CN106913351A - A kind of recognition methods of Mental Workload level - Google Patents

A kind of recognition methods of Mental Workload level Download PDF

Info

Publication number
CN106913351A
CN106913351A CN201511005142.8A CN201511005142A CN106913351A CN 106913351 A CN106913351 A CN 106913351A CN 201511005142 A CN201511005142 A CN 201511005142A CN 106913351 A CN106913351 A CN 106913351A
Authority
CN
China
Prior art keywords
eeg signals
mental workload
parameter
mental
amplitude
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
Application number
CN201511005142.8A
Other languages
Chinese (zh)
Inventor
郭孜政
张骏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Southwest Jiaotong University
Original Assignee
Southwest Jiaotong University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Southwest Jiaotong University filed Critical Southwest Jiaotong University
Priority to CN201511005142.8A priority Critical patent/CN106913351A/en
Publication of CN106913351A publication Critical patent/CN106913351A/en
Pending legal-status Critical Current

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
    • A61B5/18—Devices for psychotechnics; Testing reaction times ; Devices for evaluating the psychological state for vehicle drivers or machine operators
    • 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
    • A61B5/168—Evaluating attention deficit, hyperactivity
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—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
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7271—Specific aspects of physiological measurement analysis
    • 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
    • A61B2503/20—Workers
    • A61B2503/22—Motor vehicles operators, e.g. drivers, pilots, captains

Abstract

The invention discloses a kind of recognition methods of Mental Workload level.The method includes：EEG signals are gathered by multiple electrodes, reaction time parameter corresponding with the EEG signals is recorded；EEG signals parameter is extracted from the EEG signals for collecting；The EEG signals parameter of predetermined number is chosen from resulting EEG signals parameter as driving Mental Workload characteristic index；Based on the driving Mental Workload characteristic index, SVM identification models are set up；According to the SVM identification models, Mental Workload level is identified.By using the recognition methods of Mental Workload level provided by the present invention, it is possible to achieve the dynamic realtime identification of the Mental Workload level of driver.

Description

A kind of recognition methods of Mental Workload level
Technical field
The present invention relates to stress detection technique, more particularly to a kind of recognition methods of Mental Workload level.
Background technology
Inter-vehicle information system (for example, the system such as GPS navigation, real-time communication, vehicle-mounted audio frequency and video) it is universal Use, and the complexity of traffic control information increases so that the Mental Workload amount of driver increases.Brain high Power load situation, extends, periphery traffic events discrimination when driver will be caused to react emergent traffic incident Reduce, so as to cause operation to be slipped up, influence driving safety.Therefore, for the Mental Workload of driver Level gives the critical problem that effectively identification is driving behavior class research.
At present, driver's Mental Workload carried out both at home and abroad widely studied, result of study shows：Vehicle flowrate, The traffic environment such as traffic sign, highway layout factor can cause the increase of driver's Mental Workload amount, and mental The increase of load will cause extension of the driver to the accident reaction time, to periphery random signal, thing Part accurately identifies the decline of rate, and vehicle drive controls the reduction of performance, so as to trigger contingency occurrence probability Increase.Therefore, numerous researchers have carried out research with regard to the Mental Workload assessment method of driver, at present The main Mental Workload assessment method for using for：Based on NASA task loads index and subjective load evaluation skill The subjective assessing method of art, and the job performance mensuration based on main task and subtask behavior performance.
The above method of the prior art provides important evidence for the Mental Workload Study of recognition of driver.But, In various methods in the prior art, subjectivity test and appraisal and behavior performance test and appraisal are with the time with behavior state Hysteresis quality, is difficult to the Mental Workload test and appraisal of real-time driver.EEG signals are lived as direct reaction brain Dynamic neuro-physiological signals, have high correlation, it would therefore be highly desirable to grind with the current state of mind of driver Study carefully and propose a kind of recognition methods of Mental Workload level, so as to realize driver Mental Workload level it is dynamic State Real time identification, for the design and the integrated optimization design of transport information that are driven from dynamic auxiliary provide foundation.
The content of the invention
In view of this, the present invention provides a kind of recognition methods of Mental Workload level, such that it is able to realize driver Mental Workload level dynamic realtime identification.
What technical scheme was specifically realized in：
A kind of recognition methods of Mental Workload level, the method includes：
EEG signals are gathered by multiple electrodes, reaction time parameter corresponding with the EEG signals is recorded；
EEG signals parameter is extracted from the EEG signals for collecting；
The EEG signals parameter that predetermined number is chosen from resulting EEG signals parameter is mental negative as driving Lotus characteristic index；
Based on the driving Mental Workload characteristic index, support vector machines identification model is set up；
According to the SVM identification models, Mental Workload level is identified.
Preferably, the EEG signals parameter of being extracted from the EEG signals for collecting includes：
The EEG signals that will be collected are divided into the analytic unit that multiple time spans are k minutes, to each The EEG signals of analytic unit carry out overall filtering process with the bandwidth of 0~80Hz；
To the EEG signals of each analysis unit, from left to right slided paragraph by paragraph as time window with default step-length, and with The EEG signals of one analytic unit are divided into multiple time window signals by default time window Duplication；
The Hamming window in equal length will be multiplied in each time window signal, obtain intermediate variable f [n]；
Fast Fourier Transform (FFT) is carried out to f [n], amplitude distribution f (k) of the EEG signals in frequency domain is obtained；
Extract the average amplitude of default frequency range respectively from amplitude distribution f (k)；For each analytic unit EEG signals, when time window is slided paragraph by paragraph from front to back, obtain each default respective amplitude sequence of frequency range Row；
For the amplitude sequence of each frequency range, after removing the abnormal data outside positive and negative 3 times of standard deviations, each frequency range is obtained Amplitude arrangement set；
To amplitude arrangement set in each time window amplitude seek amplitude average value, by the amplitude average value make It is an EEG signals for analytic unit in the amplitude typical value of corresponding frequency band, i.e. EEG signals parameter.
Preferably, the value of the k is 1；The default step-length is 2 seconds；The time window Duplication is 50%.
Preferably, amplitude distribution f (k) is：
Wherein, WN=cos (2 π/N)-jsin (2 π/N), n are EEG signals sample size, and N is closest N and more than n 2 values of powers.
Preferably, the default frequency range is δ, θ, α and β frequency range；
The δ frequency ranges are 0.5~4Hz, and the θ frequency ranges are 4~8Hz, and α frequency ranges are 8~13Hz, and β is frequently Section is 13~30Hz.
Preferably, the EEG signals parameter that predetermined number is chosen from resulting EEG signals parameter Include as Mental Workload characteristic index is driven：
For arbitrary EEG signals parameter xj, it is mental with low in Mental Workload driving condition high The argument sequence obtained under load driving condition is mixed, and carries out Kruskal-Wallis inspections：
Wherein, H is test statistics, and i is Mental Workload grade scalar；I=1 is underload, and i=2 is High load capacity；Under representing the i-th type load grade, the average order of parameter sample；M is two class parameter samples Total amount；
All of EEG signals parameter is tested one by one, obtains joining with each EEG signals by tabling look-up The Probability p of the corresponding critical zone of number；
Minimum preceding m (1≤m≤4 of the Probability p of critical zone are chosen from all of EEG signals parameter × q) item EEG signals parameter as drive Mental Workload characteristic index.
Preferably, it is described based on the driving Mental Workload characteristic index, set up SVM identification model bags Include：
If it is x to drive Mental Workload characteristic index sample for t-tht=(xt1,xt2,…,xtm) (wherein T=1,2 ..., D, D are sample size)；And set Mental Workload high and low Mental Workload；
By nonlinear transformation Φ, by xtIt is mapped to high dimension linear space；
If there is optimum linearity interface meets following condition：
w·Φ(xt)+b=0；
Wherein w Φ (xt) it is vector w and Φ (xt) inner product；
Then obtain discriminant function：
It is converted into quadratic programming problem：
Wherein, w ∈ RDIt is weight factor with b ∈ R；ξtIt is relaxation function；CIt is penalty；
Use kernel functionTo replace Φ (xt), wherein,σIt is the width of kernel function Degree and t, l=1,2 ..., D, therefore decision function is written as：
For a driver Mental Workload recognition feature vector sample x=(x for UNKNOWN TYPE1,x2,…,xm), Calculate：
And the type that x belongs to is judged according to the value of y (x).
If preferably, y (x)=1, judges that x belongs to Mental Workload high；
If y (x)=- 1, judge that x belongs to low Mental Workload.
As above it is visible, in the recognition methods of Mental Workload level in the present invention, brain is gathered by multiple electrodes Electric signal, records reaction time parameter corresponding with the EEG signals, is carried from the EEG signals for collecting EEG signals parameter is taken, and chooses the EEG signals parameter of predetermined number as driving Mental Workload characteristic index； SVM identification models are then set up, Mental Workload level is identified, such that it is able to relatively accurately to driving The Mental Workload level of the person of sailing is identified, and the accuracy rate of Model Identification is high, and a step of going forward side by side demonstrates frequency Domain index can not only be identified to the Mental Workload level of cognition experiment class, also can be to being driven under simulated environment The Mental Workload level of the person of sailing effectively is recognized, so as to be built to following automatic DAS (Driver Assistant System), traffic Environment and the design of inter-vehicle information system integrated optimization provide foundation.
Brief description of the drawings
Fig. 1 is the schematic flow sheet of the recognition methods of the Mental Workload level in the embodiment of the present invention.
Fig. 2 is the behavioral data result schematic diagram in the embodiment of the present invention.
Fig. 3 is the SVM model test results schematic diagrames in the embodiment of the present invention.
Fig. 4 is the SVM model ROC curve figures in the embodiment of the present invention.
Specific embodiment
To make the objects, technical solutions and advantages of the present invention become more apparent, develop simultaneously referring to the drawings Embodiment, the present invention is described in more detail.
Present embodiments provide a kind of recognition methods of Mental Workload level.
Fig. 1 is the schematic flow sheet of the recognition methods of the Mental Workload level in the embodiment of the present invention.As schemed Shown in 1, the recognition methods of the Mental Workload level in the embodiment of the present invention mainly includes step as described below Suddenly：
Step 11, EEG signals are gathered by multiple electrodes, record reaction corresponding with the EEG signals Time parameter.
In the inventive solutions, in order to test driver is to the simple reaction time of time burst, Need to test tested personnel (for example, volunteer) accordingly, so as to obtain corresponding brain electricity Signal reaction time parameter corresponding with EEG signals and Subjective fatigue and behavior performance test and appraisal data.
For example, preferably, in the preferred embodiment, multiple tested personnel can be first convened, For example, in a specific experiment of the invention, have chosen 16 tested personnel simulator (for example, Opposite theories of vertical and horizontal alliance of states with the state of Qin as pivot VDS-S-III) on carry out underload high respectively and drive test experiments.Tested personnel's age 24 ± 2 years old, effective driving license is held, driving range is accumulative more than 10000 kilometers.
In above-mentioned specific experiment of the invention, high and low two kinds of conditions are constructed using dual task test normal form Driving Mental Workload grade, wherein main task for drive tracking task, tested personnel are informed before experiment The reasonable value of two following distances, tested person sensitive's appropriate distance is made by exercise.Mission requirements driver Control A cars, reasonable spacing is kept with B cars.Subtask is random signal detection mission, when B cars are prominent So braking, brake lamp lights, it is desirable to which tested personnel are as fast as possible to make brake response, now still needs to Keep reasonable spacing.Task load is controlled by the tracking speed of main task, and high load capacity average speed is 80KM/H, underload average speed is 40KM/H.
In experimentation kind, under the conditions of two kinds of Mental Workloads of record the actual spacing of the car of A, B two and it is reasonable between Away from difference and during the reaction to random probing signal task and accuracy, data recording frequency is 10Hz. Meanwhile, using EEG signals (EEG) record and analysis system of German Brain Products companies, press 32 conductive polar caps of international 10-20 systems extension record the EEG signals of driver, between scalp and electrode Impedance be less than 5k Ω, a width of 0.5-100Hz of record paper, sample rate is 1000Hz.
When can collect EEG signals and reaction corresponding with the EEG signals by above-mentioned mode Between parameter.
Step 12, extracts EEG signals parameter from the EEG signals for collecting.
EEG signals reflect brain activity state, when human brain enters higher load condition by low load condition, Electrical activity of brain frequency range amplitude also accordingly changes, i.e., high frequency band amplitude increases and low-frequency band amplitude drops It is low.Each frequency range amplitude change of electrical activity of brain can reflect the Mental Workload state of driver.Therefore, at this In the technical scheme of invention, frequency range can will be preset in EEG signals, for example, four kinds of signal frequency ranges：δ (0.5~4Hz), θ (4~8Hz), the amplitude of α (8~13Hz) and β (13~30Hz) are used as mental Load identification parameter.
In the inventive solutions, it is possible to use various ways extract EEG signals from EEG signals Parameter.Technical scheme will below be carried out in detail by taking a kind of specific implementation therein as an example Thin introduction.
For example, preferably, in a particular embodiment of the present invention, it is described from the EEG signals for collecting Extracting EEG signals parameter includes：
Step 121, the EEG signals that will be collected are divided into the analysis list that multiple time spans are k minutes Unit, the EEG signals to each analysis unit carry out overall filtering process with the bandwidth of 0~80Hz, to go Except the artifacts such as power frequency electric and part myoelectricity.
If for example, the EEG signals time span (abbreviation duration) for collecting be (m*k) minute, The EEG signals that can then this be collected are divided into m analytic unit, the brain telecommunications of each analysis unit Number when it is a length of k minutes.
In the inventive solutions, the value of the k can according to the need for practical situations and Pre-set.For example, preferably, in a particular embodiment of the present invention, the value of the k is 1.
Step 122, to the EEG signals of each analysis unit, with default step-length as time window from left to right Slide paragraph by paragraph, and the EEG signals of an analytic unit are divided into by multiple with default time window Duplication Time window signal.
In the inventive solutions, the value of the default step-length can be according to practical situations Need and pre-set.For example, preferably, in a particular embodiment of the present invention, the default step-length It is 2 seconds.
In the inventive solutions, the value of the time window Duplication can also be according to practical application Pre-set the need for situation.For example, preferably, in a particular embodiment of the present invention, when described Between window Duplication be 50%.
Thus, for example, when analytic unit a length of 1 minute, presetting step-length for 2000 milliseconds, time When window Duplication is 50%, then the EEG signals of an analytic unit can be divided into 60 time windows Signal, the EEG signals of each time window signal when a length of 2000 milliseconds, and each time window signal it Between Duplication be 50%.
Step 123, will multiply the Hamming window in equal length in each time window signal, obtain intermediate variable f[n]。
In the inventive solutions, in order to eliminate side lobe effect to Fast Fourier Transform (FFT) (FFT) Influence, the interior Hamming window multiplied in equal length of each time window signal can be obtained intermediate variable f[n]。
Step 124, Fast Fourier Transform (FFT) is carried out to f [n], obtains amplitude distribution of the EEG signals in frequency domain f(k)。
For example, preferably, in a particular embodiment of the present invention, amplitude distribution f (k) is：
In formula, WN=cos (2 π/N)-jsin (2 π/N), n are EEG signals sample size, and N is closest N and more than n 2 values of powers.
Step 125, extracts the average amplitude of default frequency range respectively from amplitude distribution f (k)；For each The EEG signals of analytic unit, when time window is slided paragraph by paragraph from front to back, obtain each default frequency range each Amplitude sequence.
In the inventive solutions, the value of the default frequency range can be according to practical situations Need and pre-set.For example, preferably, in a particular embodiment of the present invention, the default frequency range It is δ, θ, α and β frequency range；The δ frequency ranges are 0.5~4Hz, and the θ frequency ranges are 4~8Hz, and α is frequently Section is 8~13Hz, and β frequency ranges are 13~30Hz.
Thus, for example, in the preferred embodiment, when the EEG signals of an analytic unit are When Duplication between the Shi Changwei EEG signals of 1 minute, each time window signal is 50%, can obtain To 4 respective amplitude sequences of frequency range such as above-mentioned δ, θ, α and β.
Step 126, for the amplitude sequence of each frequency range, removes the abnormal data outside positive and negative 3 times of standard deviations Afterwards, the amplitude arrangement set of each frequency range (for example, δ, θ, α and β frequency range) is obtained.
Step 127, to amplitude arrangement set in each time window amplitude seek amplitude average value, by institute State amplitude average value as an EEG signals for analytic unit corresponding frequency band amplitude typical value, i.e. brain Electric signal parameter.
In the inventive solutions, by above-mentioned step 121~127, can be to electric from each The EEG signals that pole collects carry out above-mentioned treatment, so as to the brain telecommunications for obtaining being collected from an electrode Number EEG signals parameter.
If for example, the default frequency range is δ, θ, α and β frequency range, for from each electrode The EEG signals for collecting, can obtain 4 EEG signals parameters (i.e. amplitude typical value).Therefore, such as Fruit has used q electrode collection EEG signals, then can obtain (4 × q) item EEG signals parameter altogether, X can be designated asj(1≤j≤4×q)。
Step 13, chooses the EEG signals parameter conduct of predetermined number from resulting EEG signals parameter Drive Mental Workload characteristic index.
In the inventive solutions, it is possible to use various ways are from resulting EEG signals parameter The EEG signals parameter of predetermined number is chosen as driving Mental Workload characteristic index.Below will be with therein As a example by a kind of specific implementation, technical scheme is described in detail.
For example, preferably, in a particular embodiment of the present invention, it is described to join from resulting EEG signals The EEG signals parameter that predetermined number is chosen in number includes as Mental Workload characteristic index is driven：
Step 131, for arbitrary EEG signals parameter xj, it is driven into shape in Mental Workload high State is mixed with the argument sequence obtained under low Mental Workload driving condition, carries out Kruskal-Wallis Inspection：
Wherein, H is test statistics, and i is Mental Workload grade scalar；I=1 is underload, and i=2 is High load capacity；Under representing the i-th type load grade, the average order of parameter sample；M is two class parameter samples Total amount.
Step 132, tests one by one to all of EEG signals parameter, is obtained and each by tabling look-up The Probability p of the corresponding critical zone of EEG signals parameter.
In the inventive solutions, it is 1 chi square distribution because H obeys the free degree, therefore passes through Table look-up and can obtain the Probability p of critical zone.The p value represents sample point under the conditions of two kinds of Mental Workloads Cloth identical probability, therefore the smaller data distribution difference for representing this parameter under two type load states of p value It is bigger, p<.05 it is significant difference.
Step 133, chooses the minimum preceding m of the Probability p of critical zone from all of EEG signals parameter (1≤m≤4 × q) item EEG signals parameter is used as driving Mental Workload characteristic index.
In the inventive solutions, the above method can be based on, 4 × q electroencephalogram parameter is entered one by one Performing check, preceding m minimum composition of Probability p is then selected from 4 × q electroencephalogram parameter and drives mental Load characteristic index.
In the inventive solutions, the value of the m can according to the need for practical situations and Pre-set.
By above-mentioned step 131~133, you can the driving Mental Workload characteristic index needed for obtaining.
Step 14, based on the driving Mental Workload characteristic index, sets up SVMs (SVM) knowledge Other model.
In the inventive solutions, it is possible to use various ways set up SVM identification models.Below Technical scheme will be described in detail by taking a kind of specific implementation therein as an example.
For example, preferably, in a particular embodiment of the present invention, it is described based on the driving Mental Workload Characteristic index, setting up SVM identification models includes：
Step 141, if it is x to drive Mental Workload characteristic index sample for t-tht=(xt1,xt2,…,xtm) (its Middle t=1,2 ..., D, D are sample size)；And set Mental Workload high and low Mental Workload；
Preferably, in a particular embodiment of the present invention, Mental Workload high and low mental can be pre-set Load.For example, 1 can be preset for Mental Workload high, -1 is low Mental Workload；Now, xtCan It can be the one of which in 1 or -1 two types.
Step 142, by nonlinear transformation Φ, by xtIt is mapped to high dimension linear space.
Step 143, if there is optimum linearity interface meets following condition：
w·Φ(xt)+b=0 (3)
Wherein w Φ (xt) it is vector w and Φ (xt) inner product；
Then obtain discriminant function：
The problem can be converted into quadratic programming problem：
Wherein, w ∈ RDIt is weight factor with b ∈ R；ξtIt is relaxation function；CIt is penalty.
Step 144, uses kernel function(σ is the width of kernel function) (t, l=1,2 ..., D) replaces Φ (xt), therefore decision function is written as：
Step 145, for driver's Mental Workload recognition feature vector sample of UNKNOWN TYPE X=(x1,x2,…,xm), calculate：
And the type that x belongs to is judged according to the value of y (x).
For example, preferably, in a particular embodiment of the present invention, if y (x)=1, judging that x belongs to the 1st Class, i.e., Mental Workload high；If y (x)=- 1, judge that x belongs to the 2nd class, i.e., low Mental Workload.
By above-mentioned step 141~145, you can based on the driving Mental Workload characteristic index, set up SVM identification models.
Step 15, according to the SVM identification models, is identified to Mental Workload level.
In the inventive solutions, after above-mentioned SVM identification models are set up, you can according to institute SVM identification models are stated, Mental Workload level is identified, and obtain corresponding recognition result, from And the Mental Workload level of driver can be identified.
Therefore, by the method described in above-mentioned steps 11~15, you can build the SVM identifications Model, and the Mental Workload level of driver is identified according to the SVM identification models.
In the inventive solutions, in order to verify having for the above-mentioned SVM identification models in the present invention Effect property, it is possible to use the EEG signals obtained by experiment are identified come the Mental Workload level to driver, And above-mentioned recognition result is tested and assessed.
For example, in the inventive solutions, by experimental technique described in such as above-mentioned step 11, Driver Mental Workload recognition feature vector sample of the SVM identification models to certain UNKNOWN TYPE X=(x1,x2,…,xm) identification can produce 4 kinds of results：
(1), input sample is Mental Workload high, and identification is judged as Mental Workload high, is designated as true high load capacity (TH)；
(2), input sample is low Mental Workload, and identification is judged as high load capacity, is designated as pseudo- high load capacity (FH)；
(3), input sample is low Mental Workload, and identification is judged as low Mental Workload, is designated as true underload (TL)；
(4), input sample is high load capacity, and identification is judged as underload, is designated as pseudo- underload (FL).
Recognition effect is tested and assessed it is assumed that following three indexs can be built based on above-mentioned, i.e.,：
Wherein, THR has reacted recognition correct rate of the SVM identifiers to Mental Workload state high；SPC Recognition correct rate of the SVM identifiers to low Mental Workload is reacted；ACC has reacted SVM identifiers pair The overall recognition correct rate of all samples.
In the inventive solutions, three behavioral datas and subjective load test and appraisal data can be carried out Average is calculated with standard deviation, and its result is as shown in Figure 2.
Fig. 2 is the behavioral data result schematic diagram in the embodiment of the present invention.As shown in Fig. 2 to above-mentioned number According to paired-sample t test is carried out, as a result show, NASA subjective scales score is significantly low at low load In high load capacity (82.20vs.92.70, t (15)=- 8.197, p<.01)；Two cars are reasonable under low load condition High-load condition (2.0.2vs.2.7.6, t (15)=- 2.40, p are substantially less than when distance offsets are with reaction <.05；482.5vs.518.6, t (15)=- 11.791, p<.01), accuracy is notable under low loading conditions More than high-load condition (92.26vs.87.37, t (15)=3.05, p<.05).Old friend's above two drives Task is set, and realizes the difference of driver's Mental Workload.
Fig. 3 is the SVM model test results schematic diagrames in the embodiment of the present invention, as shown in figure 3, institute State SVM identification models judgment accuracy preferably, Model Identification accuracy up to 93.8%~96.5%, Average accuracy is 95.2%.
Fig. 4 is the SVM model ROC curve figures in the embodiment of the present invention.As shown in figure 4, described ROC curve figure reflects identification model has preferable stability for the classification of data.
In summary, in the recognition methods of Mental Workload level in the present invention, adopted by multiple electrodes Collection EEG signals, record reaction time parameter corresponding with the EEG signals, from the brain telecommunications for collecting EEG signals parameter is extracted in number, and chooses the EEG signals parameter of predetermined number as driving Mental Workload Characteristic index；SVM identification models are then set up, Mental Workload level is identified, such that it is able to Relatively accurately the Mental Workload level to driver is identified, and the accuracy rate of Model Identification is high, A step of going forward side by side demonstrates frequency-domain index and not only the Mental Workload level of cognition experiment class can be identified, Also the Mental Workload level of driver under simulated environment can effectively be recognized, so as to following automatic auxiliary Control loop is helped to build, traffic environment and the design of inter-vehicle information system integrated optimization provide foundation.
Presently preferred embodiments of the present invention is the foregoing is only, is not intended to limit the invention, it is all at this Within the spirit and principle of invention, any modification, equivalent substitution and improvements done etc. should be included in Within the scope of protection of the invention.

Claims (8)

1. a kind of recognition methods of Mental Workload level, it is characterised in that the method includes：
EEG signals are gathered by multiple electrodes, reaction time parameter corresponding with the EEG signals is recorded；
EEG signals parameter is extracted from the EEG signals for collecting；
The EEG signals parameter that predetermined number is chosen from resulting EEG signals parameter is mental negative as driving Lotus characteristic index；
Based on the driving Mental Workload characteristic index, support vector machines identification model is set up；
According to the SVM identification models, Mental Workload level is identified.
2. method according to claim 1, it is characterised in that described from the EEG signals for collecting Middle extraction EEG signals parameter includes：
The EEG signals that will be collected are divided into the analytic unit that multiple time spans are k minutes, to each The EEG signals of analytic unit carry out overall filtering process with the bandwidth of 0~80Hz；
To the EEG signals of each analysis unit, from left to right slided paragraph by paragraph as time window with default step-length, and with The EEG signals of one analytic unit are divided into multiple time window signals by default time window Duplication；
The Hamming window in equal length will be multiplied in each time window signal, obtain intermediate variable f [n]；
Fast Fourier Transform (FFT) is carried out to f [n], amplitude distribution f (k) of the EEG signals in frequency domain is obtained；
Extract the average amplitude of default frequency range respectively from amplitude distribution f (k)；For each analytic unit EEG signals, when time window is slided paragraph by paragraph from front to back, obtain each default respective amplitude sequence of frequency range Row；
For the amplitude sequence of each frequency range, after removing the abnormal data outside positive and negative 3 times of standard deviations, each frequency range is obtained Amplitude arrangement set；
To amplitude arrangement set in each time window amplitude seek amplitude average value, by the amplitude average value make It is an EEG signals for analytic unit in the amplitude typical value of corresponding frequency band, i.e. EEG signals parameter.
3. method according to claim 2, it is characterised in that：
The value of the k is 1；The default step-length is 2 seconds；The time window Duplication is 50%.
4. method according to claim 2, it is characterised in that amplitude distribution f (k) is：
f &lsqb; k &rsqb; = &Sigma; n = 0 N - 1 f &lsqb; n &rsqb; W N k n , 0 &le; k &le; N - 1 0 , o t h e r ;
Wherein, WN=cos (2 π/N)-j sin (2 π/N), n are EEG signals sample size, and N is closest N and more than n 2 values of powers.
5. method according to claim 2, it is characterised in that：
The default frequency range is δ, θ, α and β frequency range；
The δ frequency ranges are 0.5~4Hz, and the θ frequency ranges are 4~8Hz, and α frequency ranges are 8~13Hz, and β is frequently Section is 13~30Hz.
6. method according to claim 2, it is characterised in that described from resulting EEG signals The EEG signals parameter that predetermined number is chosen in parameter includes as Mental Workload characteristic index is driven：
For arbitrary EEG signals parameter xj, it is mental with low in Mental Workload driving condition high The argument sequence obtained under load driving condition is mixed, and carries out Kruskal-Wallis inspections：
H = 12 M ( M + 1 ) &Sigma; i = 1 2 R &OverBar; i 2 n i - 3 ( M + 1 ) ;
Wherein, H is test statistics, and i is Mental Workload grade scalar；I=1 is underload, and i=2 is High load capacity；Under representing the i-th type load grade, the average order of parameter sample；M is two class parameter samples Total amount；
All of EEG signals parameter is tested one by one, obtains joining with each EEG signals by tabling look-up The Probability p of the corresponding critical zone of number；
Minimum preceding m (1≤m≤4 of the Probability p of critical zone are chosen from all of EEG signals parameter × q) item EEG signals parameter as drive Mental Workload characteristic index.
7. method according to claim 2, it is characterised in that described to drive mental negative based on described Lotus characteristic index, setting up SVM identification models includes：
If it is x to drive Mental Workload characteristic index sample for t-tht=(xt1,xt2,…,xtm) (wherein T=1,2 ..., D, D are sample size)；And set Mental Workload high and low Mental Workload；
By nonlinear transformation Φ, by xtIt is mapped to high dimension linear space；
If there is optimum linearity interface meets following condition：
w·Φ(xt)+b=0；
Wherein w Φ (xt) it is vector w and Φ (xt) inner product；
Then obtain discriminant function：
y ( x t ) = 1 , w &CenterDot; &Phi; ( x t ) + b &GreaterEqual; 1 ; - 1 , w &CenterDot; &Phi; ( x t ) + b &le; - 1 ; ;
It is converted into quadratic programming problem：
m i n 1 2 w 2 + C &Sigma; t = 1 D &xi; t , S . T . y t ( w &CenterDot; &Phi; ( x t ) + b ) &GreaterEqual; 1 - &xi; t , &xi; t &GreaterEqual; 0 , t = 1 , 2 , ... , D ;
Wherein, w ∈ RDIt is weight factor with b ∈ R；ξtIt is relaxation function；C is penalty；
Use kernel functionTo replace Φ (xt), wherein, σ is the width of kernel function Degree and t, l=1,2 ..., D, therefore decision function is written as：
sgn ( &Sigma; t = 1 D y t &alpha; t k ( x t , x l ) + b ) ;
For a driver Mental Workload recognition feature vector sample x=(x for UNKNOWN TYPE1,x2,…,xm), Calculate：
y ( x ) = sgn ( &Sigma; t = 1 D y t &alpha; t k ( x t , x l ) + b ) ;
And the type that x belongs to is judged according to the value of y (x).
8. method according to claim 7, it is characterised in that：
If y (x)=1, judge that x belongs to Mental Workload high；
If y (x)=- 1, judge that x belongs to low Mental Workload.
CN201511005142.8A 2015-12-28 2015-12-28 A kind of recognition methods of Mental Workload level Pending CN106913351A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201511005142.8A CN106913351A (en) 2015-12-28 2015-12-28 A kind of recognition methods of Mental Workload level

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201511005142.8A CN106913351A (en) 2015-12-28 2015-12-28 A kind of recognition methods of Mental Workload level

Publications (1)

Publication Number Publication Date
CN106913351A true CN106913351A (en) 2017-07-04

Family

ID=59455394

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201511005142.8A Pending CN106913351A (en) 2015-12-28 2015-12-28 A kind of recognition methods of Mental Workload level

Country Status (1)

Country Link
CN (1) CN106913351A (en)

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107463250A (en) * 2017-07-11 2017-12-12 天津大学 The method for improving P300 spellings device using effect under Mental Workload state
CN107550501A (en) * 2017-08-30 2018-01-09 西南交通大学 The method of testing and system of high ferro dispatcher's Mental rotation ability
CN107550500A (en) * 2017-08-30 2018-01-09 西南交通大学 The method of testing and system of high ferro dispatcher's response inhabitation ability
CN109480871A (en) * 2018-10-30 2019-03-19 北京机械设备研究所 A kind of fatigue detection method towards RSVP brain-computer interface
CN109497997A (en) * 2018-12-10 2019-03-22 杭州妞诺科技有限公司 Based on majority according to the seizure detection and early warning system of acquisition
CN109602417A (en) * 2018-11-23 2019-04-12 杭州妞诺科技有限公司 Sleep stage method and system based on random forest
CN109953757A (en) * 2017-12-14 2019-07-02 中国航天员科研训练中心 Towards keep track control and shooting generic task Mental Workload method of real-time
CN110236532A (en) * 2019-04-30 2019-09-17 深圳和而泰家居在线网络科技有限公司 Processing of bioelectric signals method, apparatus, computer equipment and storage medium
CN111407292A (en) * 2020-03-30 2020-07-14 西北工业大学 Pilot workload assessment method based on eye movement and multi-parameter physiological data information
CN111839506A (en) * 2019-04-30 2020-10-30 清华大学 Mental load detection method and device
CN112773365A (en) * 2019-10-22 2021-05-11 上海交通大学 System for monitoring mental load of underwater vehicle during underwater operation

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103610447A (en) * 2013-12-04 2014-03-05 天津大学 Mental workload online detection method based on forehead electroencephalogram signals

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103610447A (en) * 2013-12-04 2014-03-05 天津大学 Mental workload online detection method based on forehead electroencephalogram signals

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
LI ZHIWEI等: "Classification of Mental Task EEG Signals Using Wavelet", 《 THE EIGHTH INTERNATIONAL CONFERENCE ON ELECTRONIC MEASUREMENT AND INSTRUMENTS 》 *
李南南: "N-back诱发脑力负荷信息监测与识别技术研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 *
郭孜政等: "基于EEG熵值的驾驶员脑力负荷水平识别方法", 《东南大学学报》 *

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107463250A (en) * 2017-07-11 2017-12-12 天津大学 The method for improving P300 spellings device using effect under Mental Workload state
CN107463250B (en) * 2017-07-11 2019-11-22 天津大学 The method for improving P300 spelling device using effect under Mental Workload state
CN107550501A (en) * 2017-08-30 2018-01-09 西南交通大学 The method of testing and system of high ferro dispatcher's Mental rotation ability
CN107550500A (en) * 2017-08-30 2018-01-09 西南交通大学 The method of testing and system of high ferro dispatcher's response inhabitation ability
CN109953757A (en) * 2017-12-14 2019-07-02 中国航天员科研训练中心 Towards keep track control and shooting generic task Mental Workload method of real-time
CN109480871A (en) * 2018-10-30 2019-03-19 北京机械设备研究所 A kind of fatigue detection method towards RSVP brain-computer interface
CN109602417A (en) * 2018-11-23 2019-04-12 杭州妞诺科技有限公司 Sleep stage method and system based on random forest
CN109497997A (en) * 2018-12-10 2019-03-22 杭州妞诺科技有限公司 Based on majority according to the seizure detection and early warning system of acquisition
CN110236532A (en) * 2019-04-30 2019-09-17 深圳和而泰家居在线网络科技有限公司 Processing of bioelectric signals method, apparatus, computer equipment and storage medium
CN111839506A (en) * 2019-04-30 2020-10-30 清华大学 Mental load detection method and device
CN111839506B (en) * 2019-04-30 2021-10-12 清华大学 Mental load detection method and device
CN112773365A (en) * 2019-10-22 2021-05-11 上海交通大学 System for monitoring mental load of underwater vehicle during underwater operation
CN111407292A (en) * 2020-03-30 2020-07-14 西北工业大学 Pilot workload assessment method based on eye movement and multi-parameter physiological data information

Similar Documents

Publication Publication Date Title
CN106913351A (en) A kind of recognition methods of Mental Workload level
Yao et al. A novel intelligent method for fault diagnosis of electric vehicle battery system based on wavelet neural network
Hu et al. Abnormal driving detection based on normalized driving behavior
CN103610447B (en) A kind of Mental Workload online test method based on forehead EEG signal
Chai et al. Driver fatigue classification with independent component by entropy rate bound minimization analysis in an EEG-based system
Yang et al. Comparison among driving state prediction models for car-following condition based on EEG and driving features
CN102149320A (en) Method and system for concentration detection
CN104665849A (en) Multi-physiological signal multi-model interaction-based high-speed railway dispatcher stress detecting method
CN104236911A (en) Train bogie bearing service process monitoring and fault diagnosis system and method
Kovaceva et al. Identification of aggressive driving from naturalistic data in car-following situations
CN106092600A (en) A kind of pavement identification method for proving ground strengthening road
CN103042922A (en) System and method for determining vehicle driving status information-based driver concentration
CN103471865A (en) Train suspension system failure isolation method based on LDA method
CN104807640A (en) Method of detecting life state of spatial rolling bearing based on vibration-sensitive time-frequency characteristics
CN106919948A (en) A kind of recognition methods for driving Sustained attention level
CN104688252A (en) Method for detecting fatigue status of driver through steering wheel rotation angle information
CN101843496A (en) Driving work-load kit detection method
CN102133100A (en) Sparse representation-based electroencephalogram signal detection method
CN113753059B (en) Method for predicting takeover capacity of driver under automatic driving system
CN104490391A (en) Combatant state monitoring system based on electroencephalogram signals
CN106913333A (en) A kind of choosing method of the sensivity feature index of Sustained attention level
Jiang et al. Denoising and chaotic feature extraction of electrocardial signals for driver fatigue detection by Kolmogorov entropy
CN105956574A (en) Rolling bearing service life state same-scale characterization and recognition method under different rotating speeds
You et al. A study on driver fatigue recognition based on SVM method
Zfnebi et al. Driver behavior quantitative models: Identification and classification of variables

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20170704