CN106073805B - A kind of fatigue detection method and device based on eye movement data - Google Patents
A kind of fatigue detection method and device based on eye movement data Download PDFInfo
- Publication number
- CN106073805B CN106073805B CN201610369357.6A CN201610369357A CN106073805B CN 106073805 B CN106073805 B CN 106073805B CN 201610369357 A CN201610369357 A CN 201610369357A CN 106073805 B CN106073805 B CN 106073805B
- Authority
- CN
- China
- Prior art keywords
- reading
- eye movement
- data
- search
- people
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
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
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B3/00—Apparatus for testing the eyes; Instruments for examining the eyes
- A61B3/10—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions
- A61B3/11—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions for measuring interpupillary distance or diameter of pupils
- A61B3/112—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions for measuring interpupillary distance or diameter of pupils for measuring diameter of pupils
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B3/00—Apparatus for testing the eyes; Instruments for examining the eyes
- A61B3/10—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions
- A61B3/113—Objective types, i.e. instruments for examining the eyes independent of the patients' perceptions or reactions for determining or recording eye movement
Abstract
The invention discloses a kind of fatigue detection method and device based on eye movement data.This method with people to be measured by carrying out the agility that UI interactive testings people to be measured reacts, eye movement master data is acquired by eye tracker while test and calculates eye movement fatigue characteristic data, is finally combined again by the agility and eye movement fatigue characteristic data that react and judges whether people to be measured is in fatigue state using machine learning algorithm.Wherein, the process that the agility that UI interactive testings people to be measured reacts is carried out with people to be measured is made of chracter search test and test in reading comprehension.After tested, Detection accuracy of the invention has reached 95.2%, compared to fatigue detecting method before, has many advantages, such as high accuracy, practicability, simplicity, learnability.
Description
Technical field
The present invention relates to eye movement data analysis and the applications in fatigue detecting.
Background technology
It is a series of studies have shown that different from traditional information browse and reading method, be easier to allow use using electronic curtain
Family generates feeling of fatigue, and the increase of fatigue strength can lead to low working efficiency, eyesight reduction or even more serious physical impairment, so
And many users in life and work and are unaware that this point.Therefore such as a kind of how quick, easy mode, accurately have
The fatigue state for judging user of effect is a current problem urgently to be resolved hurrily.
Have multinomial research before this and be mentioned to visual fatigue detection, but how to obtain an accurate and objective knot
This problem is not solved perfectly in fruit, most method designs.Ka Neijimeilong research institutes by testing and proving repeatedly, it is proposed that degree
The physical quantity PERCLOS (Percentage of Eyelid Closure over the Pupil) of amount fatigue/drowsiness.They
Think, the time within the unit interval shared by the eyes closed reaches certain proportion, then judgement produces visual fatigue.
In the work of Singh, it is believed that when being more than 0.5S between when the eyes are occluded, then be likely to produce fatigue state.Due to using herein
It is excessively single in the feature of detection, thus it be easy to cause erroneous judgement.Benedetto etc. is in homeostasis visual fatigue, in conjunction with flash of light
Stimulation is analyzed with frequency of wink, but one side flash stimulation may influence whether eye movement data, on the other hand influence
There are many factor of frequency of wink, and feeling of fatigue is only one of many factors.Di and Mccamy uses data analysis and questionnaire tune
Come to an end the method for conjunction, with three hours for a stage, calculates reading and pan rate of the user in multiple stages, and think to work as
When eyes are in fatigue state, rating results can be greatly reduced.However this method takes long, user experience is extremely unfriendly.
Invention content
Problem to be solved by this invention is for current existing fatigue detecting method excessively subjectivityization, without quantitative data
The phenomenon that carrying out homeostasis, a kind of method and apparatus detecting whether fatigue of design, this method and device can utilize eye tracker pair
The eyes of user are detected, and collect eye movement data in a manner of the most natural, simple " text search " and " reading ", then
Judge whether fatigue according to these eye movement datas.
To solve the above problems, the scheme that the present invention uses is as follows:
A kind of fatigue detection method based on eye movement data according to the present invention, includes the following steps:
S1:UI by carrying out chracter search task with people to be measured is interacted, and obtains search efficiency data;
S2:While executing step S1, the eye movement master data of people to be measured is acquired by eye tracker, and is acquired corresponding
Then time data calculates search eye movement fatigue characteristic data according to eye movement master data and time data;
S3:UI by with people to be measured read understanding task is interacted, and obtains reading efficiency data;
S4:While executing step S3, the eye movement master data of people to be measured is acquired by eye tracker, and is acquired corresponding
Then time data calculates eye movement in reading fatigue characteristic data according to eye movement master data and time data;
S5:It is tired to the search efficiency data, reading efficiency data, search eye movement fatigue characteristic data and eye movement in reading
Labor characteristic analyzes and determines whether people to be measured is in fatigue state using machine learning algorithm;
Described search eye movement fatigue characteristic data include:Pupil diameter ratio, frequency of wink and interest domain number;
The eye movement in reading fatigue characteristic data include:Pupil diameter ratio, frequency of wink and interest domain number;
The ratio of pupil diameter and initial pupil diameter when the pupil diameter is than to test;
The frequency of wink is the time interval averagely blinked;
The number that the interest domain number is less than specific threshold by counting eyeball saccade velocity obtains.
Further, the fatigue detection method according to the present invention based on eye movement data, described search efficiency data include word
Symbol search success rate and chracter search error rate;The step S1 includes:
S11:Search text is randomly selected from chracter search text library;
S12:The search text is shown in screen for people to be measured search, when display, from the search text
5~50 characters are randomly selected, selected character are shown with the pattern for being different from text, and to these to be different from text
Pattern show character structure spcial character sample region;
S13:The screen taps message of people's operation to be measured is obtained, and whether spcial character is fallen into according to screen taps message
Sample region determines whether that spcial character clicks message;
S14:Message is clicked according to screen taps message and spcial character, chracter search success rate is counted and chracter search is wrong
Accidentally rate.
Further, the fatigue detection method according to the present invention based on eye movement data, the reading efficiency data include readding
Read rate;The step S3 includes:
S31:Understand to randomly select in text library from reading and reads text and corresponding reading understanding topic collection;
S32:The reading text is shown in screen and reading understanding topic collection is read for people to be measured and answer reads reason
It solves a problem collection;
S33:When all reading understandings topic collection is answered correct, the reading of recording step S32, which understands, to be taken;
S34:Understood according to the total number of word of the reading text and reading and taken, calculates rate of reading.
Further, the fatigue detection method according to the present invention based on eye movement data, the eye movement in reading fatigue characteristic number
According to further including readback line number;The readback line number is more than specific threshold, and of eyeball pan by counting eyeball pan rate
The distance of point and terminal is more than specific threshold, and the abscissa of terminal is obtained less than the number of the abscissa of starting point.
Fatigue detection method based on eye movement data as described in claim 1, which is characterized in that the eye movement in reading is tired
Labor characteristic further includes staring total time accounting;The total time accounting of staring is to stare total time and read accounting for for total time
Than;Described stare is obtained total time by counting the total time of eyeball stationary state.
A kind of fatigue detection device based on eye movement data according to the present invention, comprises the following modules:
M1 is used for:UI by carrying out chracter search task with people to be measured is interacted, and obtains search efficiency data;
M2 is used for:While execution module M1, the eye movement master data of people to be measured is acquired by eye tracker, and acquires phase
Then the time data answered calculates search eye movement fatigue characteristic data according to eye movement master data and time data;
M3 is used for:UI by with people to be measured read understanding task is interacted, and obtains reading efficiency data;
M4 is used for:While execution module M3, the eye movement master data of people to be measured is acquired by eye tracker, and acquires phase
Then the time data answered calculates eye movement in reading fatigue characteristic data according to eye movement master data and time data;
M5 is used for:To the search efficiency data, reading efficiency data, search eye movement fatigue characteristic data and read eye
Dynamic fatigue characteristic data analyze and determine whether people to be measured is in fatigue state using machine learning algorithm;
Described search eye movement fatigue characteristic data include:Pupil diameter ratio, frequency of wink and interest domain number;
The eye movement in reading fatigue characteristic data include:Pupil diameter ratio, frequency of wink and interest domain number;
The ratio of pupil diameter and initial pupil diameter when the pupil diameter is than to test;
The frequency of wink is the time interval averagely blinked;
The number that the interest domain number is less than specific threshold by counting eyeball saccade velocity obtains.
Further, the fatigue detection device according to the present invention based on eye movement data, described search efficiency data include word
Symbol search success rate and chracter search error rate;The module M1 includes:
M11 is used for:Search text is randomly selected from chracter search text library;
M12 is used for:Show that the search text is searched for for people to be measured in screen, it is literary from the search when display
5~50 characters are randomly selected in this, selected character are shown with the pattern for being different from text, and to these to be different from
The sample region for the character structure spcial character that the pattern of text is shown;
M13 is used for:The screen taps message of people's operation to be measured is obtained, and whether special word is fallen into according to screen taps message
The sample region of symbol determines whether that spcial character clicks message;
M14 is used for:Message is clicked according to screen taps message and spcial character, chracter search success rate is counted and character is searched
Rope error rate.
Further, the fatigue detection device according to the present invention based on eye movement data, the reading efficiency data include readding
Read rate;The module M3 includes:
M31 is used for:Understand to randomly select in text library from reading and reads text and corresponding reading understanding topic collection;
M32 is used for:The reading text is shown in screen and reading understanding topic collection is read for people to be measured and answer is read
Read understanding topic collection;
M33 is used for:When all reading understandings topic collection is answered correct, the reading of logging modle M32, which understands, to be taken;
M34 is used for:Understood according to the total number of word of the reading text and reading and taken, calculates rate of reading.
Further, the fatigue detection device according to the present invention based on eye movement data, the eye movement in reading fatigue characteristic number
According to further including readback line number;The readback line number is more than specific threshold, and of eyeball pan by counting eyeball pan rate
The distance of point and terminal is more than specific threshold, and the abscissa of terminal is obtained less than the number of the abscissa of starting point.
Further, the fatigue detection device according to the present invention based on eye movement data, the eye movement in reading fatigue characteristic number
According to further including staring total time accounting;The total time accounting of staring is to stare total time and read the accounting of total time;It is described
It stares and is obtained by counting the total time of eyeball stationary state total time.
The technique effect of the present invention is as follows:The present invention takes full-automatic mode to collect data, can be current to user simultaneously
Fatigue state, working efficiency assessed, device process for using only needs one people of user that all operations can be completed, after tested,
The Detection accuracy of the present invention has reached 95.2%, compared to fatigue detecting method before, has high accuracy, practicality
The advantages that property, simplicity, learnability.
Specific implementation mode
The present invention is described in further details below.
The present embodiment is related to a kind of fatigue detecting machine, which includes:Eye tracker, display, host and input equipment.
Wherein eye tracker is for acquiring eye movement master data.Display and input equipment carry out UI interactions for people to be measured and host.It is defeated
It can be keyboard or mouse to enter equipment, if if display is touch screen certainly, then input equipment may be the display
Device.Host is realized for running program by the operation of program:It is realized by display and input equipment and the UI of people to be measured is handed over
The agility of people's reaction to be measured is mutually tested, and obtains the eye movement of eye tracker acquisition while testing the agility of people's reaction to be measured
Then whether master data carries out the agility of reaction and eye movement master data analyzing and determining out current people to be measured in fatigue
State.The above process that above-mentioned mainframe program is realized is the fatigue detection method based on eye movement data of the present invention.This reality
It applies in example, the fatigue detection method based on eye movement data includes the following steps:
S1:UI by carrying out chracter search task with people to be measured is interacted, and obtains search efficiency data;
S2:While executing step S1, the eye movement master data of people to be measured is acquired by eye tracker, and is acquired corresponding
Then time data calculates search eye movement fatigue characteristic data according to eye movement master data and time data;
S3:UI by with people to be measured read understanding task is interacted, and obtains reading efficiency data;
S4:While executing step S3, the eye movement master data of people to be measured is acquired by eye tracker, and is acquired corresponding
Then time data calculates eye movement in reading fatigue characteristic data according to eye movement master data and time data;
S5:It is tired to the search efficiency data, reading efficiency data, search eye movement fatigue characteristic data and eye movement in reading
Labor characteristic analyzes and determines whether people to be measured is in fatigue state using machine learning algorithm.
In above-mentioned steps, step S1 and S3 be two by the agility of UI interactive testings people to be measured reaction the step of.Step
The step of rapid S2 and S4 is execution synchronous with step S1 and S3 respectively.Therefore, step S2 and S1 can essentially be combined into one, letter
Referred to as chracter search is tested;Step S4 and S3 can also be combined into one, referred to as test in reading comprehension.Therefore the entire above process
It can be understood as:Chracter search is tested first, then test in reading comprehension, and ultimate analysis judges.Last analytical judgment process
As step S5.It should be pointed out that chracter search test and test in reading comprehension are two processes in no particular order, it can also
Test in reading comprehension is carried out first, and then chracter search is tested, and is finally analyzed and determined again.
The search efficiency data and reading efficiency data that step S1 and S3 are obtained are reacting person's development agility to be measured.
In the present embodiment, search efficiency data include two indices value:Chracter search success rate and chracter search error rate.The present embodiment
Step S1 specific implementation use following S101, S102, S11, S12, S13 and S14 step.
S101:By the explanation of screen display chracter search test assignment, and the test that starts of people to be measured to be received is waited to refer to
It enables.
S102:Starting after testing instruction for people to be measured is received, step S11, S12, S13 and S14 step is executed.
S11:Search text is randomly selected from chracter search text library.Chracter search text library is previously stored with many pieces
Article.Search text is the article that search is read for people to be measured.Article quantity in chracter search text library is generally no less than
100.The language of article is also to preset, and when such as using Chinese as the people to be measured of mother tongue, the language of article is Chinese;Needle
When to using English as the people to be measured of mother tongue, the language of article is English.It can certainly in advance be deposited in chracter search text library
The article of multilingual is stored up, then when randomly selecting search text, can select and grasp according to the language environment of operating system
Make the consistent article conduct search text of the language form of system.The number of words of each piece article in chracter search text library mutually compares
It is closer to, maximum difference is also no more than 10.
S12:The search text is shown in screen for people to be measured search, when display, from the search text
5~50 characters are randomly selected, selected character are shown with the pattern for being different from text, and to these to be different from text
Pattern show character structure spcial character sample region.5~50 characters randomly selected are known as spcial character.Also
It is when display, to search for the character in text in addition to spcial character and shown using a kind of text style, and spcial character is adopted
It is shown with another text style.For example, text style is using small No. four used by character in addition to spcial character
The Song typeface;And spcial character has 50 characters, in 50 characters, 10 characters use small No. four regular scripts, 10 characters to use
The italics of small No. four Song typefaces, 10 characters use No. five Song typefaces, 10 characters that No. four Song typefaces, 10 characters is used to adopt
With the Song typeface of No. four runics.It is to wait for people's screen operator event message to be measured after step S12.
S13:The screen taps message of people's operation to be measured is obtained, and whether spcial character is fallen into according to screen taps message
Sample region determines whether that spcial character clicks message.Specifically, judge whether mouse motor or screen touch click are clicked
On spcial character.Here the sample region of spcial character is generated by step S12.
S14:Message is clicked according to screen taps message and spcial character, chracter search success rate is counted and chracter search is wrong
Accidentally rate.Chracter search success rate uses formula Rc=Nc/NsIt is calculated.Chracter search error rate uses formula Rw=Nw/NsMeter
It obtains.Wherein, NcAnd NwShow respectively the right and wrong result number that people to be measured searches, NsIt indicates in reading material
The total number that special word occurs.
In the present embodiment, reading efficiency data are rate of reading.The specific implementation of the step S3 of the present embodiment is using as follows
S301, S302, S31, S32, S33 and S34 step.
S301:By the explanation of screen display test in reading comprehension task, and the test that starts of people to be measured to be received is waited to refer to
It enables.
S302:Starting after testing instruction for people to be measured is received, step S31, S32, S33 and S34 step is executed.
S31:Understand to randomly select in text library from reading and reads text and corresponding reading understanding topic collection.It reads and understands text
This library and chracter search text library above-mentioned are similar, are not both, and reading understands that every article institute is also stored in text library is right
The reading understanding topic collection answered.Read the reading understanding topic number for reading understanding topic collection understood in text library corresponding to each piece article
It is identical.It reads understanding topic number and is generally 8~20.The form for reading understanding topic is generally multiple-choice question.
S32:The reading text is shown in screen and reading understanding topic collection is read for people to be measured and answer reads reason
It solves a problem collection.Then it waits for people to be measured to read and answer and reads understanding topic collection.When people to be measured, which answers one, reads understanding topic, i.e.,
It can determine whether people's answer to be measured is correct.
S33:When all reading understandings topic collection is answered correct, the reading of recording step S32, which understands, to be taken.It needs to illustrate
, in the present embodiment, it is all each understanding topic of reading by especially selecting and designing to read text and read understanding topic collection
Answer can find answer in reading text.Therefore, the normal people of general IQ will not can not reply.
S34:Understood according to the total number of word of the reading text and reading and taken, calculates rate of reading.Rate of reading is
Read the total number of word of the text ratio time-consuming with understanding is read.
The search eye movement fatigue characteristic data and eye movement in reading fatigue characteristic data that step S2 and S4 are obtained are that characterization is tired
The eye movement characteristics data of labor index.The eye movement master data that eye tracker acquires people to be measured includes real-time pupil diameter, in real time
Eyeball pan point and in real time eyes open and-shut mode.In the present embodiment, search eye movement fatigue characteristic data include:Pupil diameter
Than, frequency of wink and interest domain number.Eye movement in reading fatigue characteristic data include:Pupil diameter ratio, frequency of wink, interest domain
Number and stares total time accounting at readback line number.Pupil diameter ratio, frequency of wink in search eye movement fatigue characteristic data, interest
Domain number is identical with pupil diameter ratio, frequency of wink, the acquisition pattern of interest domain number in eye movement in reading fatigue characteristic data.
The ratio of pupil diameter and initial pupil diameter when pupil diameter is than to test.Initial pupil diameter is word
The pupil diameter sampled by eye tracker before symbol search test and test in reading comprehension.Pupil diameter when test is word
In the real-time pupil diameter that symbol search test and test in reading comprehension are sampled by eye tracker in the process with initial pupil
Bore dia has the pupil diameter of maximum difference.For example, in chracter search test, obtained real-time pupil diameter is sequence
{D1,D2,D3,...,DN};Minimum pupil diameter D can then be obtainedmin=min (D1,D2,D3,...,DN), minimum pupil diameter
DminPupil diameter when as chracter search is tested.It should be noted that pupil diameter ratio here includes the pupil of left eye
Diameter is than the pupil diameter ratio with right eye.
Frequency of wink is the time interval averagely blinked when test.Blink acquires real-time eyes by eye tracker and is opened and closed shape
State obtains.The present embodiment, eyes are from open state to closed state again to the one-shot change of open state as one of eyes blink
Action." blink " in the present embodiment " time interval averagely blinked " is completed at the same time a blink action once to blink with eyes
Eye.Time point can be recorded when blink occurs every time, the adjacent time interval blinked twice is then calculated according to the time point, from
And obtain the sequence { C of the adjacent time interval blinked twice1,C2,C3,...,CM}.Thus frequency of wink=avg (C1,C2,
C3,...,CN), wherein avg is mean value calculation formula, specially
The number that interest domain number is less than specific threshold by counting eyeball saccade velocity obtains.Specific to the present embodiment
In, the statistics of interest domain number uses following steps:
S81:Real-time eyeball is acquired by eye tracker for time interval with every 250 milliseconds and sweeps point, to be swept
Point sequence { P1,P2,P3,...,PK+1};Wherein, PiIt is made of abscissa and ordinate.
S82:Calculate pan point sequence { P1,P2,P3,...,PK+1In it is two neighboring pan point distance, obtain apart from sequence
Arrange { L1,L2,L3,...,LK};Wherein, Li=Dist (Pi,Pi+1), Dist is distance calculation formula.It should be noted that here
{ L1,L2,L3,...,LKAlthough the two neighboring distance for sweeping point is represented, due between the time of sampling eyeball pan point
Every being fixed, each range data implies 250 milliseconds of time span, and therefore, the eyeball that they also represent moment is swept
Depending on speed.
S83:Calculate distance sequence { L1,L2,L3,...,LKWindow sliding average value, obtain equal value sequence { V1,V2,
V3,...,VK};Wherein Vi=avg (Li-3,Li-2,Li-1,Li,Li+1,Li+2,Li+3,Li+4) namely the size of sliding window be
8.For i-3<1, i+4 is more than the situation of K, then rejects.Such as when i=2, the V of actual count2=avg (Li-1,Li,Li+1,
Li+2,Li+3,Li+4)=avg (L1,L2,L3,L4,L5,L6).Such as when i=K-1, the V of actual countK-1=avg (LK-4,LK-3,
LK-2,LK-1,LK).Due to distance sequence { L1,L2,L3,...,LKIn each range data imply 250 milliseconds time it is long
Degree, continuous 8 250 milliseconds of times for meaning 2 seconds length, that is, equal value sequence { V1,V2,V3,...,VKEach mean value can
To be considered as the eyeball saccade velocity in 2 seconds length.
S84:Calculate equal value sequence { V1,V2,V3,...,VKMaximum value Vmax, minimum value Vmin and average value Vavg,
Then specific threshold Vlo=(Vmax+3 × Vmin+4 × Vavg)/8 is calculated.
S85:Calculate Boolean sequence { B1,B2,B3,...,BK, work as ViWhen less than specific threshold Vlo, Boolean BiIt is 1, it is no
Then Boolean BiIt is 0, wherein i ∈ [1..K].
S86:From Boolean sequence { B1,B2,B3,...,BK-7Reject and continuously repeat the Boolean for 1 after, count remaining boolean
The number that Boolean is 1 in sequence is as interest domain number.Here the Boolean continuously repeated as 1 is rejected, for example, B1,B2,B3
It is 1, and B4It is 0, at this point, B2,B3It is 1 to continuously repeat, rejects B2,B3It is left B afterwards1,B4,......。
Readback line number in eye movement in reading fatigue characteristic data is more than specific threshold, and eye by counting eyeball pan rate
The distance of the beginning and end of ball pan is more than specific threshold, and the abscissa of terminal is obtained less than the number of the abscissa of starting point
It arrives.Specific in the present embodiment, the statistics of readback line number passes through following steps:
S91:Real-time eyeball is acquired by eye tracker for time interval with every 250 milliseconds and sweeps point, to be swept
Point sequence { P1,P2,P3,...,PK+1};Wherein, PiIt is made of abscissa and ordinate.The step is identical as abovementioned steps S81.
S92:Calculate pan point sequence { P1,P2,P3,...,PK+1In it is two neighboring pan point distance, obtain apart from sequence
Arrange { L1,L2,L3,...,LK};Wherein, Li=Dist (Pi,Pi+1), Dist is distance calculation formula.This step and abovementioned steps
S82 is identical.
S93:Calculate distance sequence { L1,L2,L3,...,LKWindow sliding average value, obtain equal value sequence { V1,V2,
V3,...,VK}.This step is identical as abovementioned steps S83, repeats no more.
S94:Calculate equal value sequence { V1,V2,V3,...,VKMaximum value Vmax, minimum value Vmin and average value Vavg,
Then specific threshold Vhi=(2 × Vmax+Vmin+5 × Vavg)/8 is calculated.
S95:Calculate distance sequence { L1,L2,L3,...,LKMaximum value Lmax, minimum value Lmin and average value Lavg,
Then specific threshold Lhi=(2 × Lmax+Lmin+5 × Lavg)/8 is calculated.
S96:Calculate Boolean sequence { B1,B2,B3,...,BK, work as ViWhen more than specific threshold Vhi, and LiMore than certain threshold
Value Lhi, and terminal Pi+1Abscissa be less than starting point PiAbscissa when, count Boolean BiIt is 1, otherwise Boolean BiIt is 0.
S97:Count Boolean sequence { B1,B2,B3,...,BKIn Boolean be 1 number as readback line number.
Total time accounting of staring in eye movement in reading fatigue characteristic data is to stare total time and read the accounting of total time.
It stares and is obtained by counting the total time of eyeball stationary state total time.Staring total time accounting can be by abovementioned steps S82
Or distance sequence { the L in S921,L2,L3,...,LKEstimate to obtain.Specially:Statistical distance sequence { L1,L2,L3,...,LK}
In number Nz less than 0.00001 as total time is stared, with distance sequence { L1,L2,L3,...,LKIn K it is total as reading
Time.Total time accounting=Nz/K is stared as a result,.
Fatigue characteristic achievement data has been obtained by above-mentioned chracter search test and test in reading comprehension.Fatigue characteristic refers to
Marking data includes:Search efficiency data, reading efficiency data, search eye movement fatigue characteristic data and eye movement in reading fatigue characteristic number
According to.Search efficiency data include chracter search success rate and chracter search error rate, and chracter search success rate and chracter search are wrong
Accidentally rate is respectively labeled as:Res_suc and Res_err.Reading efficiency data include rate of reading, and rate of reading is labeled as R_
speed.Search eye movement fatigue characteristic data include pupil diameter ratio, frequency of wink and interest domain number, and search eye movement fatigue is special
Sign data pupil of left eye diameter ratio, pupil of right eye diameter ratio, frequency of wink and interest domain number be respectively labeled as SLD, SRD,
S_blink and S_AOI.Eye movement in reading fatigue characteristic data include pupil diameter ratio, frequency of wink, interest domain number, readback row
Count and stare total time accounting.Pupil of left eye diameter ratio, pupil of right eye diameter ratio, the blink frequency of eye movement in reading fatigue characteristic data
It rate, interest domain number, readback line number and stares total time accounting and is respectively labeled as:RLD、RRD、R_blink、R_AOI、R_sac
And Stare_t.
Step S5 namely to above-mentioned fatigue characteristic data Res_suc, Res_err, R_speed, SLD, SRD, S_blink,
Whether S_AOI, RLD, RRD, R_blink, R_AOI, R_sac and Stare_t analyze and determine people to be measured using machine learning algorithm
In fatigue state.In the present embodiment, machine learning algorithm uses BP neural network algorithm.BP neural network algorithm is this field
Technology known to technical staff, this specification repeat no more.It should be pointed out that before step S5 is judged, BP god
It needs, by model training, then to carry out above-mentioned fatigue characteristic data according to the model result trained through network algorithm
Classification, judges normal condition and fatigue state.And for the fatigue detecting machine of the present embodiment, BP neural network algorithm
Model training be to have first carried out in advance.
Claims (6)
1. a kind of fatigue detection method based on eye movement data, which is characterized in that include the following steps:
S1:UI by carrying out chracter search task with people to be measured is interacted, and obtains search efficiency data;
S2:While executing step S1, the eye movement master data of people to be measured is acquired by eye tracker, and acquires the corresponding time
Then data calculate search eye movement fatigue characteristic data according to eye movement master data and time data;
S3:UI by with people to be measured read understanding task is interacted, and obtains reading efficiency data;
S4:While executing step S3, the eye movement master data of people to be measured is acquired by eye tracker, and acquires the corresponding time
Then data calculate eye movement in reading fatigue characteristic data according to eye movement master data and time data;
S5:It is special to the search efficiency data, reading efficiency data, search eye movement fatigue characteristic data and eye movement in reading fatigue
It levies data and analyzes and determines whether people to be measured is in fatigue state using machine learning algorithm;
Described search eye movement fatigue characteristic data include:Pupil diameter ratio, frequency of wink and interest domain number;
The eye movement in reading fatigue characteristic data include:Pupil diameter ratio, frequency of wink and interest domain number;
The ratio of pupil diameter and initial pupil diameter when the pupil diameter is than to test;
The frequency of wink is the time interval averagely blinked;
The number that the interest domain number is less than specific threshold by counting eyeball saccade velocity obtains;
Described search efficiency data includes chracter search success rate and chracter search error rate;The step S1 includes:
S11:Search text is randomly selected from chracter search text library;
S12:Show that the search text is searched for for people to be measured in screen, when display, from the search text at random
5 ~ 50 characters are chosen, selected character are shown with the pattern for being different from text, and to these to be different from the sample of text
The sample region for the character structure spcial character that formula is shown;
S13:The screen taps message of people's operation to be measured is obtained, and whether falls into the sampling of spcial character according to screen taps message
Area determines whether that spcial character clicks message;
S14:Message is clicked according to screen taps message and spcial character, counts chracter search success rate and chracter search mistake
Rate;
The reading efficiency data include rate of reading;The step S3 includes:
S31:Understand to randomly select in text library from reading and reads text and corresponding reading understanding topic collection;
S32:The reading text is shown in screen and reading understanding topic collection is read for people to be measured and answer reads understanding topic
Collection;
S33:When all reading understandings topic collection is answered correct, the reading of recording step S32, which understands, to be taken;
S34:Understood according to the total number of word of the reading text and reading and taken, calculates rate of reading.
2. the fatigue detection method based on eye movement data as described in claim 1, which is characterized in that the eye movement in reading fatigue
Characteristic further includes readback line number;The readback line number sweeps rate more than specific threshold by counting eyeball, and eyeball is swept
Depending on the distance of beginning and end be more than specific threshold, and the abscissa of terminal is less than the number of abscissa of starting point and obtains.
3. the fatigue detection method based on eye movement data as described in claim 1, which is characterized in that the eye movement in reading fatigue
Characteristic further includes staring total time accounting;The total time accounting of staring is to stare total time and read accounting for for total time
Than;Described stare is obtained total time by counting the total time of eyeball stationary state.
4. a kind of fatigue detection device based on eye movement data, which is characterized in that comprise the following modules:
M1 is used for:UI by carrying out chracter search task with people to be measured is interacted, and obtains search efficiency data;
M2 is used for:While execution module M1, the eye movement master data of people to be measured is acquired by eye tracker, and is acquired corresponding
Then time data calculates search eye movement fatigue characteristic data according to eye movement master data and time data;
M3 is used for:UI by with people to be measured read understanding task is interacted, and obtains reading efficiency data;
M4 is used for:While execution module M3, the eye movement master data of people to be measured is acquired by eye tracker, and is acquired corresponding
Then time data calculates eye movement in reading fatigue characteristic data according to eye movement master data and time data;
M5 is used for:It is tired to the search efficiency data, reading efficiency data, search eye movement fatigue characteristic data and eye movement in reading
Labor characteristic analyzes and determines whether people to be measured is in fatigue state using machine learning algorithm;
Described search eye movement fatigue characteristic data include:Pupil diameter ratio, frequency of wink and interest domain number;
The eye movement in reading fatigue characteristic data include:Pupil diameter ratio, frequency of wink and interest domain number;
The ratio of pupil diameter and initial pupil diameter when the pupil diameter is than to test;
The frequency of wink is the time interval averagely blinked;
The number that the interest domain number is less than specific threshold by counting eyeball saccade velocity obtains;
Described search efficiency data includes chracter search success rate and chracter search error rate;The module M1 includes:
M11 is used for:Search text is randomly selected from chracter search text library;
M12 is used for:The search text is shown in screen for people to be measured search, when display, from the search text
5 ~ 50 characters are randomly selected, selected character are shown with the pattern for being different from text, and to these to be different from text
Pattern show character structure spcial character sample region;
M13 is used for:The screen taps message of people's operation to be measured is obtained, and whether spcial character is fallen into according to screen taps message
Sample region determines whether that spcial character clicks message;
M14 is used for:Message is clicked according to screen taps message and spcial character, chracter search success rate is counted and chracter search is wrong
Accidentally rate;
The reading efficiency data include rate of reading;The module M3 includes:
M31 is used for:Understand to randomly select in text library from reading and reads text and corresponding reading understanding topic collection;
M32 is used for:The reading text is shown in screen and reading understanding topic collection is read for people to be measured and answer reads reason
It solves a problem collection;
M33 is used for:When all reading understandings topic collection is answered correct, the reading of logging modle M32, which understands, to be taken;
M34 is used for:Understood according to the total number of word of the reading text and reading and taken, calculates rate of reading.
5. the fatigue detection device based on eye movement data as claimed in claim 4, which is characterized in that the eye movement in reading fatigue
Characteristic further includes readback line number;The readback line number sweeps rate more than specific threshold by counting eyeball, and eyeball is swept
Depending on the distance of beginning and end be more than specific threshold, and the abscissa of terminal is less than the number of abscissa of starting point and obtains.
6. the fatigue detection device based on eye movement data as claimed in claim 4, which is characterized in that the eye movement in reading fatigue
Characteristic further includes staring total time accounting;The total time accounting of staring is to stare total time and read accounting for for total time
Than;Described stare is obtained total time by counting the total time of eyeball stationary state.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610369357.6A CN106073805B (en) | 2016-05-30 | 2016-05-30 | A kind of fatigue detection method and device based on eye movement data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610369357.6A CN106073805B (en) | 2016-05-30 | 2016-05-30 | A kind of fatigue detection method and device based on eye movement data |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106073805A CN106073805A (en) | 2016-11-09 |
CN106073805B true CN106073805B (en) | 2018-10-19 |
Family
ID=57230292
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610369357.6A Active CN106073805B (en) | 2016-05-30 | 2016-05-30 | A kind of fatigue detection method and device based on eye movement data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106073805B (en) |
Families Citing this family (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106569928A (en) * | 2016-11-01 | 2017-04-19 | 佛山科学技术学院 | Eye usage protection method and system thereof |
CN106667429B (en) * | 2017-02-20 | 2018-05-04 | 重庆市肿瘤研究所 | The vision that signal is moved based on eye induces cinetosis detection method |
CN108596106B (en) * | 2018-04-26 | 2023-12-05 | 京东方科技集团股份有限公司 | Visual fatigue recognition method and device based on VR equipment and VR equipment |
CN109700472A (en) * | 2019-02-21 | 2019-05-03 | 北京七鑫易维信息技术有限公司 | A kind of fatigue detection method, device, equipment and storage medium |
WO2020226603A1 (en) * | 2019-05-03 | 2020-11-12 | Сергей Анатольевич ДАНИЛОВ | Automated method and system for determining an extent to which information is recognized and automated method for verifying familiarization with an electronic document |
CN110495895B (en) * | 2019-08-26 | 2020-04-28 | 重庆大学 | Fatigue detection method and system based on eye movement tracking |
EP4113483A4 (en) * | 2020-02-28 | 2024-03-13 | Daikin Ind Ltd | Efficiency estimation device |
CN113509189A (en) * | 2021-07-07 | 2021-10-19 | 科大讯飞股份有限公司 | Learning state monitoring method and related equipment thereof |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1561179A (en) * | 2001-11-08 | 2005-01-05 | 睡眠诊断学公司 | Alertness monitor |
CN101132729A (en) * | 2005-03-04 | 2008-02-27 | 睡眠诊断学公司 | Measuring alertness |
CN101686815A (en) * | 2007-06-27 | 2010-03-31 | 松下电器产业株式会社 | Human condition estimating device and method |
CN104159497A (en) * | 2012-03-09 | 2014-11-19 | 奥斯派克特公司 | Method for assessing function of the visual system and apparatus thereof |
CN104504404A (en) * | 2015-01-23 | 2015-04-08 | 北京工业大学 | Online user type identification method and system based on visual behavior |
CN104504390A (en) * | 2015-01-14 | 2015-04-08 | 北京工业大学 | On-line user state recognition method and device based on eye movement data |
CN104636890A (en) * | 2015-03-13 | 2015-05-20 | 中国民航大学 | Measurement method for workload of air traffic controller |
CN204813795U (en) * | 2015-07-24 | 2015-12-02 | 中国人民解放军空军航空医学研究所 | Device based on it is tired that eye movement pan speed detects maincenter |
CN105513280A (en) * | 2016-01-15 | 2016-04-20 | 苏州大学 | Fatigue driving detection method |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7344251B2 (en) * | 2005-02-23 | 2008-03-18 | Eyetracking, Inc. | Mental alertness level determination |
US9854966B2 (en) * | 2011-11-22 | 2018-01-02 | Dignity Health | System and method for using microsaccade dynamics to measure attentional response to a stimulus |
US20150213634A1 (en) * | 2013-01-28 | 2015-07-30 | Amit V. KARMARKAR | Method and system of modifying text content presentation settings as determined by user states based on user eye metric data |
-
2016
- 2016-05-30 CN CN201610369357.6A patent/CN106073805B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1561179A (en) * | 2001-11-08 | 2005-01-05 | 睡眠诊断学公司 | Alertness monitor |
CN101132729A (en) * | 2005-03-04 | 2008-02-27 | 睡眠诊断学公司 | Measuring alertness |
CN101686815A (en) * | 2007-06-27 | 2010-03-31 | 松下电器产业株式会社 | Human condition estimating device and method |
CN104159497A (en) * | 2012-03-09 | 2014-11-19 | 奥斯派克特公司 | Method for assessing function of the visual system and apparatus thereof |
CN104504390A (en) * | 2015-01-14 | 2015-04-08 | 北京工业大学 | On-line user state recognition method and device based on eye movement data |
CN104504404A (en) * | 2015-01-23 | 2015-04-08 | 北京工业大学 | Online user type identification method and system based on visual behavior |
CN104636890A (en) * | 2015-03-13 | 2015-05-20 | 中国民航大学 | Measurement method for workload of air traffic controller |
CN204813795U (en) * | 2015-07-24 | 2015-12-02 | 中国人民解放军空军航空医学研究所 | Device based on it is tired that eye movement pan speed detects maincenter |
CN105513280A (en) * | 2016-01-15 | 2016-04-20 | 苏州大学 | Fatigue driving detection method |
Also Published As
Publication number | Publication date |
---|---|
CN106073805A (en) | 2016-11-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106073805B (en) | A kind of fatigue detection method and device based on eye movement data | |
Hawkins et al. | Racing against the clock: Evidence-based versus time-based decisions. | |
Iacobucci et al. | A meditation on mediation: Evidence that structural equations models perform better than regressions | |
CN106920129B (en) | Eye tracking-based network advertisement effect evaluation system and method | |
CN107126222B (en) | Cognitive ability evaluation system and evaluation method thereof | |
KR101102004B1 (en) | A method and system for quantitating fatigue resulting from a three dimensional display | |
EP2542144B1 (en) | Adaptive visual performance testing system | |
Ruscio et al. | A nontechnical introduction to the taxometric method | |
US20180174288A1 (en) | SCORE WEIGHTS FOR USER INTERFACE (ui) ELEMENTS | |
CN101738998B (en) | System and method for monitoring industrial process based on local discriminatory analysis | |
CN110189383A (en) | Chinese medicine tongue color coating colour quantitative analysis method based on machine learning | |
Prudêncio et al. | Analysis of instance hardness in machine learning using item response theory | |
CN108399366A (en) | It is a kind of based on the remote sensing images scene classification extracting method classified pixel-by-pixel | |
CN113749619A (en) | Mental fatigue assessment method based on K-TRCA | |
Sokolov | Sensitivity of goodness of fit indices to lack of measurement invariance with categorical indicators and many groups | |
CN116087647A (en) | Building electrical fault diagnosis method for optimizing random forest based on PCA and sparrow algorithm | |
CN110457895A (en) | A kind of PC application program violation content monitoring method and device | |
CN108932593B (en) | Cognitive influence factor analysis method and device | |
AU2020100135A4 (en) | Method, system and apparatus for evaluating sensory assessors’ concentration ability | |
DeCastellarnau et al. | Two approaches to evaluate measurement quality in online surveys: An application using the norwegian citizen panel | |
CN113361780A (en) | Behavior data-based crowdsourcing tester evaluation method | |
CN109994207B (en) | Mental health early warning method, server and system | |
CN106295957A (en) | Occupation Competency Model system and the method for analysis | |
Goldberg et al. | Eye tracking on visualizations: Progressive extraction of scanning strategies | |
Trewin et al. | Age-specific predictive models of human performance |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |