介质
CN113208633A (zh) 一种基于eeg脑波的情绪识别方法、系统
Yong et al. Generalized morphological component analysis for EEG source separation and artifact removal
Li et al. An improved MAMA-EMD for the automatic removal of EOG artifacts
Devulapalli et al. A hybrid ICA Kalman predictor algorithm for ocular artifacts removal
Ahmed et al. Effective hybrid method for the detection and rejection of electrooculogram (EOG) and power line noise artefacts from electroencephalogram (EEG) mixtures
Gao et al. FBLPF-ABOW: an effective method for blink artifact removal in single-channel EEG signal
Mohammadi et al. Comparison of two methods of removing EOG artifacts for use in a motor imagery-based brain computer interface
Kaur et al. EEG artifact suppression based on SOBI based ICA using wavelet thresholding
AU2021104217A4 (en) A system and method for optimum wavelet basis function selection for ecg arrhythmia denoising using artificial intelligence
CN116942143A (zh) 一种基于智能耳戴设备惯性传感器的坐姿行为监测方法
CN115998249A (zh) 脑电图中的伪影处理方法、装置、设备及存储介质
Massar et al. Blind Source Separation of EEG Signals Using Wavelet and EMD Decomposition
CN112842368A (zh) 一种识别表面肌电信号的系统及方法
Visu et al. Enhanced EEG classification using adaptive DWT and heuristic-ICA algorithm
Gouy-Pailler et al. Iterative subspace decomposition for ocular artifact removal from EEG recordings
CN116383709B (zh) 信号噪声的检测方法、信号检测模型、可读存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21924394

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21924394

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 11.01.2024)

122 Ep: pct application non-entry in european phase

Ref document number: 21924394

Country of ref document: EP

Kind code of ref document: A1