CN113040788A - Electroencephalogram signal quality detection method based on spectrum analysis - Google Patents
Electroencephalogram signal quality detection method based on spectrum analysis Download PDFInfo
- Publication number
- CN113040788A CN113040788A CN202110613283.7A CN202110613283A CN113040788A CN 113040788 A CN113040788 A CN 113040788A CN 202110613283 A CN202110613283 A CN 202110613283A CN 113040788 A CN113040788 A CN 113040788A
- Authority
- CN
- China
- Prior art keywords
- electroencephalogram signal
- electroencephalogram
- fitting function
- quality
- signal quality
- 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.)
- Granted
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
-
- 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
- A61B5/7257—Details of waveform analysis characterised by using transforms using Fourier transforms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Medical Informatics (AREA)
- Signal Processing (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Physiology (AREA)
- Psychiatry (AREA)
- Pathology (AREA)
- Heart & Thoracic Surgery (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention discloses a method for detecting the quality of an electroencephalogram signal based on spectrum analysis, which comprises the steps of obtaining the electroencephalogram signal in a time window; performing fast Fourier transform on the electroencephalogram signal to obtain multiple groups of frequency domain data, and obtaining power spectral density according to the frequency domain dataPSD(F) (ii) a To power spectral densityPSD(F) In (2) the non-periodic component and the periodic component are respectively subjected to non-neutralizationAnd performing linear fitting to obtain a plurality of characteristic parameters, and performing combined evaluation on the plurality of characteristic parameters to obtain a detection result of the electroencephalogram signal. By using the invention, a decomposition model of the electroencephalogram signal can be established in a frequency domain, fitting is carried out on the non-periodic component and the periodic component, the noise type is quantized, and the interpretability of the signal quality is increased.
Description
Technical Field
The invention relates to the technical field of digital signal processing, in particular to a method for detecting electroencephalogram signal quality based on spectrum analysis.
Background
The electroencephalogram signal is used as an exploration information source for the brain of a human body, can reflect the thinking process of the human body, can also reflect the emotional changes of the human body in different states and the like; as a bioelectricity signal, the bioelectricity signal can be generally acquired through an electrode contacting with the head to obtain a voltage signal including an electroencephalogram signal, so that the electroencephalogram signal can be used as a brain-computer interface to carry out information interaction with a head-mounted device.
Because the electroencephalogram signals are irregular and unstable and are weaker than the electrocardiosignals, the electromyogram signals and the like, whether the electroencephalogram signals exist in the acquired voltage signals or whether the acquired electroencephalogram signals are credible is difficult to identify, and therefore, the quality judgment of the electroencephalogram signals has important guiding significance and practical significance.
Scalp electroencephalography (EEG) is one of the most common and economical non-invasive means of electroencephalography, and collects microvolt-level signals generated by synchronous neuronal activity within the brain by placing electrodes at specific locations on the scalp, but this approach is susceptible to various disturbances from environmental and human factors, such as mains frequency disturbances of alternating current, poor electrode contact, eye movement, and so forth. In the electroencephalogram signal acquisition process, the judgment of the quality of the electroencephalogram signal is an important ring for effectively guaranteeing the accuracy of subsequent electroencephalogram application analysis.
Most of existing electroencephalogram signal quality evaluation methods are based on time domain waveform characteristics for analysis, for example, statistics such as mean value, variance, energy and the like of electroencephalogram segments are measured in a time window with a certain length, and the signal quality is evaluated according to a set hard threshold. Moreover, the time domain waveform of the electroencephalogram signal is different in different age stages and pathology, for example, the neonatal brain wave has no obvious alpha rhythm and is often expressed as paroxysmal rhythmic activity, and the time domain waveform of the electroencephalogram signal is greatly influenced by noise. Therefore, the invention provides a method for detecting the quality of the electroencephalogram signal based on the frequency spectrum analysis, and the quality of the electroencephalogram signal is comprehensively judged by analyzing the characteristics of the electroencephalogram signal on the frequency domain.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: in order to solve the technical problem of poor reliability of the electroencephalogram signal detection method in the prior art, the invention provides the electroencephalogram signal quality detection method based on the frequency spectrum analysisPSD(F) The method comprises the steps of respectively carrying out nonlinear fitting on aperiodic components and periodic components in the power spectral density to obtain a plurality of characteristic parameters, and judging the quality of the electroencephalogram signal through the characteristic parameters.
The technical scheme adopted by the invention for solving the technical problems is as follows: a method for detecting the quality of an electroencephalogram signal based on spectral analysis is characterized by comprising the following steps:
s1: acquiring electroencephalogram signals in a time window to obtain M groups of time domain data;
s2: performing fast Fourier transform on the M groups of time domain data to obtain K groups of frequency domain data, and obtaining power spectral density according to the K groups of frequency domain dataPSD(F)；
S3: for the power spectral densityPSD(F)Carrying out nonlinear fitting on the data of the medium-aperiodic components to obtain a fitting functionL(F)Decomposing to obtain the fitting functionL(F)The first set of characteristic parameters of (1);
s4: removing the power spectral densityPSD(F)Obtaining data of periodic components from the data of non-periodic components;
s5: performing nonlinear fitting on the data of the periodic components to obtain a fitting functionG(F) n Decomposing to obtain the fitting functionG(F) n The second set of characteristic parameters of (1);
s6: and performing combined evaluation on the first characteristic parameter group and the second characteristic parameter group to obtain a detection result of the quality of the electroencephalogram signal.
The method for detecting the quality of the electroencephalogram signal based on the frequency spectrum analysis, disclosed by the invention, has the advantages that the fast Fourier transform is carried out on the acquired electroencephalogram signal, the time domain signal is changed into the frequency domain signal, and the power spectral density is establishedPSD(F) (ii) a To power spectral densityPSD(F) The non-periodic component and the periodic component are respectively subjected to non-linear fitting to obtain a plurality of characteristic parameters, and the characteristic parameters are combined and evaluated to obtain a detection result of the electroencephalogram signal. Fitting the aperiodic component and the periodic component of the electroencephalogram signal on a frequency domain by establishing a signal decomposition model, quantizing a noise parameter, and increasing the interpretability of signal quality; and the interference of power frequency to the electroencephalogram signal can be directly measured, and the accuracy of signal detection is further improved.
Further, specifically, the method further includes: fitting the fitting functionL(F)And said fitting functionG(F) n Overlapping to obtain the final fitting functionNPS(F)=L(F)+G(F) n Calculating a final fitting functionNPS(F)And the power spectral densityPSD(F)And then, performing combined evaluation on the regression evaluation index parameter, the first characteristic parameter group and the second characteristic parameter group to obtain a detection result of the electroencephalogram signal quality.
Further, specifically, the regression evaluation index parameter includes a pearson correlation coefficient R2Mean absolute error MAE, root mean square error RMSE, and mean absolute percentage error MAPE.
Further, specifically, the electroencephalogram signal is original electroencephalogram data acquired in a single channel; or the electroencephalogram signal is electroencephalogram data obtained by performing linear and nonlinear combination processing on original electroencephalogram data acquired in a plurality of channels with different spatial distributions.
Further, specifically, the electroencephalogram signal includes at least one periodic component, and a fitting function is obtained by performing nonlinear fitting on each periodic componentG(F) n Each fitting functionG(F) n A set of said second set of characteristic parameters can be resolved; and when the quality of the electroencephalogram signal is detected, at least one group of the second characteristic parameter group, the first characteristic parameter group and the regression evaluation index parameter is selected each time for combined evaluation.
Further, in particular, the fitting function
Wherein F represents a frequency variable, b represents an offset, x represents a power law distribution coefficient, and k is an inflection point; and when k and F are zero, obtaining a direct current component parameter DC, wherein the first characteristic parameter group comprises an offset b, a power law distribution coefficient x and the direct current component DC. The offset b represents the energy intensity of the low-frequency signal, common-mode noise and motion artifacts are reflected to a certain extent, and the larger the offset b is, the larger the low-frequency interference is; the power law distribution coefficient x reflects the non-periodic background signal characteristics of the electroencephalogram signal, and the distribution of the power law distribution coefficient x is normal distribution; the direct current component DC reflects the intensity of the direct current component of the electroencephalogram signal.
Further, in particular, the fitting function
Wherein F denotes a frequency variable, c denotes a center frequency, ω denotes a bandwidth, a denotes an energy level of the center frequency, and the second feature parameter group includes the bandwidth ω, the energy level a of the center frequency, and the center frequency c. The bandwidth omega can represent the bandwidth of interference frequency, and the smaller the bandwidth omega is, the more stable the voltage of power supply is; the energy height a can represent the intensity of interference frequency, the smaller a is, the smaller the power frequency interference intensity is, and whether the normal physiological signal or the power frequency interference exists can be determined according to the central frequency c.
Further, specifically, the Pearson correlation coefficient
Wherein cov (X, Y) represents the final fitting functionNPS(F)And the power spectral densityPSD(F)The covariance of (a) of (b),
representing the final fitting functionNPS(F)The variance of (a) is determined,
representing power spectral densityPSD(F)The variance of (c). Pearson's correlation coefficient represents the final fitting functionNPS(F)And power spectral densityPSD(F)The degree of correlation reflects the quality of the fitting effect, R2The value range is 0-1, and the closer to 1, the better the fitting effect is.
Further, specifically, the average absolute error
Wherein X represents the final fitting functionNPS(F)And Y represents the power spectral densityPSD(F)。
Further, specifically, the step S6 specifically includes: and evaluating and classifying the first characteristic parameter group, the second characteristic parameter group and the regression evaluation index parameter through a machine learning algorithm to obtain a detection result of the electroencephalogram signal quality.
Further, specifically, the first feature parameter group, the second feature parameter group and the regression evaluation index parameter are combined and quantified and evaluated through a feature map, so as to obtain a detection result of the electroencephalogram signal quality.
The method for detecting the quality of the electroencephalogram signal based on the spectral analysis has the advantages that the method for detecting the quality of the electroencephalogram signal based on the spectral analysis carries out fast Fourier transform on the acquired electroencephalogram signal, changes a time domain signal into a frequency domain signal and establishes the power spectral densityPSD(F) (ii) a To power spectral densityPSD(F) The periodicity of the EEG signal is divided into a plurality of periodicity components and the periodicity components are subjected to nonlinear fitting respectively to obtain a plurality of characteristic parameters, and the plurality of characteristic parameters are subjected to combined evaluation to obtain a detection result of the EEG signal. The method is characterized in that a signal decomposition model is established, fitting is carried out on the aperiodic component and the periodic component of the electroencephalogram signal in a frequency domain, noise parameters are quantized, and the interpretability of signal quality is improvedSex; the method considers the interference of factors such as electrode falling, environment white noise, power frequency interference and the like on the electroencephalogram signals, quantifies the noises through parameters, comprehensively considers the influence of various noises when evaluating the quality of the electroencephalogram signals, and improves the reliability of signal detection results. The method analyzes the quality of the electroencephalogram signal based on the frequency domain, can be used for real-time signal quality evaluation, guides a user to perform corresponding noise reduction operation on a low-quality data segment by combining with quantized parameters, and improves the usability of data.
Drawings
The invention is further illustrated with reference to the following figures and examples.
FIG. 1 is a flow chart of the method for detecting brain electrical signal quality based on spectrum analysis of the present invention.
Figure 2 is a graphical illustration of the power spectral density and the final fit function of the present invention.
FIG. 3 is a diagram of the classification results of the SVM classification model of the present invention.
Fig. 4 is a radar map of six characteristic parameters of the present invention and a schematic diagram of the area W.
Fig. 5 is a schematic diagram of a closed region Q of six characteristic parameters of the present invention.
Fig. 6 is a schematic diagram of a closed region P of six characteristic parameters of the present invention.
Fig. 7 is a schematic diagram of a closed region J of six characteristic parameters of the present invention.
Fig. 8 is a schematic diagram of a closed region T of six characteristic parameters of the present invention.
Fig. 9 is a schematic diagram of a closed region U of six characteristic parameters of the present invention.
Detailed Description
The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic views illustrating only the basic structure of the present invention in a schematic manner, and thus show only the constitution related to the present invention.
In the description of the present invention, it is to be understood that the terms "central," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," "circumferential," and the like are used in the orientations and positional relationships indicated in the drawings for convenience in describing the invention and to simplify the description, and are not intended to indicate or imply that the referenced devices or elements must have a particular orientation, be constructed and operated in a particular orientation, and are therefore not to be considered limiting of the invention. Furthermore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present invention, "a plurality" means two or more unless otherwise specified.
In the description of the present invention, it should be noted that, unless otherwise explicitly specified or limited, the terms "mounted," "connected," and "connected" are to be construed broadly, e.g., as meaning either a fixed connection, a removable connection, or an integral connection; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present invention can be understood in specific cases to those skilled in the art.
The electroencephalogram signal has periodic components and non-periodic components, the periodic components refer to components with characteristic frequencies, generally called signals of neural oscillation, such as delta, theta, alpha, beta, gamma and other specific frequency bands, and the non-periodic components refer to non-rhythmic activities or activities without characteristic frequencies.
As shown in FIG. 1, the invention provides a method for detecting the quality of an electroencephalogram signal based on spectrum analysis, which comprises the following steps:
s1: and acquiring the electroencephalogram signals in the time window to obtain M groups of time domain data.
It should be noted that the electroencephalogram signal in the acquisition time window may be original electroencephalogram data acquired through a single channel; or, the electroencephalogram signal is electroencephalogram data obtained by performing linear and nonlinear combination processing on original electroencephalogram data acquired in a plurality of channels with different spatial distributions, and may be, for example, electroencephalogram data obtained by performing weighted average processing on original electroencephalogram data acquired in a plurality of channels. In this embodiment, the length of the time window is not limited herein, and may be selected according to needs, for example, the length of the time window may be several seconds, several minutes, several hours, or several days, and the electroencephalogram signal acquired within 1 second may include 1000 sets of time domain data.
S2: performing fast Fourier transform on the M groups of time domain data to obtain K groups of frequency domain data, and obtaining power spectral density according to the K groups of frequency domain dataPSD(F)。
It should be noted that the acquired original electroencephalogram signal data is time domain data, the time domain data can be converted into frequency domain data by performing fast fourier transform (FFT for short), and the power spectral density can be obtained according to the frequency domain dataPSD(F)Power spectral densityPSD(F)The abscissa of (a) is frequency and the ordinate is power spectral density. The number of frequency domain data obtained is different according to different sampling frequencies selected by fast fourier transform.
S3: to power spectral densityPSD(F)Carrying out nonlinear fitting on the data of the medium-aperiodic components to obtain a fitting functionL (F)Decomposing to obtain a fitting functionL(F)The first set of characteristic parameters in (1).
The fitting function obtained by non-linearly fitting the data of the aperiodic component is
Wherein, F represents a frequency variable, b represents an offset, x represents a power law distribution coefficient, k is an inflection point, and when k and F are zero, a direct current component parameter DC can be obtained. In this embodiment, the first characteristic parameter group may include an offset b, a power law distribution coefficient x, and a direct current component DC, but is not limited thereto, and may also include other characteristic parameters for different kinds of electroencephalogram signals. For example, by inputting multiple sets of data with non-periodic components into a software program, a fitting function can be obtained
That is, the offset b and the power law distribution coefficient x can be obtained, and a function can be fitted to the aperiodic component in a section of electroencephalogram signalL(F)A fitting functionL(F)A set of first sets of characteristic parameters may be resolved. The brain electrical signals are transmitted from the neuron population to the scalp, and the attenuation of the signals is in power law distribution. There are four main categories of human brain waves: alpha waves, beta waves, Q waves and delta waves, alpha waves (8-13 hz) also known as "alpha rhythm", which is also known as "resting waves" (brain waves when a person is in a state of waking to be most calm) and "longevity waves" (which are beneficial to good health and longevity), which are sufficient to be of vital importance to the mental and physical health of a person. The frequency of alpha rhythm is closely related to age, alpha rhythm is not formed in the occipital region of infant, and the initial alpha rhythm appears at about 3 years old, which is about 8 Hz; near adult levels by age 10, up to 10Hz and mixed with delta and theta waves. In the prior art, when the quality of the electroencephalogram signals is detected by adopting time domain analysis, because the time domain waveforms of the electroencephalogram signals at different ages are different in performance, the electroencephalogram signals cannot be judged by using a uniform threshold value, and the judgment result is easy to deviate. In different levels of neuronal population activity, the shape changes of the neural power spectrum reveal different dynamics within the population. A change in the power law distribution coefficient x implies a change in the correlation between neurons, while a shift in the offset b implies an overall increase or decrease in population activity.
S4: removing power spectral densityPSD(F) And obtaining data of periodic components from the data of non-periodic components.
It is noted that the power spectral density is removedPSD(F) Of (2) a non-periodic component such that the power spectral densityPSD(F) Only the periodic component is left, so that the subsequent nonlinear fitting of the periodic component is facilitated, and the interference is eliminated.
S5: carrying out nonlinear fitting on the data of the periodic components to obtain a fitting functionG(F) n Decomposing to obtain a fitting functionG(F) n The second set of characteristic parameters in (1).
It should be noted that the periodic component is at powerThe spectrum can be presented as a peak value which reflects the power of specific frequency, the electroencephalogram signal can comprise a plurality of periodic components (a plurality of peak values), the peak values obey Gaussian distribution, and the fitting function can be obtained by carrying out nonlinear fitting on the periodic components
Where F denotes a frequency variable, c denotes a center frequency, ω denotes a bandwidth, and a denotes an energy height of the center frequency, in this embodiment, the second characteristic parameter group may include the bandwidth ω, the energy height a of the center frequency, and the center frequency c, but is not limited thereto, and the second characteristic parameter group may further include other characteristic parameters. It will be appreciated that a peak may be fitted to a functionG(F) n A function ofG(F) n A second set of characteristic parameters can be analyzed, the aperiodic component in the electroencephalogram signal can comprise a physiological signal and a noise signal, and whether the peak is the physiological signal or the noise signal can be analyzed according to the second set of characteristic parameters. For example, when the electroencephalogram signal receives power frequency interference (for example, 50Hz or 60 Hz), corresponding bandwidth and energy height appear on a fitting curve of gaussian distribution, so that the power frequency interference can be directly measured, and the quality of the electroencephalogram signal can be conveniently evaluated subsequently.
S6: and carrying out combined evaluation on the obtained first characteristic parameter group and the second characteristic parameter group to obtain a detection result of the quality of the electroencephalogram signal.
It should be noted that, when performing electroencephalogram signal quality detection, each time, a group of second characteristic parameter groups and the first characteristic parameter groups are selected to perform combined evaluation, so as to obtain a quality detection result of the electroencephalogram signal. And if the acquired electroencephalogram signal contains a plurality of peaks, sequentially carrying out combined analysis on a second characteristic parameter group and a first characteristic parameter group obtained by fitting each peak, and obtaining a quality detection result of the whole electroencephalogram signal.
In this embodiment, the method further comprises fitting a function to the dataL(F)And fitting functionG(F) n Overlapping to obtain the final fitting functionNPS(F)=L(F)+G(F) n Calculating a final fitting functionNPS(F)And power spectral densityPSD(F)And then, carrying out combined evaluation on the regression evaluation index parameter, the first characteristic parameter group and the second characteristic parameter group to obtain a detection result of the quality of the electroencephalogram signal. It should be noted that the regression evaluation index parameter may include a pearson correlation coefficient R2Mean absolute error MAE, root mean square error RMSE, and mean absolute percentage error MAPE, among others. For example, Pearson's correlation coefficient
Wherein cov (X, Y) represents the final fitting functionNPS(F)And power spectral densityPSD(F)The covariance of (a) of (b),
representing the final fitting functionNPS(F)The variance of (a) is determined,
representing power spectral densityPSD(F)The variance of (c). Coefficient of correlation R2Representing the final fitting functionNPS(F)And power spectral densityPSD(F)The degree of correlation reflects the quality of the fitting result, R2The value range is 0-1, and the closer to 1, the better the fitting effect is. E.g. mean absolute error
Wherein X represents the final fitting functionNPS(F)Y said power spectral densityPSD(F)。The smaller the error value, the better the fitting effect.
For example, the first feature parameter group, the second feature parameter group and the regression evaluation index parameter may be evaluated and classified by a machine learning algorithm to obtain a detection result of electroencephalogram signal quality. For example, the machine learning algorithm may be a SVM classification algorithm, a neural network classification algorithm, or the like.
For example, the first feature parameter group, the second feature parameter group, and the regression evaluation index parameter may be combined and quantitatively evaluated through the feature map to obtain a detection result of electroencephalogram signal quality. The characteristic legend may be, for example, a radar chart, a bar chart or a line chart, etc.
As shown in FIG. 2, the power spectral density of the acquired brain electrical signalPSD(F)Fitting function of aperiodic composition as shown by curve EL(F)The final fit function is shown as curve FNPS(F)As shown in curve G, where the three peaks of curve G are the result of a periodic component gaussian fit. As can be seen from FIG. 2, the central frequencies of the three peaks are 15Hz, 50Hz and 100Hz, respectively, and the energy height at the frequency of 50Hz is higher, which indicates that the electroencephalogram signal is subjected to 50Hz power frequency interference.
As shown in fig. 3, taking SVM classification as an example, the SVM classification model adopted in this embodiment is a trained SVM classification model (for example, by collecting electroencephalogram data for 20 minutes for 15 subjects, each subject has (15 × 1200) × 6 feature parameter data, and training the model by using a soft interval SVM algorithm according to a five-fold cross-validation method, and then using the model for online evaluation of electroencephalogram signal quality), the quality of an electroencephalogram signal can be identified, for example, output result 0 indicates that the electroencephalogram signal quality is very good, output result 1 indicates that the electroencephalogram signal quality is fair, and output result 2 indicates that the electroencephalogram signal quality is poor. As shown in FIG. 3, FIG. 3(a) shows the EEG signal in a certain time window, which is about 1600 seconds, from which it can be found that the EEG signal is relatively stable for most of the time, but fluctuates at some time. Fig. 3(b) is a real label of electroencephalogram signals labeled by experts, wherein 0 represents that the electroencephalogram signal quality is very good, 1 represents that the electroencephalogram signal quality is still good, and 2 represents that the electroencephalogram signal quality is poor. FIG. 3(c) is a detection result obtained by the SVM classification model according to a plurality of characteristic parameters, wherein 0 represents that the quality of the electroencephalogram signal is very good, 1 represents that the quality of the electroencephalogram signal is good, and 2 represents that the quality of the electroencephalogram signal is poor. As can be seen by comparing the graph in FIG. 3(b) with the graph in FIG. 3(c), the detection result of the method is basically consistent with the real label marked by the expert, and the method can effectively and rapidly detect the quality of the electroencephalogram signal.
The feature map is obtained by taking a radar map as an example, and when the radar map is drawn, the selected feature parameters can be a first feature parameter group and a second feature parameter groupThe radar map can be a regular polygon (such as a regular quadrangle, a regular pentagon, a regular hexagon and the like), a nearly circular shape or a circular shape in various forms, and the distribution sequence of variables of the radar map is not unique. The six characteristic parameters (x, R) are shown below2ω, a, b, DC) drawing a radar map as an example, and the quantitative evaluation through the feature map specifically includes: firstly, six characteristic parameters x, R are determined2Numerical ranges of ω, a, b, DC; drawing a radar map according to the numerical range, and adjusting the axis scale of each characteristic parameter to enable the outer boundary of the radar map to form a regular hexagon; determining an area W which represents good quality of the electroencephalogram signal on a radar map; determining the position points of the measured numerical value of each characteristic parameter in the radar map, and connecting the position points into a closed area; if the closed area is completely inside the area W, the quality of the electroencephalogram signal is considered to be good; if the closed area is not completely inside the area W, the quality of the electroencephalogram signal is considered to be poor.
At the time of quantization processing, the characteristic parameter R2ω, a, b and DC are all measured raw values, and x is a positive-distribution transform (e.g., a mapping obtained by 3sigma principle) of the raw values. In this embodiment, the numerical ranges of the six characteristic parameters are respectively: x is more than or equal to 0 and less than or equal to 5, and R is more than or equal to 0.52The method is characterized in that omega is not less than 1, omega is not less than 0 and not more than 3, a is not less than 1 and not more than 6, b is not less than 0 and not more than 5, and DC is not less than 6 and not more than 11, the numerical range is obtained by selecting 15 testers, collecting 1200 groups of scalp electroencephalogram data of each tester, and respectively carrying out Gaussian probability distribution evaluation analysis on 2 ten thousand groups of collected observation data, all conditions of the quality of the scalp electroencephalogram signals can be reflected to a certain extent, it needs to be noted that the numerical ranges of different types of electroencephalogram signals are different, and the electroencephalogram signals can be scalp electroencephalogram signals or intracranial electroencephalogram signals and the like. The radar drawing can adopt tools such as excel or FineReport report software and the like to determine a circle center, draw six axes in a radioactive ray form, and set the scale of each axis, so that the drawn radar drawing forms a regular hexagon, and each axis represents the numerical range of a characteristic parameter (for example, each axis represents the numerical range of a characteristic parameter)As shown in fig. 4). According to the observed values of the characteristic parameters of the obtained large amount of observed data, 95% unilateral/bilateral confidence intervals of the characteristic parameters under Gaussian distribution are calculated, and an area W (shown in FIG. 4) indicating good electroencephalogram signal quality can be determined. Respectively determining the position points of the numerical values of the six characteristic parameters obtained by real-time measurement in the radar map, connecting the six position points into a closed area, and if the closed area is completely inside the area W, considering that the electroencephalogram signal quality is good; if the closed area is not completely inside the area W, the quality of the electroencephalogram signal is considered to be poor. For example, the closed region O in fig. 4 is a range where all six characteristic parameters fall into the region W, which indicates that the quality of the electroencephalogram signal is good; the parameter a of the closed region Q in fig. 5 exceeds the range of the region W, which indicates that the electroencephalogram signal has poor quality and is interfered by 50Hz power frequency. Parameters of the closed region P in fig. 6xAnd R2When the range of the area W is exceeded, the quality of the electroencephalogram signal is poor, and myoelectricity noise exists. Parameters of the closed region J in fig. 7bAndxif the signal exceeds the area W, the quality of the electroencephalogram signal is poor and low-frequency noise exists. Parameters of the closed region T in fig. 8bDC andxwhen the signal exceeds the area W, the quality of the electroencephalogram signal is poor, and the electrode falling interference exists. Parameters of the closed region U in fig. 9bAnd the DC exceeds the area W, which indicates that the quality of the electroencephalogram signal is poor and invalid leads exist.
In summary, the electroencephalogram signal quality detection method based on the spectrum analysis of the invention changes the time domain signal into the frequency domain signal by performing the fast Fourier transform on the acquired electroencephalogram signal, and establishes the power spectral densityPSD(F) (ii) a To power spectral densityPSD(F) The periodicity in the method is divided into and is a periodic component, nonlinear fitting is respectively carried out on the periodic component to obtain a plurality of characteristic parameters, and the characteristic parameters are combined and evaluated to obtain the result of the quality of the electroencephalogram signal. Fitting the aperiodic component and the periodic component of the electroencephalogram signal on a frequency domain by establishing a signal decomposition model, quantizing a noise parameter, and increasing the interpretability of signal quality; and the interference of power frequency to the electroencephalogram signal can be directly measured, and the accuracy of signal detection is further improved. The method analyzes the quality of the electroencephalogram signal based on the frequency domain, can be used for real-time signal quality evaluation, guides a user to perform corresponding noise reduction operation on a low-quality data segment by combining with quantized parameters, and improves the usability of data.
In light of the foregoing description of the preferred embodiment of the present invention, many modifications and variations will be apparent to those skilled in the art without departing from the spirit and scope of the invention. The technical scope of the present invention is not limited to the contents of the specification, and must be determined by the scope of the claims.
Claims (11)
1. A method for detecting the quality of an electroencephalogram signal based on spectral analysis is characterized by comprising the following steps:
s1: acquiring electroencephalogram signals in a time window to obtain M groups of time domain data;
s2: performing fast Fourier transform on the M groups of time domain data to obtain K groups of frequency domain data, and obtaining power spectral density according to the K groups of frequency domain dataPSD(F)；
S3: for the power spectral densityPSD(F)Carrying out nonlinear fitting on the data of the medium-aperiodic components to obtain a fitting functionL (F)Decomposing to obtain the fitting functionL(F)The first set of characteristic parameters of (1);
s4: removing the power spectral densityPSD(F)Obtaining data of periodic components from the data of non-periodic components;
s5: performing nonlinear fitting on the data of the periodic components to obtain a fitting functionG(F) n Decomposing to obtain the fitting functionG(F) n The second set of characteristic parameters of (1);
s6: and performing combined evaluation on the first characteristic parameter group and the second characteristic parameter group to obtain a detection result of the quality of the electroencephalogram signal.
2. The method for detecting brain electrical signal quality based on spectral analysis of claim 1, wherein said method further comprises:
fitting the fitting functionL(F)And said fitting functionG(F) n Overlapping to obtain the final fitting functionNPS(F)=L(F)+ G(F) n Calculating a final fitting functionNPS(F)And the power spectral densityPSD(F)And then, performing combined evaluation on the regression evaluation index parameter, the first characteristic parameter group and the second characteristic parameter group to obtain a detection result of the electroencephalogram signal quality.
3. The method of detecting brain electrical signal quality based on spectral analysis of claim 2, wherein said regression evaluation index parameter includes Pearson's correlation coefficient R2Mean absolute error MAE, root mean square error RMSE, and mean absolute percentage error MAPE.
4. The method of detecting brain electrical signal quality based on spectral analysis of claim 1, wherein said brain electrical signal is raw brain electrical data acquired in a single channel; or the electroencephalogram signal is electroencephalogram data obtained by performing linear and nonlinear combination processing on original electroencephalogram data acquired in a plurality of channels with different spatial distributions.
5. The method of claim 2, wherein the electroencephalogram signal includes at least one of the periodic components, and each of the periodic components is non-linearly fitted to obtain a fitting functionG(F) n Each fitting functionG(F) n A set of said second set of characteristic parameters can be resolved; and when the quality of the electroencephalogram signal is detected, at least one group of the second characteristic parameter group, the first characteristic parameter group and the regression evaluation index parameter is selected each time for combined evaluation.
6. The method for detecting EEG signal quality based on spectrum analysis of claim 1Characterized in that said fitting function
Wherein F represents a frequency variable, b represents an offset, x represents a power law distribution coefficient, and k is an inflection point; and when k and F are zero, obtaining a direct current component parameter DC, wherein the first characteristic parameter group comprises an offset b, a power law distribution coefficient x and the direct current component DC.
7. The method of detecting brain electrical signal quality based on spectral analysis of claim 1, wherein said fitting function
Wherein F denotes a frequency variable, c denotes a center frequency, ω denotes a bandwidth, a denotes an energy level of the center frequency, and the second feature parameter group includes the bandwidth ω, the energy level a of the center frequency, and the center frequency c.
8. The method of detecting brain electrical signal quality based on spectral analysis of claim 3, wherein said Pearson's correlation coefficient
Wherein cov (X, Y) represents the final fitting functionNPS(F)And the power spectral densityPSD(F)The covariance of (a) of (b),
representing the final fitting functionNPS(F)The variance of (a) is determined,
representing power spectral densityPSD(F)The variance of (c).
9. The method of detecting brain electrical signal quality based on spectral analysis of claim 3, wherein said average absolute error
Wherein X represents the final fitting functionNPS(F)And Y represents the power spectral densityPSD(F)。
10. The method for detecting the quality of the electroencephalogram signal based on the spectral analysis as recited in claim 2, wherein the step S6 specifically includes: and evaluating and classifying the first characteristic parameter group, the second characteristic parameter group and the regression evaluation index parameter through a machine learning algorithm to obtain a detection result of the electroencephalogram signal quality.
11. The method of claim 2, wherein the first feature parameter set, the second feature parameter set and the regression evaluation index parameter are combined and quantitatively evaluated through a feature map to obtain the result of detecting the electroencephalogram signal quality.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110613283.7A CN113040788B (en) | 2021-06-02 | 2021-06-02 | Electroencephalogram signal quality detection method based on spectrum analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110613283.7A CN113040788B (en) | 2021-06-02 | 2021-06-02 | Electroencephalogram signal quality detection method based on spectrum analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113040788A true CN113040788A (en) | 2021-06-29 |
| CN113040788B CN113040788B (en) | 2021-09-21 |
Family
ID=76518651
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110613283.7A Active CN113040788B (en) | 2021-06-02 | 2021-06-02 | Electroencephalogram signal quality detection method based on spectrum analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113040788B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113662565A (en) * | 2021-08-09 | 2021-11-19 | 清华大学 | Video playing quality evaluation method and device based on electroencephalogram characteristics |
| CN116401534A (en) * | 2023-06-08 | 2023-07-07 | 中国空气动力研究与发展中心高速空气动力研究所 | Pulse pressure modal component separation method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050256743A1 (en) * | 2004-05-11 | 2005-11-17 | Dale Richard B | Medical imaging-quality assessment and improvement system (QAISys) |
| US20130131465A1 (en) * | 2010-07-26 | 2013-05-23 | Sharp Kabushiki Kaisha | Biomeasurement device, biomeasurement method, control program for a biomeasurement device, and recording medium with said control program recorded thereon |
| CN110916631A (en) * | 2019-12-13 | 2020-03-27 | 东南大学 | Student classroom learning state evaluation system based on wearable physiological signal monitoring |
| CN111012341A (en) * | 2020-01-08 | 2020-04-17 | 东南大学 | Artifact removal and electroencephalogram signal quality evaluation method based on wearable electroencephalogram equipment |
-
2021
- 2021-06-02 CN CN202110613283.7A patent/CN113040788B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050256743A1 (en) * | 2004-05-11 | 2005-11-17 | Dale Richard B | Medical imaging-quality assessment and improvement system (QAISys) |
| US20130131465A1 (en) * | 2010-07-26 | 2013-05-23 | Sharp Kabushiki Kaisha | Biomeasurement device, biomeasurement method, control program for a biomeasurement device, and recording medium with said control program recorded thereon |
| CN110916631A (en) * | 2019-12-13 | 2020-03-27 | 东南大学 | Student classroom learning state evaluation system based on wearable physiological signal monitoring |
| CN111012341A (en) * | 2020-01-08 | 2020-04-17 | 东南大学 | Artifact removal and electroencephalogram signal quality evaluation method based on wearable electroencephalogram equipment |
Non-Patent Citations (1)
| Title |
|---|
| THOMAS DONOGHUE 等: "Parameterizing neural power spectra into periodic and aperiodic components", 《NATURE NEUROSCIENCE》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113662565A (en) * | 2021-08-09 | 2021-11-19 | 清华大学 | Video playing quality evaluation method and device based on electroencephalogram characteristics |
| CN113662565B (en) * | 2021-08-09 | 2022-06-28 | 清华大学 | Video playing quality evaluation method and device based on electroencephalogram characteristics |
| CN116401534A (en) * | 2023-06-08 | 2023-07-07 | 中国空气动力研究与发展中心高速空气动力研究所 | Pulse pressure modal component separation method |
| CN116401534B (en) * | 2023-06-08 | 2023-08-01 | 中国空气动力研究与发展中心高速空气动力研究所 | Pulse pressure modal component separation method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113040788B (en) | 2021-09-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10743809B1 (en) | Systems and methods for seizure prediction and detection | |
| Wilson et al. | Spike detection: a review and comparison of algorithms | |
| AU2019424265B2 (en) | Fatigue Classiffication Method Based on Brain Function Network Constructed via Generalized Consistency and Relevant Vector Machine | |
| Fraser et al. | Automated biosignal quality analysis for electromyography using a one-class support vector machine | |
| Kiymik et al. | Automatic recognition of alertness level by using wavelet transform and artificial neural network | |
| KR101235441B1 (en) | Bio-signals based automatic sleep-stage identification system | |
| Song et al. | Automatic epileptic seizure detection in EEGs based on optimized sample entropy and extreme learning machine | |
| Sabeti et al. | Entropy and complexity measures for EEG signal classification of schizophrenic and control participants | |
| Song et al. | Automatic recognition of epileptic EEG patterns via extreme learning machine and multiresolution feature extraction | |
| US5083571A (en) | Use of brain electrophysiological quantitative data to classify and subtype an individual into diagnostic categories by discriminant and cluster analysis | |
| Dobrowolski et al. | Multiresolution MUAPs decomposition and SVM-based analysis in the classification of neuromuscular disorders | |
| Anuragi et al. | Automated FBSE-EWT based learning framework for detection of epileptic seizures using time-segmented EEG signals | |
| Gruszczyńska et al. | Epilepsy identification based on EEG signal using RQA method | |
| Ghosh et al. | Automatic eyeblink and muscular artifact detection and removal from EEG signals using k-nearest neighbor classifier and long short-term memory networks | |
| CN113040788B (en) | Electroencephalogram signal quality detection method based on spectrum analysis | |
| Kalaivani et al. | Prediction of biomedical signals using deep learning techniques | |
| Alotaibi et al. | Ensemble Machine Learning Based Identification of Pediatric Epilepsy. | |
| Song | A review of developments of EEG-based automatic medical support systems for epilepsy diagnosis and seizure detection | |
| Furui et al. | Non-Gaussianity detection of EEG signals based on a multivariate scale mixture model for diagnosis of epileptic seizures | |
| KR20190111570A (en) | A system of detecting epileptic seizure waveform based on coefficient in multi-frequency bands from electroencephalogram signals, using feature extraction method with probabilistic model and machine learning | |
| CN115429292A (en) | Electroencephalogram signal quality detection device and system based on spectrum analysis | |
| Greene et al. | Classifier models and architectures for EEG-based neonatal seizure detection | |
| Dora et al. | Adaptive SSA based muscle artifact removal from single channel EEG using neural network regressor | |
| Fickling et al. | Good data? The EEG quality index for automated assessment of signal quality | |
| Sharma et al. | Comparative investigation of machine learning algorithms for detection of epileptic seizures |
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 | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20210629 Assignee: Boruikang medical technology (Shanghai) Co.,Ltd. Assignor: NEURACLE TECHNOLOGY (CHANGZHOU) CO.,LTD. Contract record no.: X2022980011241 Denomination of invention: Detection method of EEG signal quality based on spectrum analysis Granted publication date: 20210921 License type: Common License Record date: 20220728 |
|
| EE01 | Entry into force of recordation of patent licensing contract |