CN109389806A - Fatigue driving detection method for early warning, system and medium based on multi-information fusion - Google Patents
Fatigue driving detection method for early warning, system and medium based on multi-information fusion Download PDFInfo
- Publication number
- CN109389806A CN109389806A CN201811325931.3A CN201811325931A CN109389806A CN 109389806 A CN109389806 A CN 109389806A CN 201811325931 A CN201811325931 A CN 201811325931A CN 109389806 A CN109389806 A CN 109389806A
- Authority
- CN
- China
- Prior art keywords
- pulse
- feature
- fatigue
- considered
- fatigue characteristic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
- G08B21/02—Alarms for ensuring the safety of persons
- G08B21/06—Alarms for ensuring the safety of persons indicating a condition of sleep, e.g. anti-dozing alarms
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
- A61B5/0205—Simultaneously evaluating both cardiovascular conditions and different types of body conditions, e.g. heart and respiratory condition
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
- A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
- A61B5/1126—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb using a particular sensing technique
- A61B5/1128—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb using a particular sensing technique using image analysis
-
- 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
- A61B5/24—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
- A61B5/68—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
- A61B5/6887—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient mounted on external non-worn devices, e.g. non-medical devices
- A61B5/6893—Cars
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Surgery (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- General Health & Medical Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Physiology (AREA)
- Cardiology (AREA)
- Psychology (AREA)
- Psychiatry (AREA)
- Dentistry (AREA)
- Radiology & Medical Imaging (AREA)
- Child & Adolescent Psychology (AREA)
- Developmental Disabilities (AREA)
- Educational Technology (AREA)
- Hospice & Palliative Care (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Social Psychology (AREA)
- Business, Economics & Management (AREA)
- Emergency Management (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Pulmonology (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
- Image Processing (AREA)
Abstract
The present disclosure discloses fatigue driving detection method for early warning, system and media based on multi-information fusion, the facial image information under driver's driving condition is obtained by high-definition camera built in automobile cab, and identification and eye feature extraction, analysis are carried out to the image information of acquisition；Pulse data is acquired by the electronics bracelet with photoelectric sphyg sensor, carries out feature-extraction analysis；Wheel steering data are obtained by steering wheel angle sensor；Extracted feature is carried out fusion to judge whether by MCU in fatigue driving state, and makes corresponding warning.Compared with prior art, the disclosure has been put forward for the first time facial information, pulse data, steering wheel rotation angle and has driven four kinds of information fusions of duration, data information acquisition process reduces the interference to driver to greatest extent, and handles fatigue driving by multi-source information and judge that accuracy is higher.
Description
Technical field
This disclosure relates to traffic safety protection field, and in particular to the fatigue driving based on multi-information fusion detects the pre- police
Method, system and medium.
Background technique
With the continuous development of global economy, the demand of volume of transport is continuously increased, and the ownership of personal car is also held
One of an important factor for continuous rapid growth, traffic accident increase therewith, and wherein fatigue driving is induced as traffic accident, becomes friendship
Lead to the main hidden danger of safety.
Driving fatigue, that is, driver generates physiological function and the heart under uninterrupted driving for a long time or non-energetic state
The imbalance of function is managed, and the phenomenon that driving efficiency decline is objectively occurring.Driver's poor sleeping quality or deficiency, if for a long time
Vehicle is driven, then is easy to appear fatigue.Driving fatigue influences whether attention, feeling, consciousness, thinking, the judgement, meaning of driver
The aspects such as will, decision and movement.Fatigue is divided into actively tired and passive fatigue, tired due to caused by sleep insufficiency or disease
Labor is called actively fatigue；Lack the referred to as passive fatigue of fatigue caused by excitation since operating environment is dull.Fatigue is subsequent
It continuing and sails, driver can feel sleepy tired, limbs fatigue, blurred vision, and attention can not be concentrated, it is slow in reacting, judge energy
Power decline, or even absent-minded immediate memory occur and disappear, operation of making a fault causes to lose control of one's vehicle, easily to lead to major motor vehicle
Safety accident.
According to statistics, U.S.'s traffic accident as caused by driver tired driving every year is annual in China there are about 100,000
Traffic accident also have more than 30% be by driver tired driving caused by, and accident casualty is heavy, seriously affects the people's
Safety of life and property.Therefore, it is particularly important to improve vehicle active safety for a kind of fatigue driving detection device of researching and designing.
As the detection method of preventing driver fatigue driving, a large amount of scholars study it, so far
The method of use is broadly divided into following two major classes:
(1) using information of vehicles as predominantly detecting object.By acquiring the running condition information of vehicle, such as speed change
Change, vehicle driving trace etc. judge whether driver enters fatigue state indirectly.The deficiency of this method is accuracy of identification not
Height, the interference such as road conditions is larger when collection analysis data, can not accurately identify the driving condition of driver.
(2) using driver as predominantly detecting object.By acquiring Driver physiological data, such as brain wave, pulse, the heart
Rate, breathing etc., and by the facial image information of camera acquisition driver, the states such as eyes, the mouth of driver are analyzed,
To judge the fatigue conditions of driver.But the deficiency of this method is not interfere data in the case where driver's normal driving
Acquisition difficulty is big, is typically necessary earphone, glasses, the cap etc. worn and have sensor, causes to the normal driving of driver
Interference increases driving safety hidden danger, and is related to the security etc. of driver.
In conclusion main problem of the existing technology is: how relieving fatigue drive identification process in the presence of
In place of the deficiencies of signal source is single, accuracy of identification is low and the practicability is poor.
Summary of the invention
In order to solve the deficiencies in the prior art, present disclose provides the fatigue drivings based on multi-information fusion to detect the pre- police
Method, system and medium, with accuracy of identification height, the high effect of practicability；
In a first aspect, present disclose provides the fatigue drivings based on multi-information fusion to detect method for early warning；
Fatigue driving based on multi-information fusion detects method for early warning, comprising:
Extract human eye state feature, pulse characteristics and the steering wheel hyperspin feature of driver；
Fusion Features and tired are carried out to human eye state feature, pulse characteristics and steering wheel hyperspin feature based on SVM-DS algorithm
Labor identification carries out early warning to tired recognition result.
In some possible implementations, further includes: by the high-definition camera installed on automobile rearview mirror, acquisition is driven
The facial image for the person of sailing carries out the identification of driver identity by facial image.The beneficial effect is that if vehicle belongs to taxi
The vehicles in use such as company, it is understood that there may be the replacement class phenomenon of driver can be by driving record reality by the identification of driver identity
When be uploaded to monitoring server and remotely monitored, start convenient for traffic accident caused by the Identity Management of driver and fatigue driving
Thing person's identity is screened.
In some possible implementations, the specific steps of the human eye state feature of driver are acquired are as follows:
Step (101): the face under driver's driving condition is obtained by the high-definition camera installed on automobile rearview mirror
Image；
Step (102): realizing Face detection to Face datection algorithm of the face-image in YCbCr space based on the colour of skin,
Obtain Face detection image；
Step (103): on Face detection image, human eye area positioning is realized according to gray-level projection；
Step (104): according to the human eye area of positioning, eyes is calculated and open duration, eyes closed duration, frequency of wink
With ratio PERCLOS (the Percentage of Eyelid Closure over the Pupil of eyes closed in the unit time
over Time)；
Step (105): eyes are opened into the feature that duration is more than given threshold and are considered as fatigue characteristic；Eyes are opened into duration
It is considered as non-fatigue characteristic less than or equal to the feature of given threshold；
The feature that eyes closed duration is more than given threshold is considered as fatigue characteristic；Eyes closed duration is less than or equal to set
The feature for determining threshold value is considered as non-fatigue characteristic；
The feature that frequency of wink is less than given threshold is considered as fatigue characteristic；Frequency of wink is more than or equal to given threshold
Feature is considered as non-fatigue characteristic；
The ratio PERCLOS of eyes closed in the unit time feature for being less than given threshold is considered as fatigue characteristic；It will be single
The feature that the ratio PERCLOS of eyes closed is greater than given threshold in the time of position is considered as non-fatigue characteristic；
Fatigue characteristic is divided into training set and test set；Non- fatigue characteristic is also divided into training set and test set.
The beneficial effect that high-definition camera is installed on rearview mirror is can to grasp to avoid camera installation site to driver
Make the influence of activity space, and can be blocked to avoid to driver's range of observation.
In some possible implementations, the specific steps of step (103) are as follows:
Step (1031): Face detection image is chosenIt arrivesImage, h be Face detection picture altitude, find out It arrivesImage minimum point bm, taking image ordinate is [bm-30,bm+ 30] region is approximate eye areas,；
Step (1032): pairing approximation eye areas image carries out the calculating of vertical gray-level projection function, to the function
It is normalized, finds out eyeball minimum point coordinate and is denoted as am, obtain [am,bm]；
Step (1033): from the eyeball minimum point coordinate [a of positioningm,bm] pixel that respectively takes 22 to the left and right, 16 are respectively taken up and down
Pixel, and then determine human eye area, histogram equalization then carried out to human eye area image, then carries out binary conversion treatment,
Finally obtain the two dimensional character figure calculated for human eye feature；
Step (1034): the high l of record two dimensional character figureyWith wide lx, the ratio L of Gao Yukuan is as eyes folding angle value；
Step (1035): the eyes opening degree of owner's Vitrea eye area image is normalized to [0,1], the eyes less than 20%
Opening degree is considered human eye and is in closed-eye state, and then calculates eyes and open time, eyes closed time, frequency of wink and list
The ratio PERCLOS of eyes closed in the time of position.
In some possible implementations, the pulse characteristics of driver are acquired:
Pulse data is obtained by the wireless bracelet with photo-electric pulse transducer, based on wavelet transformation to pulse data
It is filtered, extracts pulse characteristics.
The wireless bracelet, comprising: control chip, the control chip respectively with power module, pulse collection module, shake
Dynamic model block, LCD MODULE, bluetooth module are connected with GPS module.The pulse collection module is photo-electric pulse transducer.
The LCD MODULE is liquid crystal display, for showing temporal information.
In some possible implementations, Pulse Rate is obtained by the wireless bracelet with photo-electric pulse transducer
According to being filtered based on wavelet transformation to pulse data, extract the specific steps of pulse characteristics are as follows:
Step (200): signal collected for photo-electric pulse transducer carries out structure solution using db6 small echo, is denoted as ω
(m, n), for the coefficient matrix that noisy acoustical signal obtains after wavelet transformation, m is contraction-expansion factor, and n is shift factor；
Step (201): calculating coefficient R (m, n), and R (m, n) indicates the product of contraction-expansion factor and shift factor；
Step (202): coefficient R (m, n) is normalized:
Wherein, NR (m, n)Correlation matrix after indicating normalization；Z indicates integer.
Step (203): compare NR(m,n)Absolute value and ω (m, n) absolute value size；If NR(m,n)Absolute value it is big,
Then think that ω (m, n) is that ω (m, n) is assigned to reconstruction signal function ω from original signalf(m, n), and ω (m, n) is set
Zero；If the absolute value of ω (m, n) is big, then it is assumed that ω (m, n) comes from noise signal, retains ω (m, n)；
Step (204): noise mean square deviation is calculatedWith unbiased esti-mator σ (m, n) than being λ；
Wherein, L indicates zero setting points；
Step (205): if λ value is greater than 1, the process of iterative step (200) to step (204)；If λ is less than or equal to
1, obtain reconstruction signal ωf(m, n), denoising are completed；
Step (206): main wave crest location is extracted in reconstruction signal:
It selects orthogonal wavelet as wavelet basis, three layers of small wavelength-division is carried out to pulse information using the method for orthogonal wavelet transformation
Solution；The third layer high frequency coefficient after decomposing is extracted, and third layer high-frequency signal is reconstructed with third layer high frequency coefficient；In third layer height
In frequency coefficient, using Adaptive Thresholding, the maximum of points in each periodic regime is detected；The maximum of points conduct that will test
Datum mark, and correspond in original signal；In original signal, front and back respectively takes M point as search range, detects within the scope of this
The maximum of points of original signal, the point are the main wave crest location of pulse；
The time difference between two adjacent main wave wave crests is found out, x is denoted asi, i is positive integer, then xiMean value be main wave wave crest
Mean value, xiStandard deviation be main wave wave crest standard deviation；
Pulse signal is calculated separately into the pulse signal for being converted into frequency domain in high fdrequency component to frequency domain by FFT transform
The power HF of high fdrequency component is obtained pulse compared with both power LF of low frequency component by the power LF of power HF and low frequency component
Low-and high-frequency power ratio；
By the low-and high-frequency power of the main wave crest location of pulse, main wave wave crest mean value, main wave wave crest standard deviation and pulse it
Than being considered as pulse characteristics；
The pulse characteristics that the main wave crest location of pulse is more than given threshold are considered as fatigue characteristic；By the main wave wave of pulse
Peak position is considered as non-fatigue characteristic lower than the pulse characteristics of given threshold；
The pulse characteristics that main wave wave crest mean value is more than given threshold are considered as fatigue characteristic；Main wave wave crest mean value is lower than and is set
The pulse characteristics for determining threshold value are considered as non-fatigue characteristic；
The pulse characteristics that main wave wave crest standard deviation is more than given threshold are considered as fatigue characteristic；Main wave wave crest standard deviation is low
It is considered as non-fatigue characteristic in the pulse characteristics of given threshold；
The pulse characteristics that the low-and high-frequency power ratio of pulse is more than given threshold are considered as fatigue characteristic；By the height of pulse
Frequency power ratio is considered as non-fatigue characteristic lower than the pulse characteristics of given threshold；
Fatigue characteristic is divided into training set and test set；Non- fatigue characteristic is also divided into training set and test set.
In some possible implementations, the steering wheel hyperspin feature of driver is acquired:
Direction disc rotation frequency and rotation angle are obtained by angular sensor；
The feature that direction disc rotation frequency is more than given threshold is considered as fatigue characteristic；Direction disc rotation frequency is lower than and is set
The feature for determining threshold value is considered as non-fatigue characteristic；
The feature that angle is more than given threshold will be rotated and be considered as fatigue characteristic；Feature of the angle lower than given threshold will be rotated
It is considered as non-fatigue characteristic；
Fatigue characteristic is divided into training set and test set；Non- fatigue characteristic is also divided into training set and test set.
In some possible implementations, the specific steps of early warning are carried out to tired recognition result are as follows:
If it is determined that fatigue driving, equipment is by the alerting drivers in such a way that voice and bracelet shake；
It is used if taxi company or fleet, vehicle drive information is uploaded onto the server.
In some possible implementations, based on SVM-DS algorithm to human eye state feature, pulse characteristics and steering wheel
Hyperspin feature carries out the specific steps of Fusion Features and fatigue identification are as follows:
The fatigue characteristic of training set and non-fatigue characteristic are input in support vector machines, support vector machines is instructed
Practice, obtains trained support vector machines；
The feature of test set is input in support vector machines, the posterior probability p of each support vector machines is calculatedi；
The feature of test set is input in support vector machines, the confusion matrix of each support vector machines is obtained；
The local confidence level of corresponding support vector machines is calculated based on each confusion matrix；
When support vector machines is w to the sample class that a certain sample x is identifiediWhen, the result of support vector machines output can
Reliability is PC (wi)；
Posterior probability p based on each support vector machinesiResult credibility with support vector machines output is PC (wi) meter
Calculate BPA when obtaining Decision fusion；
ml(wi)=Pi×PC(wi)
Wherein, ml(wi) presentation class device l belongs to w to sample xiThe probability assignment of class；
It merges to obtain final tired recognition result by DS.
Second aspect, the disclosure additionally provide the detection early warning system of the fatigue driving based on multi-information fusion；
Fatigue driving based on multi-information fusion detects early warning system, comprising:
Data acquisition module: the human eye state feature, pulse characteristics and steering wheel hyperspin feature of driver are acquired；
Identify warning module: based on SVM-DS algorithm to human eye state feature, pulse characteristics and steering wheel hyperspin feature into
Row Fusion Features and fatigue identification, carry out early warning to tired recognition result.
The third aspect, the disclosure additionally provide a kind of electronic equipment, including memory and processor and are stored in storage
The computer instruction run on device and on a processor when the computer instruction is run by processor, is completed first aspect and is appointed
Method in one possible implementation.
Fourth aspect, the disclosure additionally provide a kind of computer readable storage medium, described for storing computer instruction
When computer instruction is executed by processor, either in the completion any possible implementation of first aspect the step of method；
Compared with prior art, the beneficial effect of the disclosure is:
Based on SVM-DS algorithm to human eye state feature, pulse characteristics, steering wheel hyperspin feature and drive duration characteristics into
Row Fusion Features and fatigue identification, carry out early warning to tired recognition result can be improved by the fusion recognition to multiple features
The accuracy of fatigue detecting, and data information acquisition process reduces the interference to driver to greatest extent.
Detailed description of the invention
The accompanying drawings constituting a part of this application is used to provide further understanding of the present application, and the application's shows
Meaning property embodiment and its explanation are not constituted an undue limitation on the present application for explaining the application.
Fig. 1 is the overall flow figure of the disclosure；
Fig. 2 be the disclosure P80 (when eyelid covers pupil area and is more than 80%, the time shared by unit time eyes closed
Than) calculation method schematic diagram；
Fig. 3 is the wireless hand ring moulds block hardware structural diagram of the disclosure；
Fig. 4 is the pulse signal Acquisition Circuit schematic diagram of the disclosure；
Fig. 5 is the pulse collection image schematic diagram of the disclosure；
Fig. 6 is the tired recognizer model schematic of the disclosure.
Specific embodiment
It is noted that following detailed description is all illustrative, it is intended to provide further instruction to the application.Unless another
It indicates, all technical and scientific terms used herein has usual with the application person of an ordinary skill in the technical field
The identical meanings of understanding.
It should be noted that term used herein above is merely to describe specific embodiment, and be not intended to restricted root
According to the illustrative embodiments of the application.As used herein, unless the context clearly indicates otherwise, otherwise singular
Also it is intended to include plural form, additionally, it should be understood that, when in the present specification using term "comprising" and/or " packet
Include " when, indicate existing characteristics, step, operation, device, component and/or their combination.
The disclosure provide it is a kind of can automatic identification driver whether enter fatigue state and carry out the vehicle of early warning to it
Safety device is carried, as shown in figures 1 to 6, the main process of algorithm is divided into the data acquisition of sensor and locates in advance overall flow
It reason, the extraction of feature, the fusion of three kinds of features, fatigue identification and warns, implementation step is as follows:
(1) face-image under driver's driving condition is obtained by high definition camera built in automobile cab
Information；Due to inherently watching front vehicle information by rearview mirror when driver drives, and camera is mounted on rearview mirror
On the interference of driving sight will not be caused to driver, be the optimum position for acquiring driver's video information.
Carry out Face detection.
YCbCr space image, which is converted, by image carries out cluster calculation and skin color segmentation, YmaxAnd YminFor in cluster areas
Maxima and minima, following parameter: Y is had according to test resultmax=153, Ymin=31, Cbmax=128, Cbmin=113,
Crmax=153, Crmin=126.
When Cb and Cr component is within the scope of this in image, it is identified as the potential region of face, otherwise is identified as non-
The potential region of face.By the above flesh tone algorithms to image carry out skin color segmentation, to face location carry out preliminary judgement, then
The potential region of face-image is set as white, and the potential region of non-face image is set as black, will treated image binaryzation.
In order to keep image recognition more accurate, Morphological scale-space is carried out to image.As the basis fortune in morphological operation
It calculates, erosion operation and dilation operation have the function of noise reduction filtering, feature extraction, border detection, Contrast enhanced etc..
Erosion operation can consumed boundary point, can filter off small without the region of meaning.Using structural element B to image
Carry out erosion operation processing, expression formula are as follows:
Dilation operation can make to expand outside outer boundary, merge the region that part is accidentally isolated.Using structural element B to figure
As carrying out dilation operation processing, expression formula are as follows:
Carry out following post-processing:
(1) region of the connected domain boundary rectangle area less than 15 × 15 is filtered off；
(2) human face region boundary rectangle the ratio of width to height is P, and value is not too large will not be too small, by P value in [0.4,1.2]
Region in range filters.
(3) expansive working carried out again to image, obtain comprising facial area there are the bianry images of single connected domain
As exposure mask template.
Carry out human eye positioning.
Gray level image is converted by the human face region after determination, median filtering is recycled to be filtered noise reduction process, is utilized
Histogram equalization operation is adjusted image overall intensity, and image overall intensity is extended in [0,255] range, so that
Each section pixel quantity tends to be identical, and intensity profile more evenly, reduces illumination effect.
If variable I (a, b) is that coordinate is the grayscale information of the point of (a, b) on image, then it is located at [a1,a2]、[b1,b2] model
Horizontal gray-level projection function H (b) and vertical integral projection function V (a) expression formula in enclosing are i.e. are as follows:
In digital picture, discrete expression are as follows:
Image range is contracted to original imageIt arrivesThe range of (h is picture altitude) finds out image minimum point bm,
In order to completely retain the effective information of eye areas, taking image ordinate is [bm-30,bm+ 30] region.Image is carried out vertical
Gray-level projection function calculates, which is normalized, and finding out eyeball minimum point coordinate is am, obtain [am,bm]。
PERCLOS P80 parameter calculates schematic diagram as shown in Fig. 2, calculation formula is as follows:
The definition of P80: when eyelid covers pupil area and is more than 80%, time ratio shared by unit time eyes closed.P is ratio
Example, t1~t4Four time points represented be respectively eyelid cover pupil 80%, eye closing, eye opening, eyelid cover pupil 80% this
Process.
It respectively taking 20 pixel to the left and right from the human eye coordinates of positioning, respectively takes 15 pixel up and down, region includes whole eyes,
Then histogram equalization is carried out, then carries out binary conversion treatment and finally obtains the two dimensional character figure calculated for human eye feature.Note
Record its high lyWith wide lx, ratio L is as eyes folding angle value；All opening degrees are normalized to [0,1], less than 20% folding
Degree is considered at closed state.And then it extracts eyes and opens the spies such as time, eyes closed time, frequency of wink, PERCLOS
Sign.
(2) pulse information is acquired by the wireless bracelet with photo-electric pulse transducer.
This pulse signal acquisition terminal uses STM8S207R8 chip as control chip.Bracelet hardware configuration such as Fig. 3 institute
Show, control chip respectively with power module, pulse collection module, shock module, LCD MODULE, bluetooth module and GPS mould
Block connection.Pulse signal is acquired using photoelectricity heart rate sensor.After AD conversion, signal is transmitted to host by wireless blue tooth.
In addition, terminal has OLED liquid crystal display, for showing the information such as time.Terminal using normal voltage is 3.8V, capacity is
The lithium battery power supply power supply of 250mAH constitutes charging module using AP5056 chip etc..Charging interface is Micro-USB mouthfuls, is filled
Electric current should be between 0.5A-1A.Battery supply makes voltage stabilization after 3.3V by TPS7333 voltage regulator circuit, and supply is internal
Each module uses.
Pulse signal acquisition module is constituted using YK1303 photoelectricity heart rate sensor, HR6706 heart rate chip etc., signal is adopted
It is as shown in Figure 4 to collect schematic diagram.No. 1 pin of photoelectricity heart rate sensor YK1303 passes through resistance R8 connection 3.3V power supply VCC；Photoelectricity
No. 2 pins of heart rate sensor YK1303 are grounded；No. 3 pins of photoelectricity heart rate sensor YK1303 are connected by resistance R10
3.3V power supply VCC；No. 4 pins of photoelectricity heart rate sensor YK1303 connect 3.3V power supply VCC；Photoelectricity heart rate sensor YK1303
No. 5 pins ground connection；No. 6 pins of photoelectricity heart rate sensor YK1303 are grounded by resistance R11；Photoelectricity heart rate sensor
No. 6 pins of YK1303 also pass through resistance R12 and capacitor C18 and connect with 3.3V power supply VCC；Photoelectricity heart rate sensor YK1303's
No. 6 pins also pass sequentially through capacitor C16, resistance R9 and connect with No. 2 pins of HR6706 heart rate chip；HR6706 heart rate chip
No. 1 pin passes sequentially through capacitor C15 and resistance R6 and connect with No. 6 pins of HR6706 heart rate chip；The 2 of HR6706 heart rate chip
Number pin passes sequentially through resistance R18 and resistance R7 and connect with No. 1 pin of HR6706 heart rate chip；The 2 of HR6706 heart rate chip
Number pin is grounded by capacitor C14 and resistance R5；No. 1 pin of HR6706 heart rate chip is grounded by resistance R5；The HR6706 heart
No. 3 pins of rate chip are connect by capacitor C18 with 3.3V power supply VCC；No. 3 pins and the HR6706 heart of HR6706 heart rate chip
No. 5 pins of rate chip connect；No. 4 pins of HR6706 heart rate chip are grounded；No. 6 pins of HR6706 heart rate chip are by simultaneously
The capacitor C17 and resistance R13 of connection are connect with No. 7 pins of HR6706 heart rate chip；No. 7 pins of HR6706 heart rate chip are successively
It is grounded by resistance R16 and resistance R17；No. 8 pins of HR6706 heart rate chip are connect with 3.3V power supply VCC.Resistance R16 and electricity
The tie point between R17 is hindered as signal output point, and bracelet controller STM8 is output in the form of digital signal.
Heart rate sensor is incuded the pulse information of human body and is extracted by the way of photo-electric capacity trace PPG, most
After export pulse wave.HR6707 is designed to a heart rate IC of cooperation YK1303P heart rate sensor, exports pulse wave
MCU is transmitted to be AD converted.Fig. 5 is pulse collection image.
Structure solution is carried out using db6 small echo for collected signal, being denoted as ω (m, n) is noisy acoustical signal through wavelet transformation
Coefficient matrix afterwards.
Coefficient R (m, n) indicates the product of m decomposition scale and the adjacent scale of n, and R (m, n) is normalized:
Compare NR(m,n)With the size of ω (m, n) absolute value, if NR(m,n)Absolute value it is big, then it is assumed that ω (m, n) is to come from
It is assigned to ω by original signalf(m, n) and zero setting, if the absolute value of ω (m, n) to retain since noise signal depending on it greatly
ω(m,n)。
Wherein, L indicates zero setting points；
It obtains noise mean square deviation and unbiased esti-mator ratio is λ, λ value is greater than 1, then iterative step (200)-step (204) process.Most
Reconstruct ω eventuallyf(m, n), denoising are completed.
Main wave position is extracted in reconstruction signal.It is calculated to simplify, selects the preferable orthogonal wavelet of symmetry
Coiflet1 is calculated as wavelet basis using the method for orthogonal wavelet transformation, the specific steps are as follows:
(1) using Coifletl as wavelet basis, three layers of wavelet decomposition are carried out to pulse signal；
(2) the third layer high frequency coefficient after decomposing individually is extracted, and third layer high-frequency signal is reconstructed with this；
(3) in third layer high frequency coefficient, using Adaptive Thresholding, the maximum of points in each periodic regime is detected；
(4) it using the maximum of points detected in step (3) as datum mark, and corresponds in original signal.In original signal
In, front and back respectively takes 100 points as search range, detects the maximum of points of original signal within the scope of this, which is pulse
Main wave crest location.
The time difference between two adjacent main waves is found out, x is denoted asi(i=1,2,3 ..., n), then xiMean value be main wave
Interphase mean value, xiStandard deviation be main wave interphase standard deviation.Pulse signal is calculated separately by FFT transform to frequency domain
, in the power HF and LF of high frequency (0.15~0.4Hz) and low frequency (0.04~0.15Hz) component, the two, which is compared, just obtains pulse for it
Low-and high-frequency power ratio.
(3) direction disc rotation frequency and angle-data are obtained by angular sensor.
When normal driving, the rotational frequency and angle of steering wheel can be maintained in a normal range (NR).Work as driver fatigue
When, it can not concentrate one's energy, reaction speed is slow, and corresponding steering wheel rotational frequency can be decreased obviously, or even will appear low-angle
Urgency beat steering wheel.Direction disc rotation frequency f is recorded by being mounted on the angular sensor at steering wheel shaft, and
Rotate gyration a.
(4) Fusion Features and fatigue identify.
DS composition algorithm based on matrix analysis
N category feature, the phase of m kind target type is assigned with matrix by the case where identifying a target simultaneously for n category feature
Mutual independent basic trusted apportioning cost mijWith uncertain probability θiIt is expressed as
Since same feature assigns the mutually independent basic trusted apportioning cost m of m kind dbjective stateijIt is general with uncertainty
Rate θiThe sum of should be 1, so the sum of the element of every a line of matrix should meet normalizing condition, i.e.,
mi1+mi2+…+mim+θi=1 (i=1,2 ..., n)
It is multiplied to obtain the new matrix R of one (m+1) × (m+1) with the transposition of a line in matrix with another row
Wherein, uncertain factor k is the sum of the off diagonal element of preceding m × m rank submatrix in matrix, i.e.,
1. single feature identification.
On the basis of image preprocessing, three classes intrinsic dimensionality is extracted, three category features are carried out just with support vector machines
Step identification.
2.BPA construction of function.
To " one-to-one " classify SVM fatigue identification in such a way that the probability results that multiple 2 classify are combined into
Row probabilistic Modeling.The probability of pairing class is estimated with sigmoid function, i.e.,
rij≈ p (y=i | y=i or j, x)
For posterior probability piHave:
Then learning sample collection is tested, obtains recognition correct rate qi, then BPA function may be defined as:
mj(A)=piqi
3. Decision fusion and judgment rule.
If Ai(i=1,2,3,4) is driving condition, AωFor dbjective state；
Evidence is obtained to A in frame ΘiReliability and evidence uncertain mjAfter (θ), categorised decision need to follow following rule
Then:
①m(Aw)=max { m (Ai), i.e., the state of maximum reliability is dbjective state.
②m(Aw)-m(Ai)>ε1(ε1> 0), i.e., dbjective state will have to be larger than a certain door with the reliability difference of other states
Limit.
③m(Aw)-m(θ)>ε2(ε2> 0), i.e. the reliability of dbjective state has to be larger than uncertain credits assigned value.
④m(θ)<ε3(ε3> 0), i.e., uncertain credits assigned value is necessarily less than a certain thresholding, that is, to dbjective state
The uncertain reliability of evidence cannot be too big.
(5) driver is warned and data uploads.
If being judged as fatigue driving, MCU sends instructions to bracelet by bluetooth, so that bracelet is generated vibration and is driven with reminding
Member, while MCU also passes through external speaker and issues voice alarm.It is used if taxi company or fleet etc., MCU is by vehicle row
It sails data and differentiates that result is uploaded to server.
The foregoing is merely preferred embodiment of the present application, are not intended to limit this application, for the skill of this field
For art personnel, various changes and changes are possible in this application.Within the spirit and principles of this application, made any to repair
Change, equivalent replacement, improvement etc., should be included within the scope of protection of this application.
Claims (10)
1. the fatigue driving based on multi-information fusion detects method for early warning, characterized in that include:
Extract human eye state feature, pulse characteristics and the steering wheel hyperspin feature of driver；
Fusion Features are carried out to human eye state feature, pulse characteristics and steering wheel hyperspin feature based on SVM-DS algorithm and fatigue is known
Not, early warning is carried out to tired recognition result.
2. the fatigue driving based on multi-information fusion detects method for early warning as described in claim 1, characterized in that acquisition drives
The specific steps of the human eye state feature of member are as follows:
Step (101): the face-image under driver's driving condition is obtained by the high-definition camera installed on automobile rearview mirror；
Step (102): Face detection is realized to Face datection algorithm of the face-image in YCbCr space based on the colour of skin, is obtained
Face detection image；
Step (103): on Face detection image, human eye area positioning is realized according to gray-level projection；
Step (104): according to the human eye area of positioning, eyes is calculated and open duration, eyes closed duration, frequency of wink and list
The ratio PERCLOS of eyes closed in the time of position；
Step (105): eyes are opened into the feature that duration is more than given threshold and are considered as fatigue characteristic；Eyes opening duration is less than
Or it is considered as non-fatigue characteristic equal to the feature of given threshold；
The feature that eyes closed duration is more than given threshold is considered as fatigue characteristic；Eyes closed duration is less than or equal to setting threshold
The feature of value is considered as non-fatigue characteristic；
The feature that frequency of wink is less than given threshold is considered as fatigue characteristic；Frequency of wink is more than or equal to the feature of given threshold
It is considered as non-fatigue characteristic；
The ratio PERCLOS of eyes closed in the unit time feature for being less than given threshold is considered as fatigue characteristic；When by unit
The feature that the ratio PERCLOS of interior eyes closed is greater than given threshold is considered as non-fatigue characteristic；
Fatigue characteristic is divided into training set and test set；Non- fatigue characteristic is also divided into training set and test set.
3. the fatigue driving based on multi-information fusion detects method for early warning as claimed in claim 2, characterized in that step
(103) specific steps are as follows:
Step (1031): Face detection image is chosenIt arrivesImage, h be Face detection picture altitude, find outIt arrivesImage minimum point bm, take image ordinate [bm-30,bm+ 30] region is as approximate eye areas；
Step (1032): pairing approximation eye areas image carries out the calculating of vertical gray-level projection function, carries out to the function
Normalization, finds out eyeball minimum point coordinate and is denoted as am, obtain [am,bm]；
Step (1033): from the eyeball minimum point coordinate [a of positioningm,bm] pixel that respectively takes 22 to the left and right, 16 picture is respectively taken up and down
Element, and then determine human eye area, histogram equalization then is carried out to human eye area image, then carry out binary conversion treatment, finally
Obtain the two dimensional character figure calculated for human eye feature；
Step (1034): the high l of record two dimensional character figureyWith wide lx, the ratio L of Gao Yukuan is as eyes folding angle value；
Step (1035): being normalized to [0,1] for the eyes opening degree of owner's Vitrea eye area image, the eyes folding less than 20%
Degree is considered human eye and be in closed state, and then when calculating eyes opening time, eyes closed time, frequency of wink and unit
The ratio PERCLOS of interior eyes closed.
4. the fatigue driving based on multi-information fusion detects method for early warning as described in claim 1, characterized in that acquisition drives
The pulse characteristics of member:
Pulse data is obtained by the wireless bracelet with photo-electric pulse transducer, pulse data is carried out based on wavelet transformation
Pulse characteristics are extracted in filtering processing.
5. the fatigue driving based on multi-information fusion detects method for early warning as claimed in claim 4, characterized in that by having
The wireless bracelet of photo-electric pulse transducer obtains pulse data, is filtered, is mentioned to pulse data based on wavelet transformation
Take the specific steps of pulse characteristics are as follows:
Step (200): signal collected for photo-electric pulse transducer carries out structure solution using db6 small echo, is denoted as ω (m, n)
For the coefficient matrix that noisy acoustical signal obtains after wavelet transformation, m is contraction-expansion factor, and n is shift factor；
Step (201): calculating coefficient R (m, n), and R (m, n) indicates the product of contraction-expansion factor and shift factor；
Step (202): coefficient R (m, n) is normalized:
Wherein, NR(m,n)Correlation matrix after indicating normalization；Z indicates integer；
Step (203): compare NR(m,n)Absolute value and ω (m, n) absolute value size；If NR(m,n)Absolute value it is big, then recognize
It is that ω (m, n) is assigned to reconstruction signal function ω from original signal for ω (m, n)f(m, n), and by ω (m, n) zero setting；If
The absolute value of ω (m, n) is big, then it is assumed that ω (m, n) comes from noise signal, retains ω (m, n)；
Step (204): noise mean square deviation is calculatedWith unbiased esti-mator σ (m, n) than being λ；
Wherein, L indicates zero setting points；
Step (205): if λ value is greater than 1, iterative step (200)-step (204) process；If λ is less than or equal to 1, obtain
Reconstruction signal ωf(m, n), denoising are completed；
Step (206): main wave crest location is extracted in reconstruction signal:
It selects orthogonal wavelet as wavelet basis, three layers of wavelet decomposition is carried out to pulse information using the method for orthogonal wavelet transformation；
The third layer high frequency coefficient after decomposing is extracted, and third layer high-frequency signal is reconstructed with third layer high frequency coefficient；In third layer high frequency
In coefficient, using Adaptive Thresholding, the maximum of points in each periodic regime is detected；The maximum of points that will test is as base
On schedule, it and corresponds in original signal；In original signal, front and back respectively takes M point as search range, detects former within the scope of this
The maximum of points of beginning signal, the point are the main wave crest location of pulse；
The time difference between two adjacent main wave wave crests is found out, x is denoted asi, i is positive integer, then xiMean value be that main wave wave crest is equal
Value, xiStandard deviation be main wave wave crest standard deviation；
Pulse signal is calculated separately into the power for being converted into the pulse signal of frequency domain in high fdrequency component to frequency domain by FFT transform
The power HF of high fdrequency component is obtained the height of pulse compared with both power LF of low frequency component by the power LF of HF and low frequency component
The ratio between low frequency power；
The low-and high-frequency power ratio of the main wave crest location of pulse, main wave wave crest mean value, main wave wave crest standard deviation and pulse is regarded
For pulse characteristics；
The pulse characteristics that the main wave crest location of pulse is more than given threshold are considered as fatigue characteristic；By the main wave wave crest position of pulse
The pulse characteristics set low in given threshold are considered as non-fatigue characteristic；
The pulse characteristics that main wave wave crest mean value is more than given threshold are considered as fatigue characteristic；By main wave wave crest mean value lower than setting threshold
The pulse characteristics of value are considered as non-fatigue characteristic；
The pulse characteristics that main wave wave crest standard deviation is more than given threshold are considered as fatigue characteristic；Main wave wave crest standard deviation is lower than and is set
The pulse characteristics for determining threshold value are considered as non-fatigue characteristic；
The pulse characteristics that the low-and high-frequency power ratio of pulse is more than given threshold are considered as fatigue characteristic；By the low-and high-frequency function of pulse
The ratio between rate is considered as non-fatigue characteristic lower than the pulse characteristics of given threshold；
Fatigue characteristic is divided into training set and test set；Non- fatigue characteristic is also divided into training set and test set.
6. the fatigue driving based on multi-information fusion detects method for early warning as described in claim 1, characterized in that acquisition drives
The steering wheel hyperspin feature of member:
The speed and rotation angle of steering wheel are obtained by angular sensor；
The feature that direction disc rotation frequency is more than given threshold is considered as fatigue characteristic；By direction disc rotation frequency lower than setting threshold
The feature of value is considered as non-fatigue characteristic；
The feature that angle is more than given threshold will be rotated and be considered as fatigue characteristic；Feature of the angle lower than given threshold will be rotated to be considered as
Non- fatigue characteristic；
Fatigue characteristic is divided into training set and test set；Non- fatigue characteristic is also divided into training set and test set.
7. the fatigue driving based on multi-information fusion detects method for early warning as described in claim 1, characterized in that
Fusion Features are carried out to human eye state feature, pulse characteristics and steering wheel hyperspin feature based on SVM-DS algorithm and fatigue is known
Other specific steps are as follows:
The fatigue characteristic of training set and non-fatigue characteristic are input in support vector machines, support vector machines is trained, is obtained
To trained support vector machines；
The feature of test set is input in support vector machines, the posterior probability P of each support vector machines is calculatedi；
The feature of test set is input in support vector machines, the confusion matrix of each support vector machines is obtained；
The local confidence level of corresponding support vector machines is calculated based on each confusion matrix；
When support vector machines is w to the sample class that a certain sample x is identifiediWhen, the result credibility of support vector machines output is
PC(wi)；
Posterior probability P based on each support vector machinesiResult credibility with support vector machines output is PC (wi) calculate
BPA when to Decision fusion；
ml(wi)=Pi×PC(wi)
Wherein, ml(wi) presentation class device l belongs to w to sample xiThe probability assignment of class；
It merges to obtain final tired recognition result by DS.
8. the fatigue driving based on multi-information fusion detects early warning system, characterized in that include:
Data acquisition module: the human eye state feature, pulse characteristics and steering wheel hyperspin feature of driver are acquired；
It identifies warning module: human eye state feature, pulse characteristics and steering wheel hyperspin feature being carried out based on SVM-DS algorithm special
Sign fusion and fatigue identification, carry out early warning to tired recognition result.
9. a kind of electronic equipment, characterized in that on a memory and on a processor including memory and processor and storage
The computer instruction of operation when the computer instruction is run by processor, completes step described in claim 1-7 either method
Suddenly.
10. a kind of computer readable storage medium, characterized in that for storing computer instruction, the computer instruction is located
When managing device execution, step described in claim 1-7 either method is completed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811325931.3A CN109389806B (en) | 2018-11-08 | 2018-11-08 | Fatigue driving detection early warning method, system and medium based on multi-information fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811325931.3A CN109389806B (en) | 2018-11-08 | 2018-11-08 | Fatigue driving detection early warning method, system and medium based on multi-information fusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109389806A true CN109389806A (en) | 2019-02-26 |
| CN109389806B CN109389806B (en) | 2020-07-24 |
Family
ID=65427099
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811325931.3A Active CN109389806B (en) | 2018-11-08 | 2018-11-08 | Fatigue driving detection early warning method, system and medium based on multi-information fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109389806B (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109823337A (en) * | 2019-02-28 | 2019-05-31 | 重庆交通大学 | The autonomous avoiding system of vehicle and method under a kind of operating passenger car driver abnormal conditions |
| CN110012114A (en) * | 2019-05-05 | 2019-07-12 | 北京市众诚恒祥能源投资管理有限公司 | A kind of Environmental security early warning system based on Internet of Things |
| CN110664417A (en) * | 2019-09-18 | 2020-01-10 | 朔黄铁路发展有限责任公司 | Train safe driving early warning equipment and system |
| CN111062300A (en) * | 2019-12-11 | 2020-04-24 | 深圳市赛梅斯凯科技有限公司 | Driving state detection method, device, equipment and computer readable storage medium |
| CN111080940A (en) * | 2019-11-28 | 2020-04-28 | 同济大学 | Fatigue driving early warning method and system based on threshold system |
| CN111345783A (en) * | 2020-03-26 | 2020-06-30 | 山东大学 | Vestibular dysfunction detection system based on inertial sensor |
| CN111583585A (en) * | 2020-05-26 | 2020-08-25 | 苏州智华汽车电子有限公司 | Information fusion fatigue driving early warning method, system, device and medium |
| CN111797794A (en) * | 2020-07-13 | 2020-10-20 | 中国人民公安大学 | Facial dynamic blood flow distribution detection method |
| CN112233276A (en) * | 2020-10-13 | 2021-01-15 | 重庆科技学院 | Steering wheel corner statistical characteristic fusion method for fatigue state recognition |
| CN113312948A (en) * | 2020-03-26 | 2021-08-27 | 香港生产力促进局 | Method, equipment and system for detecting drowsiness by using deep learning model |
| CN113647955A (en) * | 2021-07-13 | 2021-11-16 | 华东师范大学 | Steering wheel capable of collecting electrocardiogram data, and vital sign monitoring system and method |
| CN114298189A (en) * | 2021-12-20 | 2022-04-08 | 深圳市海清视讯科技有限公司 | Fatigue driving detection method, device, equipment and storage medium |
| CN116502047A (en) * | 2023-05-23 | 2023-07-28 | 成都市第四人民医院 | Method and system for processing biomedical data |
| RU2814302C1 (en) * | 2023-04-04 | 2024-02-28 | Открытое Акционерное Общество "Российские Железные Дороги" | Automated system for continuous monitoring of vigilance of train driver and method for continuously monitoring vigilance of train driver using this system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2011248850A (en) * | 2010-04-28 | 2011-12-08 | Imasen Electric Ind Co Ltd | Doze prevention device |
| CN104952210A (en) * | 2015-05-15 | 2015-09-30 | 南京邮电大学 | Fatigue driving state detecting system and method based on decision-making level data integration |
| CN105261153A (en) * | 2015-11-03 | 2016-01-20 | 北京奇虎科技有限公司 | Vehicle running monitoring method and device |
| CN108216254A (en) * | 2018-01-10 | 2018-06-29 | 山东大学 | The road anger Emotion identification method merged based on face-image with pulse information |
| CN108710756A (en) * | 2018-05-18 | 2018-10-26 | 上海电力学院 | The method for diagnosing faults of lower multicharacteristic information Weighted Fusion is analyzed based on spectral clustering |
| CN108765876A (en) * | 2018-05-31 | 2018-11-06 | 东北大学 | Driving fatigue depth analysis early warning system based on multimode signal and method |
-
2018
- 2018-11-08 CN CN201811325931.3A patent/CN109389806B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2011248850A (en) * | 2010-04-28 | 2011-12-08 | Imasen Electric Ind Co Ltd | Doze prevention device |
| CN104952210A (en) * | 2015-05-15 | 2015-09-30 | 南京邮电大学 | Fatigue driving state detecting system and method based on decision-making level data integration |
| CN105261153A (en) * | 2015-11-03 | 2016-01-20 | 北京奇虎科技有限公司 | Vehicle running monitoring method and device |
| CN108216254A (en) * | 2018-01-10 | 2018-06-29 | 山东大学 | The road anger Emotion identification method merged based on face-image with pulse information |
| CN108710756A (en) * | 2018-05-18 | 2018-10-26 | 上海电力学院 | The method for diagnosing faults of lower multicharacteristic information Weighted Fusion is analyzed based on spectral clustering |
| CN108765876A (en) * | 2018-05-31 | 2018-11-06 | 东北大学 | Driving fatigue depth analysis early warning system based on multimode signal and method |
Non-Patent Citations (1)
| Title |
|---|
| 张万枝等: "驾驶员疲劳检测中的眼睛定位与状态分析", 《重庆大学学报》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109823337A (en) * | 2019-02-28 | 2019-05-31 | 重庆交通大学 | The autonomous avoiding system of vehicle and method under a kind of operating passenger car driver abnormal conditions |
| CN110012114A (en) * | 2019-05-05 | 2019-07-12 | 北京市众诚恒祥能源投资管理有限公司 | A kind of Environmental security early warning system based on Internet of Things |
| CN110664417A (en) * | 2019-09-18 | 2020-01-10 | 朔黄铁路发展有限责任公司 | Train safe driving early warning equipment and system |
| CN111080940A (en) * | 2019-11-28 | 2020-04-28 | 同济大学 | Fatigue driving early warning method and system based on threshold system |
| CN111062300A (en) * | 2019-12-11 | 2020-04-24 | 深圳市赛梅斯凯科技有限公司 | Driving state detection method, device, equipment and computer readable storage medium |
| CN111345783A (en) * | 2020-03-26 | 2020-06-30 | 山东大学 | Vestibular dysfunction detection system based on inertial sensor |
| CN113312948A (en) * | 2020-03-26 | 2021-08-27 | 香港生产力促进局 | Method, equipment and system for detecting drowsiness by using deep learning model |
| CN111583585B (en) * | 2020-05-26 | 2021-12-31 | 苏州智华汽车电子有限公司 | Information fusion fatigue driving early warning method, system, device and medium |
| CN111583585A (en) * | 2020-05-26 | 2020-08-25 | 苏州智华汽车电子有限公司 | Information fusion fatigue driving early warning method, system, device and medium |
| CN111797794A (en) * | 2020-07-13 | 2020-10-20 | 中国人民公安大学 | Facial dynamic blood flow distribution detection method |
| CN112233276A (en) * | 2020-10-13 | 2021-01-15 | 重庆科技学院 | Steering wheel corner statistical characteristic fusion method for fatigue state recognition |
| CN113647955A (en) * | 2021-07-13 | 2021-11-16 | 华东师范大学 | Steering wheel capable of collecting electrocardiogram data, and vital sign monitoring system and method |
| CN114298189A (en) * | 2021-12-20 | 2022-04-08 | 深圳市海清视讯科技有限公司 | Fatigue driving detection method, device, equipment and storage medium |
| RU2814302C1 (en) * | 2023-04-04 | 2024-02-28 | Открытое Акционерное Общество "Российские Железные Дороги" | Automated system for continuous monitoring of vigilance of train driver and method for continuously monitoring vigilance of train driver using this system |
| CN116502047A (en) * | 2023-05-23 | 2023-07-28 | 成都市第四人民医院 | Method and system for processing biomedical data |
| CN116502047B (en) * | 2023-05-23 | 2024-05-07 | 成都市第四人民医院 | Method and system for processing biomedical data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109389806B (en) | 2020-07-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109389806A (en) | Fatigue driving detection method for early warning, system and medium based on multi-information fusion | |
| Ramzan et al. | A survey on state-of-the-art drowsiness detection techniques | |
| CN106530623B (en) | A kind of fatigue driving detection device and detection method | |
| Dasgupta et al. | A vision-based system for monitoring the loss of attention in automotive drivers | |
| CN103839379B (en) | Automobile and driver fatigue early warning detecting method and system for automobile | |
| Fuletra et al. | A survey on drivers drowsiness detection techniques | |
| Hossain et al. | IOT based real-time drowsy driving detection system for the prevention of road accidents | |
| Fatima et al. | Driver fatigue detection using viola jones and principal component analysis | |
| CN108875642A (en) | A kind of method of the driver fatigue detection of multi-index amalgamation | |
| CN103714659B (en) | Fatigue driving identification system based on double-spectrum fusion | |
| Batista | A drowsiness and point of attention monitoring system for driver vigilance | |
| CN104200192A (en) | Driver gaze detection system | |
| CN110728241A (en) | Driver fatigue detection method based on deep learning multi-feature fusion | |
| Flores et al. | Driver drowsiness detection system under infrared illumination for an intelligent vehicle | |
| CN109977930A (en) | Method for detecting fatigue driving and device | |
| Du et al. | A multimodal fusion fatigue driving detection method based on heart rate and PERCLOS | |
| CN104881956A (en) | Fatigue driving early warning system | |
| CN107563346A (en) | One kind realizes that driver fatigue sentences method for distinguishing based on eye image processing | |
| Tang et al. | Real-time image-based driver fatigue detection and monitoring system for monitoring driver vigilance | |
| Devi et al. | Fuzzy based driver fatigue detection | |
| Ursulescu et al. | Driver drowsiness detection based on eye analysis | |
| Zhao et al. | Deep convolutional neural network for drowsy student state detection | |
| CN116503794A (en) | Fatigue detection method for cockpit unit | |
| Ribarić et al. | A neural-network-based system for monitoring driver fatigue | |
| Wathiq et al. | Optimized driver safety through driver fatigue detection methods |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |