CN107468260A - A kind of brain electricity analytical device and analysis method for judging ANIMAL PSYCHE state - Google Patents
A kind of brain electricity analytical device and analysis method for judging ANIMAL PSYCHE state Download PDFInfo
- Publication number
- CN107468260A CN107468260A CN201710949199.6A CN201710949199A CN107468260A CN 107468260 A CN107468260 A CN 107468260A CN 201710949199 A CN201710949199 A CN 201710949199A CN 107468260 A CN107468260 A CN 107468260A
- Authority
- CN
- China
- Prior art keywords
- eeg signals
- data
- eeg
- characteristic value
- purifying
- 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.)
- Pending
Links
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/16—Devices for psychotechnics; Testing reaction times ; Devices for evaluating the psychological state
- A61B5/165—Evaluating the state of mind, e.g. depression, anxiety
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/369—Electroencephalography [EEG]
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/398—Electrooculography [EOG], e.g. detecting nystagmus; Electroretinography [ERG]
-
- 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
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7253—Details of waveform analysis characterised by using transforms
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B2503/00—Evaluating a particular growth phase or type of persons or animals
- A61B2503/40—Animals
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Psychiatry (AREA)
- Veterinary Medicine (AREA)
- Public Health (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Animal Behavior & Ethology (AREA)
- Surgery (AREA)
- Artificial Intelligence (AREA)
- Signal Processing (AREA)
- Physiology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Psychology (AREA)
- Educational Technology (AREA)
- Developmental Disabilities (AREA)
- Hospice & Palliative Care (AREA)
- Child & Adolescent Psychology (AREA)
- Social Psychology (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Ophthalmology & Optometry (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The embodiment of the invention discloses a kind of brain electricity analytical device and analysis method for judging ANIMAL PSYCHE state.The analytical equipment includes brain wave acquisition module, to gather the original EEG signals of animal；Pretreatment module, obtain purifying EEG signals to remove the noise jamming in the original EEG signals；Characteristic extracting module, to the characteristic value in extraction purification EEG signals；Identification module, to carry out analysis identification to the characteristic value according to the disaggregated model built in advance, to obtain the recognition result for representing ANIMAL PSYCHE state.The brain wave of animal is acquired by above-mentioned module, denoising, and carry out the extraction of characteristic value, calculated and identified by the disaggregated model built in advance, draw the analysis result for representing ANIMAL PSYCHE state, so as to accurately catch the psychological condition change of animal, the present invention particularly when applied to working dog, can preferably aid in human work.
Description
Technical field
The present invention relates to brain-computer interface equipment field, more particularly to a kind of brain electricity analytical device for judging ANIMAL PSYCHE state
And analysis method.
Background technology
In brain-computer interface technical field, prior art carries out E.E.G analysis taking human as object designs wearable device.In
The A of state patent document CN 105011932 disclose a kind of fatigue driving eeg monitoring method based on meditation degree and focus, profit
Focus parameter, the meditation degree parameter exported with TGAM chips, with reference to the average value of original EEG signals, to the fatigue of driver
EEG signals are monitored.But focus parameter, meditation degree parameter of the prior art using the output of TGAM chips, only can
E.E.G parameter is calculated with 1 hertz of frequency, and the change of such as psychological condition brain electricity condition is often written in water, it is necessary to more
High temporal resolution can just detect.
In addition, the data SNR after being removed in the prior art to eye electrical noise in primary signal etc. is low, reliably
Property it is poor, and the removal of other eye electrical noises in the prior art relies on principal component analysis and Independent Component Analysis, it is necessary to individually
Eye electricity acquisition electrode of the setting around the eyes, add the complexity of equipment, while the method for both denoisings is both needed to
Duration that will be longer, it is difficult to meet the requirement of real-time.In sorting technique, frequently resulted in using PCA dimensionality reductions thin in signal
Save the loss of information, it is difficult to catch trickle psychological condition change.
Furthermore existing brain electricity analytical equipment or method are typically stopped to obtain the eeg data after denoising, and these are counted
According to the intermediate link as scientific experimentation, i.e. data are often only used as scientific research use, can not brought to social application beneficial to effect
Fruit.
The content of the invention
It is an object of the invention to the original EEG signals denoising of collection can be obtained into the brain electricity number compared with high s/n ratio
According to, and calculated and identified by disaggregated model, general public is apparent from the psychological condition of animal, more preferably
Realize the application of eeg data socially.
In order to achieve the above object, the present invention proposes a kind of brain electricity analytical device for judging ANIMAL PSYCHE state, and it includes：
Brain wave acquisition module, to gather the original EEG signals of animal；
Pretreatment module, obtain purifying EEG signals to remove the noise jamming in the original EEG signals；
Characteristic extracting module, to the characteristic value in extraction purification EEG signals；
Identification module, to carry out analysis identification to the characteristic value according to the disaggregated model built in advance, to obtain generation
The recognition result of table ANIMAL PSYCHE state.
Wherein, the pretreatment module, including：
An electric unit is removed, to remove eye electrical noise to the original EEG signals, obtains initial filter EEG signals；
Wavelet de-noising unit, for carrying out Wavelet Denoising Method to the initial filter EEG signals, obtain correcting EEG signals；
Data segment extraction unit, to select the amendment EEG signals, and amplitude is limited, removal, which exceeds, to be limited
The data segment of tentering value, obtain normal data section；
Equal value cell is removed, to calculate the average of the normal data section, is subtracted with the amendment EEG signals described equal
Value；
Baseline drift unit is removed, to remove baseline drift to subtracting the amendment EEG signals after average；
Filter unit, to be filtered to the amendment EEG signals after removal baseline drift, so as to obtain purifying brain electricity
Signal.
Wherein, it is described to remove an electric unit, including：
Signal intercepts subelement, to intercept the original EEG signals of a specific duration, as master reference signal；
Filtering subunit, to carry out bandpass filtering according to default eye electrical noise frequency range to the master reference signal
To obtain benchmark initial filter EEG signals；
Comparing subunit, to by each data point amplitude of the benchmark initial filter EEG signals and default eye electrical noise amplitude threshold
Value is compared, when a certain data point amplitude less than default eye electrical noise amplitude threshold by becoming equal to default eye electrical noise amplitude
During threshold value, one before setting the data point a length of first test blink period when specific, the spy after the data point is set
Regularly a length of second test blink period, the first test blink period and the second test blink period form first and surveyed
The examination blink cycle；The benchmark initial filter EEG signals in the first test blink cycle are extracted as the first test blink cycle
The first interior signal of blinking；
Computation subunit, described first in all first test blink cycles in the specific duration to be blinked
Signal is averaged, and obtains First view electrical noise average waveform；
Subelement is superimposed, to the original EEG signals and the First view electrical noise average waveform are superimposed, with
The noise jamming in the first test blink cycle is offset, obtains initial filter EEG signals.
Wherein, described device also includes：
Modeling module, preserved to build disaggregated model and provide it to database；
The modeling module includes：
Control unit, to call brain wave acquisition module collection animal caused brain electricity number under different mental state
According to；The eeg data for calling the pretreatment module to gather brain wave acquisition module is selected and noise reduction process, obtains pure
Change eeg data, call the characteristic value in the characteristic extracting module extraction purifying eeg data；
Model training unit, to be instructed using machine learning algorithm to the characteristic value in the purifying eeg data
Practice, build disaggregated model.
Wherein, described device also includes：
Characteristic optimization module, enter to the characteristic value in the purifying eeg data that is obtained to the characteristic extracting module
The cleaning of row characteristic value, characteristic value selection and/or characteristic value conversion, obtain the validity feature value of the purifying eeg data；
Accordingly, the model training unit, to be trained using machine learning algorithm to the validity feature value,
Build disaggregated model.
On the other hand, the present invention also proposes a kind of brain electricity analytical method for judging ANIMAL PSYCHE state, and it includes：
Gather the original EEG signals of animal；
The noise jamming in the original EEG signals is removed to obtain purifying EEG signals；
Characteristic value in extraction purification EEG signals；
Analysis identification is carried out to the characteristic value according to the disaggregated model built in advance, ANIMAL PSYCHE state is represented to obtain
Recognition result.
Wherein, the noise jamming removed in the original EEG signals obtains purifying EEG signals, including：
Eye electrical noise is removed to the original EEG signals, obtains initial filter EEG signals；
Wavelet Denoising Method is carried out to the initial filter EEG signals, obtains correcting EEG signals；
The amendment EEG signals are selected, and limit amplitude, removes beyond the data segment for limiting amplitude, obtains
Normal data section；
The average of the normal data section is calculated, the average is subtracted with the amendment EEG signals；
Baseline drift is removed to subtracting the amendment EEG signals after average；
Amendment EEG signals after removal baseline drift are filtered, so as to obtain purifying EEG signals.
Wherein, it is described that eye electrical noise is removed to the original EEG signals, including：
The original EEG signals of a specific duration are intercepted, as master reference signal；
Bandpass filtering is carried out to obtain benchmark initial filter according to default eye electrical noise frequency range to the master reference signal
EEG signals；
By each data point amplitude of the benchmark initial filter EEG signals compared with default eye electrical noise amplitude threshold, when certain
When one data point amplitude less than default eye electrical noise amplitude threshold by becoming equal to default eye electrical noise amplitude threshold, the number is set
One before strong point a length of first test blink period when specific, set after the data point one it is specific when a length of second test
Blink the period, the first test blink period and the second test blink period formed for the first test blink cycle；Extraction
Benchmark initial filter EEG signals in the first test blink cycle are as the first blink letter in the first test blink cycle
Number；
First signal of blinking in all first test blink cycles in the specific duration is averaged, obtained
To First view electrical noise average waveform；
The original EEG signals and the First view electrical noise average waveform are superimposed, to offset first test
The noise jamming blinked in the cycle, obtains initial filter EEG signals.
Wherein, methods described also includes：
Build disaggregated model and provide it to database and preserved；
The structure disaggregated model simultaneously provides it to database and preserved, including：
Acquisition module gathers animal caused eeg data under different mental state；
The eeg data is selected and noise reduction process, obtain purifying eeg data；
Extract the characteristic value in the purifying eeg data；
The characteristic value in the purifying eeg data is trained using machine learning algorithm, builds disaggregated model.
Wherein, methods described also includes：
Characteristic value cleaning, characteristic value selection and/or characteristic value conversion are carried out to the characteristic value in the purifying eeg data,
Obtain the validity feature value of the purifying eeg data；
It is described that the characteristic value in the purifying eeg data is trained using machine learning algorithm, specially using machine
Device learning algorithm is trained to the validity feature value, builds disaggregated model.
Brief description of the drawings
By reading the detailed description of hereafter preferred embodiment, it is various other the advantages of and benefit it is common for this area
Technical staff will be clear understanding.Accompanying drawing is only used for showing the purpose of preferred embodiment, and is not considered as to the present invention
Limitation.And in whole accompanying drawing, identical part is denoted by the same reference numerals.In the accompanying drawings：
Fig. 1 is a kind of structured flowchart for judging ANIMAL PSYCHE state brain electricity analytical device provided in an embodiment of the present invention；
Fig. 2 is a kind of structural frames for judgement ANIMAL PSYCHE state brain electricity analytical device that another embodiment of the present invention provides
Figure；
Fig. 3 is a kind of flow chart for judging ANIMAL PSYCHE state brain electricity analytical method provided in an embodiment of the present invention；
Fig. 4 is the thin of step S200 in a kind of judgement ANIMAL PSYCHE state brain electricity analytical method provided in an embodiment of the present invention
Split flow figure；
Fig. 5 is a kind of flow chart for judgement ANIMAL PSYCHE state brain electricity analytical method that another embodiment of the present invention provides；
Fig. 6 is that one kind that another embodiment of the present invention provides judges step S500 in ANIMAL PSYCHE state brain electricity analytical method
Subdivision flow chart.
Embodiment
The exemplary embodiment of the disclosure is more fully described below with reference to accompanying drawings.Although the disclosure is shown in accompanying drawing
Exemplary embodiment, it being understood, however, that may be realized in various forms the disclosure without should be by embodiments set forth here
Limited.On the contrary, these embodiments are provided to facilitate a more thoroughly understanding of the present invention, and can be by the scope of the present disclosure
Completely it is communicated to those skilled in the art.
To be fully understood by the purpose of the present invention, feature and effect, now by following specific embodiments, and coordinate accompanying drawing,
The present invention is described in detail as follows：
Term in the present invention：
Original EEG signals：The animal brain telecommunications that the device of the present invention gathers in real time in real-time analyzing animal psychological condition
Number；
Original eeg data：The animal EEG signals that the device of the present invention gathers in real time when establishing disaggregated model.
Fig. 1 is a kind of structured flowchart for judging ANIMAL PSYCHE state brain electricity analytical device provided in an embodiment of the present invention.Such as
Shown in Fig. 1, judgement ANIMAL PSYCHE state brain electricity analytical device of the present invention mainly includes brain wave acquisition module 10, pretreatment
Module 20, characteristic extracting module 30 and identification module 40, wherein：Described brain wave acquisition module 10, to gather animal
Original EEG signals；Described pretreatment module 20, purified to remove the noise jamming in the original EEG signals
EEG signals；Described characteristic extracting module 30, to the characteristic value in extraction purification EEG signals；Described identification module
40, to carry out analysis identification to the characteristic value according to the disaggregated model built in advance, ANIMAL PSYCHE state is represented to obtain
Recognition result.
In the present embodiment, brain wave acquisition module 10 gathers animal by least one electrode being arranged on animal bodies
Brain signal typically there is relatively low signal to noise ratio as original EEG signals, the original EEG signals.
Pretreatment module 20 to reduce or eliminate the noise jamming in original EEG signals and defeated purifying EEG signals, its
Including removing an electric unit, wavelet de-noising unit, data segment extraction unit, removing equal value cell, go baseline drift unit and filtering
Unit, wherein：An electric unit is removed, to remove eye electrical noise to the original EEG signals, obtains initial filter EEG signals.Small echo
Noise reduction unit, for carrying out Wavelet Denoising Method to the initial filter EEG signals, obtain correcting EEG signals.In the present embodiment, small echo
Denoising is completed by three basic steps.First, to carrying out wavelet transformation containing noisy initial filter EEG signals；Secondly,
Wavelet threshold denoising is carried out to the wavelet coefficient that conversion obtains, removes the noise that initial filter electric signal is included；Again, to denoising
Wavelet coefficient after reason carries out wavelet inverse transformation, so as to obtain correcting EEG signals.Wavelet Denoising Method is advantageous in that energy after denoising
Enough it is successfully reserved signal characteristic.Data segment extraction unit, to select the amendment EEG signals, and limit width
Value, remove beyond the data segment for limiting amplitude, obtain normal data section.Wherein, default restriction amplitude is preferably 250mV.
Equal value cell is removed, to calculate the average of the normal data section, the average is subtracted with the amendment EEG signals.Remove baseline
Drift cells, to remove baseline drift to subtracting the amendment EEG signals after average.Filter unit, to removing baseline drift
Amendment EEG signals after shifting are filtered, so as to obtain purifying EEG signals.This implementation controls, and can specifically pass through 1-30Hz
Bandpass filter the amendment eeg data after removing baseline drift is filtered, with the purifying brain electricity number of wave band needed for obtaining
According to.It is intelligible, in actual applications, the setting of device can be filtered according to actual filtering demands, such as：Pass through 1-50Hz's
Bandpass filter etc. obtains its all band, and this embodiment of the present invention is not specifically limited.
It is described to remove an electric unit in the present embodiment, further comprise following subelement：Signal intercepts subelement, to cut
The original EEG signals of a specific duration are taken, as master reference signal.Filtering subunit, to the master reference
Signal carries out bandpass filtering to obtain benchmark initial filter EEG signals according to default eye electrical noise frequency range.Comparing subunit, use
With by each data point amplitude of the benchmark initial filter EEG signals compared with default eye electrical noise amplitude threshold, when a certain data
When point amplitude less than default eye electrical noise amplitude threshold by becoming equal to default eye electrical noise amplitude threshold, set the data point it
It is preceding one it is specific when a length of first test blink period, set after the data point one it is specific when a length of second test blink when
Section, the first test blink period and the second test blink period formed for the first test blink cycle；Extract described
Benchmark initial filter EEG signals in one test blink cycle are as the first signal of blinking in the first test blink cycle.Calculate
Subelement, first signal of blinking in all first test blink cycles in the specific duration to be put down
, First view electrical noise average waveform is obtained.Subelement is superimposed, the original EEG signals and the First view electricity to be made an uproar
Sound average waveform is superimposed, to offset the noise jamming in the first test blink cycle, obtains initial filter EEG signals.
Characteristic extracting module 30 extracts characteristic value from purifying EEG signals, and this feature value can be purifying EEG signals
It is at least one in Energy distribution state, the variable condition of signal amplitude distribution, LzC complexities or AR model coefficients.Wherein, carry
The characteristic value of each effectively EEG signals is taken to may be accomplished by：Calculate the power spectrum in the effectively EEG signals
It is at least one in degree, Wavelet Entropy, Renyi entropys, Tsallis entropys, LzC complexities, AR parameter model coefficients.Wherein, Wavelet Entropy
Reflect to signal spectrum energy each Subspace Distribution orderly or unordered degree.Using a certain wavelet basis function, fixed point
Depth is solved, wavelet decomposition is carried out to signal, energy and the summation of each node is calculated, obtains gross energy and obtain entropy.Renyi
Entropy and Tsallis entropys react the changes in distribution situation of EEG signals sequence amplitude.Brain telecommunications is calculated by different modes
Number entropy, the feature as signal.Complexity and order degree of the LzC complexities from one-dimensional angle reaction time sequence.
In another embodiment of the present invention, described device also includes data storage storehouse, to store disaggregated model.
Identification module 40 is to receive features described above value, and the disaggregated model prestored by database enters to characteristic value
Row analysis identification, so as to obtain the recognition result (class label) for representing ANIMAL PSYCHE state.
In another embodiment of the present invention, described device also includes the result feedback module not shown in accompanying drawing that will identify
The ANIMAL PSYCHE state outcome of module output is presented in a manner of the mankind can recognize that, user is understood the psychological condition of animal.Enter one
Step ground, as a result feedback module may include one of display module, sounding module, can also have the two modules simultaneously, so as to
Prompted in giving user by way of figure and/or sound.
The original EEG signals denoising of collection can be obtained the eeg data compared with high s/n ratio by the embodiment of the present invention, and
Calculated and identified by disaggregated model, the recognition result for representing ANIMAL PSYCHE state is changed into the mode that the mankind can recognize that
Output, makes general public to be apparent from the psychological condition of animal, preferably realizes eeg data socially
Using.
In another embodiment of the present invention, as shown in Fig. 2 described device also includes modeling module 50, described modeling module
50, preserved to build disaggregated model and provide it to database；
Further, the modeling module 50 includes control unit and model training unit, wherein：Described control list
Member, to call brain wave acquisition module collection animal caused eeg data under different mental state；Call described pre-
The eeg data that processing module is gathered to brain wave acquisition module is selected and noise reduction process, obtains purifying eeg data, adjusts
With the characteristic value in the characteristic extracting module extraction purifying eeg data；Described model training unit, to use
Machine learning algorithm is trained to the characteristic value in the purifying eeg data, builds disaggregated model.
In the present embodiment, modeling module 50, pass through the ANIMAL PSYCHE state EEG signals to different cultivars, all ages and classes
Collection and analysis obtain above-mentioned disaggregated model, and the disaggregated model is sent into database and preserved.Pre-processed in the step
The noise reduction process mode of module 20 includes：
Wherein, control unit is chosen in the eeg data for calling the pretreatment module to gather brain wave acquisition module
Choosing and noise reduction process specifically include following steps：
1) eye electrical noise is removed to the original eeg data, obtains initial filter eeg data.Eye electricity is removed in this step to make an uproar
The method of sound includes：First, the original eeg data of a special time length is intercepted, as master reference data；Secondly, to this
Master reference data carry out bandpass filtering so as to obtain benchmark initial filter eeg data according to default eye electrical noise frequency range；Again
It is secondary, by each data point amplitude of benchmark initial filter eeg data compared with default eye electrical noise amplitude threshold, when a certain data
When point amplitude less than default eye electrical noise amplitude threshold by becoming equal to default eye electrical noise amplitude threshold, set the data point it
It is preceding one it is specific when a length of 3rd test blink period, set after the data point one it is specific when a length of 4th test blink when
Section, the 3rd test blink period and the 4th test blink period formed for the second test blink cycle；Described second is extracted to survey
Benchmark initial filter eeg data in the examination blink cycle is as the second blink data in the second test blink cycle；Then, will
The second signal of blinking in all second test blink cycles in the specific duration is averaged, and is obtained second electrical noise and is put down
Equal waveform；Then, it is second electrical noise average waveform of original eeg data and this is superimposed, to balance out for the second blink cycle
The interior noise jamming caused by the blink of animal, so as to obtain initial filter eeg data.
2) Wavelet Denoising Method is carried out to the initial filter eeg data, obtains correcting eeg data.Wavelet Denoising Method is by three
What basic step was completed.First, to carrying out wavelet transformation containing noisy initial filter eeg data；Secondly, conversion is obtained small
Wave system number carries out wavelet threshold denoising, removes the noise that initial filter eeg data is included；Again, to the wavelet systems after denoising
Number carries out wavelet inverse transformation, so as to obtain correcting eeg data.Wavelet Denoising Method can be successfully reserved after being advantageous in that denoising
Signal characteristic.
3) amendment eeg data is selected, and limits amplitude as 250mV, removed beyond the data for limiting amplitude
Section, obtains reference data section.
4) average of calculating benchmark data segment, the average is subtracted with amendment eeg data.
5) baseline drift is removed to subtracting the amendment eeg data after average.Due to the physical characteristic and ring of channel sensor
Null offset caused by the factor of border can influence the accuracy of signal and follow-up data processing, therefore to going the amendment after average
Eeg data carries out drift and handled.
6) the amendment eeg data after removal baseline drift is filtered by 1-30Hz bandpass filter, obtains institute
Need the purifying eeg data of wave band.It is intelligible, in actual applications, setting for device can be filtered according to actual filtering demands
Put, such as：Bandpass filter by 1-50Hz etc. obtains its all band, and this embodiment of the present invention is not specifically limited.
In the present embodiment, model training unit, to realize the training of disaggregated model, bag using following machine learning algorithm
Include：K arest neighbors (kNN, k-NearestNeighbor) sorting algorithm, SVMs (SVM), decision tree, Bayes's classification are calculated
It is at least one in method, artificial neural network, convolutional neural networks, integrated study (Ensemble Learning).Moreover,
In the method for Classification and Identification, the sorting technique used in the embodiment of the present invention is based on big data and more fully uses data message
To build disaggregated model, more accurately different conditions can be classified.Same sorting algorithm has many adjustable ginsengs
Number, can be by parameter optimization mode in the disaggregated model structure stage, and finding makes category of model accuracy highest parameter.Using
Multiple sorting algorithms can obtain multiple result of calculations when calculating characteristic value, by these result of calculations to used point
Class algorithm carries out preferably, so as to select a sorting algorithm as disaggregated model and be stored in database, that is, using sorting algorithm
To define a kind of classifying rules.
In actual applications, the purpose of feature selecting be used for improve learning algorithm, be also feature subset selection, refer to from
N number of feature is selected in existing M feature so that the specific indexes of system optimize, and are selected from primitive character
Most effective feature is an important means for improving learning algorithm performance to reduce the process of data set dimension, and pattern is known
Not middle crucial data prediction step.For a learning algorithm, good learning sample is the key of training pattern.For
This, in the present embodiment, described device also include accompanying drawing not shown in characteristic optimization module, described characteristic optimization module,
Characteristic value cleaning, characteristic value choosing are carried out to the characteristic value in the purifying eeg data that is obtained to the characteristic extracting module
Select and/or characteristic value converts, obtain the validity feature value of the purifying eeg data.
Accordingly, the model training unit, to be trained using machine learning algorithm to the validity feature value,
Build disaggregated model.
In the embodiment of the present invention, feature cleaning, specific bag are carried out to the characteristic value in the obtained purifying eeg data
Include following steps：
Obtain the extreme characteristic value in the characteristic value；
The extreme characteristic value in the characteristic value is deleted, or, calculates the average or intermediate value of the characteristic value, by the feature
Extreme characteristic value in value is updated to the average or intermediate value, or deletes to be measured corresponding to the extreme characteristic value in the characteristic value
EEG signals.
In a specific embodiment, specifically can be by judging extremum with statistical method, and remove.
Specifically, extreme value criterion includes following methods：
1) average plus-minus three times standard deviation；
2) case figure.
Specifically, minimizing technology is as follows：
1) feature is removed；
2) sample is removed；
3) extremum is replaced, such as with average, intermediate value.
In the specific implementation, it is necessary to remove after finding extreme characteristic value by extreme value criterion.Pass through following three kinds of sides
Formula optimizes processing to it：Firstth, the extreme characteristic value found is removed, for example Wavelet Entropy finds extremum, then removes small
Ripple entropy characteristic value；Second, EEG signals sample to be measured corresponding to the extreme characteristic value found is removed；Third, the pole in feature
End value is replaced with average or intermediate value.
It should be noted that it can be realized in actual applications according to real needs using one or more modes therein
The feature of characteristic value is cleared up.
In the embodiment of the present invention, feature selecting, specific bag are carried out to the characteristic value in the obtained purifying eeg data
Include following steps：
According to the correlation between the diversity of each characteristic value and/or characteristic value and desired value, from the characteristic value
Choose the spy that diversity is higher than the second predetermined threshold value higher than the correlation between the first predetermined threshold value and/or characteristic value and desired value
Value indicative, or
The characteristic value is subjected to random combine, various features combination is obtained, target is carried out to each combinations of features respectively
Model training, evaluation index corresponding to each combinations of features is calculated, search evaluation index is higher than the 3rd predetermined threshold value at least one
Individual combinations of features, the characteristic value at least one combinations of features found is chosen from the characteristic value, or
Each characteristic value is trained using default machine learning algorithm, obtains weights system corresponding to each characteristic value
Number, select weight coefficient higher than the 4th predetermined threshold value or the feature of predetermined number is chosen according to the descending order of weight coefficient
Value.
In actual applications, the purpose of feature selecting be used for improve learning algorithm, be also feature subset selection, refer to from
N number of feature is selected in existing M feature so that the specific indexes of system optimize, and are selected from primitive character
Most effective feature is an important means for improving learning algorithm performance to reduce the process of data set dimension, and pattern is known
Not middle crucial data prediction step.For a learning algorithm, good learning sample is the key of training pattern.
Specifically, it is necessary to select significant feature to input engineering after data complete the correlation pretreatment such as denoising
The algorithm and model of habit is trained.As a rule, feature can be selected from the aspect of following two：
1) whether feature dissipates：If a feature does not dissipate, such as variance is close to 0, that is to say, that sample is at this
Difference is there is no in feature, what use differentiation of this feature for sample does not have.
2) correlation of feature and target：This puts more obvious, the high feature with target correlation, should prioritizing selection.
In addition to variance method, other method mostly considers from correlation.
In the embodiment of the present invention, it can respectively be realized by the following method and feature selecting is carried out to obtained characteristic value.
First, according to the correlation between the diversity of each characteristic value and/or characteristic value and desired value, from the feature
Diversity is chosen in value and is higher than the second predetermined threshold value higher than the correlation between the first predetermined threshold value and/or characteristic value and desired value
Characteristic value.It can specifically be realized, i.e., each feature be commented according to diversity or correlation by filtration method (Filter)
Point, the number of given threshold or threshold value to be selected, select feature.
In a specific example, characteristic value A, B, C, D, E are tried to achieve from one section of EEG signals, this five kinds of features are all respective
Include one group of numerical value, calculate the variance or coefficient correlation (such as Pearson's coefficient and mutual information coefficient) of five kinds of features respectively,
The threshold value of variance or coefficient correlation is set as P, the feature more than this threshold range is A, C, D, is B, E less than P, then choose A,
C, the feature after D is used in machine learning.
Second, the characteristic value is subjected to random combine, various features combination is obtained, each combinations of features is carried out respectively
Object module is trained, and calculates evaluation index corresponding to each combinations of features, searches evaluation index higher than the 3rd predetermined threshold value extremely
A few combinations of features, chooses the characteristic value at least one combinations of features found from the characteristic value.It can specifically lead to
Pack (Wrapper) realization is crossed, i.e., according to object function (being typically the object module training that prediction effect scores), every time choosing
Some features are selected, or exclude some features.Wherein, take turns can be carried out using a basic mode type by recursion elimination feature more
Training, after often wheel training, the feature of some weight coefficients is eliminated, then next round training is carried out based on new feature set.By feature
The selection of collection is considered as a search problem, can first prepare the assembled scheme of several feature, then assess, be compared to each other, evaluation
High feature is preferentially selected.This method needs to select a kind of index of assessment models effect, such as Area Under the
Curve (AUC), Mean Absolute Error (MAE) or Mean Squared Error (MSE).
In a specific example, five kinds of features A, B, C, D, E, using preceding to Method for Feature Selection or backward Method for Feature Selection
The combination of feature is carried out, model training is carried out for each combination and is calculated according to evaluation index, accuracy rate is high then
It is selected.
3rd, each characteristic value is trained using default machine learning algorithm, obtained corresponding to each characteristic value
Weight coefficient, select weight coefficient higher than the 4th predetermined threshold value or choose predetermined number according to the descending order of weight coefficient
Characteristic value.It can specifically be realized, i.e., first be carried out using the algorithm and model of some machine learning by embedding inlay technique (Embedded)
Training, obtains the weight coefficient of each feature, feature is selected from big to small according to coefficient.Similar to Filter methods, but it is logical
Training is crossed to determine the quality of feature.
In a specific example, five kinds of features A, B, C, D, E, first instructed using the algorithm and model of machine learning
Practice, obtain the weight coefficient of each feature, feature is selected according to coefficient from big to small, the big feature of weight coefficient is preferentially chosen
Choosing.
In the embodiment of the present invention, eigentransformation, specific bag are carried out to the characteristic value in the obtained purifying eeg data
Include following steps：The characteristic value is returned using linear normalization method, standard deviation Standardization Act or non-linear normalizing method
One change is handled.
In actual applications, further can be by carrying out eigentransformation to characteristic value, by spy after obtaining characteristic value
The number field restriction of sign in certain limit, and then advantageously in improve electroencephalogramsignal signal analyzing precision.
In actual applications, some graders need to calculate the distance between sample (such as Euclidean distance), such as KNN.Such as
One feature codomain scope of fruit is very big, then distance calculates and just depends primarily on this feature, so as to be runed counter to actual conditions
(for example at this moment actual conditions are that the small feature of codomain scope is more important).Therefore.By to extracting in the embodiment of the present invention
The characteristic value come is normalized, so that the optimization of characteristic value is better achieved.Method for normalizing is specific as follows：
1) linear normalization.This method for normalizing is relatively useful in situation of the numeric ratio compared with concentration.This method has individual
Defect, if max and min are unstable, it is easy to so that normalization unstable result so that follow-up using effect is also unstable.
Max and min can be substituted in actual use with experience constant.
2) standard deviation standardizes.Treated data fit standardized normal distribution, i.e. average are 0, standard deviation 1, its
Converting function is：
Wherein, μ is the average of all sample datas, and σ is the standard deviation of all sample datas.
3) non-linear normalizing.It is frequently used in data and breaks up bigger scene, some numerical value is very big, some very littles.It is logical
Some mathematical functions are crossed, original value is mapped.This method includes log, index, tangent etc..Need according to data distribution
Situation, determine the curve of nonlinear function.
It should be noted that it can choose phase therein according to the data characteristicses of obtained characteristic value and fit in actual applications
A kind of mode answered realizes the normalized to characteristic value, to obtain preferably effect.
The following examples are using police dog as data acquisition object, it is intended to illustrate how modeling module builds point based on dog
Class model.
First, all ages and classes of the dog age more than 8 months, the police dog of different cultivars of health are selected.Then, to them
EEG signals carry out timesharing, one by one acquisition process.For example, at least one electrode of brain wave acquisition module is worn on first
On the body of police dog, then to the stimulation (article for such as hearing various smells) in the police a variety of smell of dog, and gather
The police dog caused EEG signals after these smells are smelt.Daily right place is carried out by the analytical equipment of the present invention
Collection, to ensure the stability to police dog institute's gathered data under different conditions.For example, point not homogeneous is not heard to the police dog
Vinegar, sesame oil, drugs, explosive etc..Control unit in the modeling module of analytical equipment of the present invention calls brain wave acquisition mould first
Block gathers eeg data when the police dog smells above-mentioned article, recalls what pretreatment module was gathered to brain wave acquisition module
Eeg data is selected and denoising, is calculated by characteristic extracting module and extracts the purifying eeg data after denoising
Characteristic value, then features described above value is calculated with least one sorting algorithm by identification module, it is police so as to obtain this
At least one result of calculation of dog.EEG signals are carried out to the police dog of other different cultivars all ages and classes with same method again
Collection, purification, calculate, so as to obtain multiple result of calculations.Preferably go out one optimal point according to the result of calculation of all police dogs
Class algorithm is stored into database as disaggregated model.In the embodiment, the data of collection need to reach certain magnitude.
Final disaggregated model can obtain through the above way, can also first pass through the EEG signals to a police dog
Preliminary classification model is acquired, handles and then establishes, then by carrying out brain electricity to a plurality of police dog of same kind all ages and classes
Disaggregated model is corrected in the collection of signal, processing, then carry out to the police dog of different cultivars the collection of EEG signals, processing is entered
One step improves disaggregated model, finally obtains the more comprehensive final classification model of characteristic value.
Fig. 3 is a kind of flow chart for judging ANIMAL PSYCHE state brain electricity analytical method provided in an embodiment of the present invention.Such as Fig. 3
Shown, judgement ANIMAL PSYCHE state brain electricity analytical method provided in an embodiment of the present invention specifically includes following steps：
Step S100, gather the original EEG signals of animal.
By real by brain wave acquisition module to analytical equipment of the invention on monitored animal wears in the present embodiment
When gather animal original EEG signals, the collection duration can be 1 second, to reach the purpose monitored in real time.
Step S200, remove the noise jamming in the original EEG signals and obtain purifying EEG signals.
The processing method of noise jamming is removed in this step, as shown in figure 4, comprising the following steps：
Step S210, eye electrical noise is removed to the original EEG signals, obtains initial filter EEG signals.Removed in this step
The method of eye electrical noise includes：The original EEG signals of a special time length are intercepted, as master reference signal；It is original to this
Reference signal carries out bandpass filtering so as to obtain benchmark initial filter EEG signals according to default eye electrical noise frequency range；By the benchmark
Each data point amplitude of initial filter EEG signals compared with default eye electrical noise amplitude threshold, when a certain data point amplitude by less than
When default eye electrical noise amplitude threshold becomes equal to default eye electrical noise amplitude threshold, one before setting the data point it is specific when
A length of first test blink period, set after the data point one it is specific when a length of second test blink period, this first is surveyed
Examination blink period and the second test blink period formed for the first test blink cycle；Extract in the first test blink cycle
Benchmark initial filter EEG signals as this first test blink cycle in first blink data；Will be all in the specific duration
The first signal of blinking in first test blink cycle is averaged, and obtains First view electrical noise average waveform；By original brain electricity
Signal and the First view electrical noise average waveform are superimposed, are caused with balancing out in the first blink cycle due to the blink of animal
Noise jamming, so as to obtain initial filter EEG signals.
Step S220, Wavelet Denoising Method is carried out to the initial filter EEG signals, obtain correcting EEG signals.Wavelet Denoising Method is logical
Cross what three basic steps were completed.First, to carrying out wavelet transformation containing noisy initial filter EEG signals；Secondly, change is got in return
The wavelet coefficient arrived carries out wavelet threshold denoising, removes the noise that initial filter electric signal is included；Again, to small after denoising
Wave system number carries out wavelet inverse transformation, so as to obtain correcting EEG signals.Wavelet Denoising Method can be successfully after being advantageous in that denoising
Stick signal feature.
Step S230, amendment EEG signals are selected, and limit amplitude as 250mV, removal, which exceeds, limits amplitude
Data segment, obtain normal data section.
Step S240, the average of normal data section is calculated, the average is subtracted with amendment EEG signals.
Step S250, baseline drift is removed to subtracting the amendment EEG signals after average.Due to the physics of channel sensor
Null offset caused by characteristic and environmental factor can influence the accuracy of signal and follow-up data processing, therefore to going average
Amendment EEG signals afterwards carry out drift and handled.
Step S260, the amendment EEG signals after removal baseline drift are filtered by 1-30Hz bandpass filter
Ripple, obtain the purifying EEG signals of required wave band.It is intelligible, in actual applications, it can be filtered according to actual filtering demands
The setting of ripple device, such as：Bandpass filter by 1-50Hz etc. obtains its all band, this embodiment of the present invention is not done specifically
Limit.
Step S300, the characteristic value in extraction purification EEG signals；
Step S400, analysis identification is carried out to the characteristic value according to the disaggregated model built in advance, moved with obtaining to represent
The recognition result of thing psychological condition.
In an alternative embodiment of the invention, as shown in figure 5, methods described also includes step S500；
Step S500, build disaggregated model and provide it to database and preserved；
In the present embodiment, as shown in fig. 6, step S500 further comprises the steps：
Step S510, acquisition module collection animal caused eeg data under different mental state, as sample
Data are for training pattern.Wherein, the different mental state refers to animal caused heart under the stimulation in different stimulated source
Reason state.
Step S520, is selected the eeg data and noise reduction process, obtains purifying eeg data；
Step S530, extract the characteristic value in the purifying eeg data；
Step S540, the characteristic value in the purifying eeg data is trained using machine learning algorithm, structure point
Class model.
Wherein, after step S530, methods described also includes：Characteristic value in the purifying eeg data is carried out special
Value indicative cleaning, characteristic value selection and/or characteristic value conversion, obtain the validity feature value of the purifying eeg data；
Further, it is described that the characteristic value in the purifying eeg data is trained using machine learning algorithm, have
Body builds disaggregated model to be trained using machine learning algorithm to the validity feature value.
The following examples are still using police dog as eeg monitoring object, it is intended to illustrate the present invention is how to analyze brain telecommunications
Number and exported in a manner of the mankind can recognize that (explanation of being given for example only property of this example, do not represent police dog smell it is above-mentioned
True reflection during article).
It is connected for precedent, after the device of the present invention is provided with foregoing disaggregated model in advance, by brain wave acquisition module
At least one electrode be worn on the body of police dog of execution task, acquisition module is with certain period of time (such as 1 second) to adopt
Collect chronomere, gather the police dog EEG signals and be sent to pretreatment module as original EEG signals.When the police dog exists
When suspected explosive is smelt in task process, dog produces certain psychological condition in itself, reflects the EEG signals of the psychological condition
Collected module gathers and is sent to pretreatment module in real time.Pretreatment module obtains after carrying out noise reduction to the original EEG signals
Purifying EEG signals simultaneously calculate extraction characteristic value therein, and features described above value is carried out by the disaggregated model stored in database
Calculate and identify and show that recognition result (one kind in class label, i.e. numeral, letter or character string) is sent to police dog instruction and guide
Result feedback module in the hand-held Portable intelligent terminal of member, as a result feedback module the mark is converted to what the mankind can recognize that
Figure (such as admiration mark) is shown by the display screen of the Portable intelligent terminal, knows police dog Administrative Assistant police
Dog is found that suspected explosive.In addition, result feedback module the mark can also be converted to voice (such as " suspicious item " or
" explosive ") police dog Administrative Assistant played to by the loudspeaker of the Portable intelligent terminal.As a result feedback module passes through display
Module or sounding module can be set result output by the user of Portable intelligent terminal according to use occasion.
The analytical equipment of the present invention is particularly suitable for use in working dog, except available for assisting police's drug law enforcement, explosion-proof, it may also be used for
Customs's drug law enforcement, search contraband are assisted, most probably in the work of auxiliary seeing-eye dog.
In general, drug detector dog or bomb-sniffing dog need (such as to hear various drugs or band to dog by very long training stage
Have the material of blast characteristics, barked with instructing and guiding them when smelling these articles) and can just be come into operation by screening, this is with regard to big
It is big to add culture training cost.In addition, drug detector dog or bomb-sniffing dog find drugs or material with blast characteristics need it is logical
Specific action (such as bark or sit down at once) is crossed to notify Administrative Assistant that it is found that suspicious item, and trains canine to do these actions
It is also required to put into substantial amounts of time and efforts, and in some these actions of specific occasions and improper.It is but of the present invention
Judgement ANIMAL PSYCHE state brain electricity analytical device, it becomes possible to preferably solve the above problems.First, the present invention by obtain compared with
The eeg data of high s/n ratio, compared with the disaggregated model in the database built in advance, show that working dog is real-time
Psychological condition, and (such as voice or figure) is output to intelligent terminal (such as hand in a manner of the mankind can recognize that by these states
Machine, tablet personal computer, other portable terminals or remote computer), the user of intelligent terminal is learnt in the very first time
Dog is found that any article.
Further for example,, will be of the present invention when performing the explosion-proof inspection of convention by taking bomb-sniffing dog as an example
ANIMAL PSYCHE state brain electricity analytical device is worn on dog, by Administrative Assistant with dog patrol meeting-place, when dog smell certain or certain
When one article is explosive, the intelligent terminal that Administrative Assistant holds shows a figure for representing explosive or with word " blast at once
Thing ", " danger ", " attention " or "！" display, and or represented with the prompt tone compared with amount of bass or vibration intelligent terminal.Instruction is made with this
The person of leading learns that dog is found that situation.
It can be seen that the present invention applies the advantages below with working dog by upper example：
First, by constructing disaggregated model in advance, analytical equipment is by continuing to monitor the EEG signals of the bomb-sniffing dog simultaneously
Comparing is carried out to know that dog is found that suspicious item so as to notify Administrative Assistant.Therefore, the bomb-sniffing dog need not move through it is very long before
Phase trains, and any health, smell is good, dog that is energetic and concentrating only can serve as bomb-sniffing dog use.This
Greatly reduce the spending of police and the input of manpower.
Secondly, in convention occasion, if dog, which is absorbed in certain or a certain thing and bark, easily causes public concern,
So as to the scene to cause confusion.But the analytical equipment of the present invention can select to remind instruction and guide in a manner of figure or text importing
Member, therefore, need not bark when dog is found that suspected explosive.This just eliminates danger for the police are safe and orderly, evacuates the masses
Provide good guarantee.
Again, seeing-eye dog runs into stair when leading owner to walk and will stopped automatically, and the body that body blocks owner is crossed in side
Body, and if at that time being not in good state of owner, easily cause danger when perception is poor.By point that the present invention is worn to seeing-eye dog
Analysis apparatus, the psychological condition of seeing-eye dog is continued to monitor, when seeing-eye dog is in " it was found that stair " corresponding psychological condition, owner
The intelligent terminal of carrying can send the prompt tones such as " attention underfooting " or buzzer, prompt user to pay attention to, so as to aid in
The work of seeing-eye dog, ensure the personal safety of blind person.
In addition, the present invention is not only only limited to working dog use, other animals are also applied for.For example, pet cat or pet
The present invention is worn on the pet of oneself by the owner of pig, it is possible to their psychological condition is understood in real time, so as to add
Raise the enjoyment of pet.For example, after pet cat is worn the analytical equipment of the present invention, the analysis of present invention when it is hungry
The EEG signals of pet cat just can in real time be gathered and carry out filter and made an uproar processing by device, then extract characteristic value and with classification mould
Type is calculated and identified, the recognition result (class label) for representing " being hungry " is sent into result feedback module, as a result fed back
Module by voice or figure or the two all have in a manner of prompt the owner of pet cat that oneself cat is hungry now, this just makes owner couple
The pet of oneself becomes more apparent upon, and also adds the enjoyment of interaction.
It is to be understood that the present invention is described by some embodiments, those skilled in the art are not departing from this
In the case of the spirit and scope of invention, various changes or equivalence replacement can be carried out to these features and embodiment.In addition,
Under the teachings of the present invention, these features and embodiment can be modified with adapt to particular situation and material without departing from
The spirit and scope of the present invention.Therefore, the present invention is not limited to the particular embodiment disclosed, and falls with the application
Right in embodiment belong to protection scope of the present invention.
Claims (10)
- A kind of 1. brain electricity analytical device for judging ANIMAL PSYCHE state, it is characterised in that including：Brain wave acquisition module, to gather the original EEG signals of animal；Pretreatment module, obtain purifying EEG signals to remove the noise jamming in the original EEG signals；Characteristic extracting module, to the characteristic value in extraction purification EEG signals；Identification module, to carry out analysis identification to the characteristic value according to the disaggregated model built in advance, moved with obtaining to represent The recognition result of thing psychological condition.
- 2. device as claimed in claim 1, it is characterised in that the pretreatment module, including：An electric unit is removed, to remove eye electrical noise to the original EEG signals, obtains initial filter EEG signals；Wavelet de-noising unit, for carrying out Wavelet Denoising Method to the initial filter EEG signals, obtain correcting EEG signals；Data segment extraction unit, to select the amendment EEG signals, and amplitude is limited, removal, which exceeds, limits width The data segment of value, obtain normal data section；Equal value cell is removed, to calculate the average of the normal data section, the average is subtracted with the amendment EEG signals；Baseline drift unit is removed, to remove baseline drift to subtracting the amendment EEG signals after average；Filter unit, to be filtered to the amendment EEG signals after removal baseline drift, so as to obtain purifying EEG signals.
- 3. device as claimed in claim 2, it is characterised in that it is described to remove an electric unit, including：Signal intercepts subelement, to intercept the original EEG signals of a specific duration, as master reference signal；Filtering subunit, to carry out bandpass filtering according to default eye electrical noise frequency range to obtain to the master reference signal To benchmark initial filter EEG signals；Comparing subunit, each data point amplitude of the benchmark initial filter EEG signals to be entered with default eye electrical noise amplitude threshold Row compares, when a certain data point amplitude less than default eye electrical noise amplitude threshold by becoming equal to default eye electrical noise amplitude threshold When, one before setting the data point a length of first test blink period when specific, set after the data point one it is specific when A length of second test blink period, the first test blink period and the second test blink period form the first test and blinked The eye circumference phase；The benchmark initial filter EEG signals in the first test blink cycle are extracted as in the first test blink cycle First signal of blinking；Computation subunit, to first signal of blinking that all first tests in the specific duration were blinked in the cycles It is averaged, obtains First view electrical noise average waveform；Subelement is superimposed, to the original EEG signals and the First view electrical noise average waveform are superimposed, to offset Noise jamming in the first test blink cycle, obtains initial filter EEG signals.
- 4. the device as described in claim any one of 1-3, it is characterised in that also include：Modeling module, preserved to build disaggregated model and provide it to database；The modeling module includes：Control unit, to call brain wave acquisition module collection animal caused eeg data under different mental state； The eeg data for calling the pretreatment module to gather brain wave acquisition module is selected and noise reduction process, obtains purifying brain Electric data, call the characteristic value in the characteristic extracting module extraction purifying eeg data；Model training unit, to be trained using machine learning algorithm to the characteristic value in the purifying eeg data, structure Build disaggregated model.
- 5. device as claimed in claim 4, it is characterised in that also include：Characteristic optimization module, carried out to the characteristic value in the purifying eeg data that is obtained to the characteristic extracting module special Value indicative cleaning, characteristic value selection and/or characteristic value conversion, obtain the validity feature value of the purifying eeg data；Accordingly, the model training unit, to be trained using machine learning algorithm to the validity feature value, build Disaggregated model.
- A kind of 6. brain electricity analytical method for judging ANIMAL PSYCHE state, it is characterised in that including：Gather the original EEG signals of animal；The noise jamming in the original EEG signals is removed to obtain purifying EEG signals；Characteristic value in extraction purification EEG signals；Analysis identification is carried out to the characteristic value according to the disaggregated model built in advance, to obtain the knowledge for representing ANIMAL PSYCHE state Other result.
- 7. method as claimed in claim 6, it is characterised in that the noise jamming removed in the original EEG signals obtains To purifying EEG signals, including：Eye electrical noise is removed to the original EEG signals, obtains initial filter EEG signals；Wavelet Denoising Method is carried out to the initial filter EEG signals, obtains correcting EEG signals；The amendment EEG signals are selected, and limit amplitude, removes beyond the data segment for limiting amplitude, obtains standard Data segment；The average of the normal data section is calculated, the average is subtracted with the amendment EEG signals；Baseline drift is removed to subtracting the amendment EEG signals after average；Amendment EEG signals after removal baseline drift are filtered, so as to obtain purifying EEG signals.
- 8. method as claimed in claim 7, it is characterised in that described that eye electrical noise, bag are removed to the original EEG signals Include：The original EEG signals of a specific duration are intercepted, as master reference signal；Bandpass filtering is carried out to obtain benchmark initial filter brain electricity according to default eye electrical noise frequency range to the master reference signal Signal；By each data point amplitude of the benchmark initial filter EEG signals compared with default eye electrical noise amplitude threshold, when a certain number When strong point amplitude less than default eye electrical noise amplitude threshold by becoming equal to default eye electrical noise amplitude threshold, the data point is set Before one it is specific when a length of first test blink period, set after the data point one it is specific when a length of second test blink Period, the first test blink period and the second test blink period formed for the first test blink cycle；Described in extraction Benchmark initial filter EEG signals in first test blink cycle are as the first signal of blinking in the first test blink cycle；First signal of blinking in all first tests blink cycles in the specific duration are averaged, obtain the Electrical noise average waveform at a glance；The original EEG signals and the First view electrical noise average waveform are superimposed, blinked with offsetting first test Noise jamming in cycle, obtain initial filter EEG signals.
- 9. the method as described in claim any one of 6-8, it is characterised in that methods described also includes：Build disaggregated model and provide it to database and preserved；The structure disaggregated model simultaneously provides it to database and preserved, including：Acquisition module gathers animal caused eeg data under different mental state；The eeg data is selected and noise reduction process, obtain purifying eeg data；Extract the characteristic value in the purifying eeg data；The characteristic value in the purifying eeg data is trained using machine learning algorithm, builds disaggregated model.
- 10. method as claimed in claim 9, it is characterised in that methods described also includes：Characteristic value cleaning, characteristic value selection and/or characteristic value conversion are carried out to the characteristic value in the purifying eeg data, obtained The validity feature value of the purifying eeg data；It is described that the characteristic value in the purifying eeg data is trained using machine learning algorithm, specially using engineering Practise algorithm to be trained the validity feature value, build disaggregated model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710949199.6A CN107468260A (en) | 2017-10-12 | 2017-10-12 | A kind of brain electricity analytical device and analysis method for judging ANIMAL PSYCHE state |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710949199.6A CN107468260A (en) | 2017-10-12 | 2017-10-12 | A kind of brain electricity analytical device and analysis method for judging ANIMAL PSYCHE state |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107468260A true CN107468260A (en) | 2017-12-15 |
Family
ID=60606327
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710949199.6A Pending CN107468260A (en) | 2017-10-12 | 2017-10-12 | A kind of brain electricity analytical device and analysis method for judging ANIMAL PSYCHE state |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107468260A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109325402A (en) * | 2018-08-06 | 2019-02-12 | 高维度（深圳）生物信息智能应用有限公司 | A kind of signal processing method, system and computer storage medium |
| CN109614570A (en) * | 2018-11-15 | 2019-04-12 | 北京英视睿达科技有限公司 | Predict the method and device of section water quality parameter data |
| CN110210559A (en) * | 2019-05-31 | 2019-09-06 | 北京小米移动软件有限公司 | Object screening technique and device, storage medium |
| CN110623680A (en) * | 2019-08-30 | 2019-12-31 | 中国民用航空飞行学院 | Method for testing psychological health of civil aviation flight trainees |
| CN111274825A (en) * | 2020-01-16 | 2020-06-12 | 北京电子工程总体研究所 | Brain-computer pet language translation device and use method |
| CN111728607A (en) * | 2020-06-23 | 2020-10-02 | 北京脑陆科技有限公司 | Eye noise removing method based on brain wave signal characteristics |
| CN112545535A (en) * | 2020-12-07 | 2021-03-26 | 杭州沃维医疗科技有限公司 | Sleep-wake cycle analysis method based on amplitude integrated electroencephalogram |
| CN114081504A (en) * | 2021-11-23 | 2022-02-25 | 青岛理工大学 | Driving intention identification method and system based on electroencephalogram signals |
| CN117942076A (en) * | 2024-01-18 | 2024-04-30 | 好心情健康产业集团有限公司 | Psychological state identification method and device based on single-conduction electroencephalogram signal |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080002775A1 (en) * | 2006-06-30 | 2008-01-03 | Ricci Carlos A | Signal analysis employing matched wavelet |
| CN101690659A (en) * | 2009-09-29 | 2010-04-07 | 华东理工大学 | Brain wave analysis method |
| CN102178514A (en) * | 2011-05-09 | 2011-09-14 | 浙江大学 | Coma degree evaluating method based on multiple indexes of non-linearity and complexity |
| CN103019382A (en) * | 2012-12-17 | 2013-04-03 | 北京大学 | Brain-computer interaction method for reflecting subjective motive signals of brain through induced potentials |
| CN103054549A (en) * | 2012-12-29 | 2013-04-24 | 西安交通大学 | Wearable portable device and method for analyzing eye movement |
| CN103761424A (en) * | 2013-12-31 | 2014-04-30 | 杭州电子科技大学 | Electromyography signal noise reducing and aliasing removing method based on second-generation wavelets and ICA (independent component analysis) |
| CN103885445A (en) * | 2014-03-20 | 2014-06-25 | 浙江大学 | Brain-controlling animal robot system and brain-controlling method of animal robot |
| CN103933739A (en) * | 2013-01-19 | 2014-07-23 | 袁囡囡 | Intelligent mind-controlling toy car |
| CN105054928A (en) * | 2015-07-17 | 2015-11-18 | 张洪振 | Emotion display equipment based on BCI (brain-computer interface) device electroencephalogram acquisition and analysis |
| CN105942974A (en) * | 2016-04-14 | 2016-09-21 | 禅客科技（上海）有限公司 | Sleep analysis method and system based on low frequency electroencephalogram |
| CN106529476A (en) * | 2016-11-11 | 2017-03-22 | 重庆邮电大学 | Deep stack network-based electroencephalogram signal feature extraction and classification method |
-
2017
- 2017-10-12 CN CN201710949199.6A patent/CN107468260A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080002775A1 (en) * | 2006-06-30 | 2008-01-03 | Ricci Carlos A | Signal analysis employing matched wavelet |
| CN101690659A (en) * | 2009-09-29 | 2010-04-07 | 华东理工大学 | Brain wave analysis method |
| CN102178514A (en) * | 2011-05-09 | 2011-09-14 | 浙江大学 | Coma degree evaluating method based on multiple indexes of non-linearity and complexity |
| CN103019382A (en) * | 2012-12-17 | 2013-04-03 | 北京大学 | Brain-computer interaction method for reflecting subjective motive signals of brain through induced potentials |
| CN103054549A (en) * | 2012-12-29 | 2013-04-24 | 西安交通大学 | Wearable portable device and method for analyzing eye movement |
| CN103933739A (en) * | 2013-01-19 | 2014-07-23 | 袁囡囡 | Intelligent mind-controlling toy car |
| CN103761424A (en) * | 2013-12-31 | 2014-04-30 | 杭州电子科技大学 | Electromyography signal noise reducing and aliasing removing method based on second-generation wavelets and ICA (independent component analysis) |
| CN103885445A (en) * | 2014-03-20 | 2014-06-25 | 浙江大学 | Brain-controlling animal robot system and brain-controlling method of animal robot |
| CN105054928A (en) * | 2015-07-17 | 2015-11-18 | 张洪振 | Emotion display equipment based on BCI (brain-computer interface) device electroencephalogram acquisition and analysis |
| CN105942974A (en) * | 2016-04-14 | 2016-09-21 | 禅客科技（上海）有限公司 | Sleep analysis method and system based on low frequency electroencephalogram |
| CN106529476A (en) * | 2016-11-11 | 2017-03-22 | 重庆邮电大学 | Deep stack network-based electroencephalogram signal feature extraction and classification method |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109325402A (en) * | 2018-08-06 | 2019-02-12 | 高维度（深圳）生物信息智能应用有限公司 | A kind of signal processing method, system and computer storage medium |
| CN109325402B (en) * | 2018-08-06 | 2022-04-15 | 高维度（深圳）生物信息智能应用有限公司 | Signal processing method, system and computer storage medium |
| CN109614570A (en) * | 2018-11-15 | 2019-04-12 | 北京英视睿达科技有限公司 | Predict the method and device of section water quality parameter data |
| CN109614570B (en) * | 2018-11-15 | 2023-04-18 | 北京英视睿达科技股份有限公司 | Method and device for predicting section water quality parameter data |
| CN110210559B (en) * | 2019-05-31 | 2021-10-08 | 北京小米移动软件有限公司 | Object screening method and device and storage medium |
| CN110210559A (en) * | 2019-05-31 | 2019-09-06 | 北京小米移动软件有限公司 | Object screening technique and device, storage medium |
| CN110623680A (en) * | 2019-08-30 | 2019-12-31 | 中国民用航空飞行学院 | Method for testing psychological health of civil aviation flight trainees |
| CN111274825A (en) * | 2020-01-16 | 2020-06-12 | 北京电子工程总体研究所 | Brain-computer pet language translation device and use method |
| CN111728607A (en) * | 2020-06-23 | 2020-10-02 | 北京脑陆科技有限公司 | Eye noise removing method based on brain wave signal characteristics |
| CN112545535A (en) * | 2020-12-07 | 2021-03-26 | 杭州沃维医疗科技有限公司 | Sleep-wake cycle analysis method based on amplitude integrated electroencephalogram |
| CN114081504A (en) * | 2021-11-23 | 2022-02-25 | 青岛理工大学 | Driving intention identification method and system based on electroencephalogram signals |
| CN114081504B (en) * | 2021-11-23 | 2024-03-01 | 青岛理工大学 | Driving intention recognition method and system based on electroencephalogram signals |
| CN117942076A (en) * | 2024-01-18 | 2024-04-30 | 好心情健康产业集团有限公司 | Psychological state identification method and device based on single-conduction electroencephalogram signal |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107468260A (en) | A kind of brain electricity analytical device and analysis method for judging ANIMAL PSYCHE state | |
| CN107518894A (en) | A kind of construction method and device of animal brain electricity disaggregated model | |
| CN109389059B (en) | P300 detection method based on CNN-LSTM network | |
| CN110070105B (en) | Electroencephalogram emotion recognition method and system based on meta-learning example rapid screening | |
| CN108376220A (en) | A kind of malice sample program sorting technique and system based on deep learning | |
| CN107066514A (en) | The Emotion identification method and system of the elderly | |
| CN107714038A (en) | The feature extracting method and device of a kind of EEG signals | |
| CN108567418A (en) | A kind of pulse signal inferior health detection method and detecting system based on PCANet | |
| CN107669266A (en) | A kind of animal brain electricity analytical system | |
| CN107704881A (en) | A kind of data visualization processing method and processing device based on animal electroencephalogramrecognition recognition | |
| Esfahanian et al. | On contour-based classification of dolphin whistles by type | |
| Arefi et al. | Development of an expert system based on wavelet transform and artificial neural networks for the ripe tomato harvesting robot | |
| An et al. | Electroencephalogram emotion recognition based on 3D feature fusion and convolutional autoencoder | |
| CN114027846B (en) | Intelligent electrocardiosignal processing method | |
| CN112438741B (en) | Driving state detection method and system based on electroencephalogram feature transfer learning | |
| CN111259949A (en) | Fault identification model construction method, model and identification method for aircraft environmental control system | |
| CN113069117A (en) | Electroencephalogram emotion recognition method and system based on time convolution neural network | |
| CN113076878A (en) | Physique identification method based on attention mechanism convolution network structure | |
| Anders et al. | Compensating class imbalance for acoustic chimpanzee detection with convolutional recurrent neural networks | |
| CN113974655A (en) | Epileptic seizure prediction method based on electroencephalogram signals | |
| CN113207511A (en) | Pesticide application method and system based on pesticide resistance monitoring and readable storage medium | |
| CN117877734A (en) | Mental disorder patient risk level judging system | |
| Karthiga et al. | A Deep Learning Approach to classify the Honeybee Species and health identification | |
| Ibrahim et al. | Palm leaf nutrient deficiency detection using convolutional neural network (CNN) | |
| Ingole et al. | Enhancing Plant Health through Deep Neural Network based Leaf Disease Detection |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171215 |
|
| RJ01 | Rejection of invention patent application after publication |