International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 11 Issue: 03 | Mar 2024 www.irjet.net p-ISSN: 2395-0072
A Survey on Driver Drowsiness Prediction System Using Machine Learning
Dr Kalli Srinivasa Nageswara Prasad1 , Yatham Zeeharika2 , Vempati Reshmitha3 , Siram Naga Tulasi Sriram4 , Tangella Pujitha Ramya5
1Associate Professor, Dept. of Computer Science and Engineering, Sasi Institute of Technology & Engineering, Tadepalligudem, A.P., India
2,3,4,5 Students of Computer Science and Engineering, Sasi Institute of Technology & Engineering, Tadepalligudem, A.P., India ***
Abstract - The vast majority of accidents nowadays have been caused by driver fatigue for many years. Numerous road collisions are caused by drowsy driving. Despite the development of various sleepy systems over the past decade, existing systems still require improvement in efficiency, accuracy, cost, speed, andavailability. Accidentsbroughtonby fatigue and lack of sleep frequently involve drowsiness. In an effort to lessen these collisions, driver sleepiness detecting devices were developed. The passengers' safetycan beensured using this method in real-time systems.Imagesarecapturedin this using a webcam. Deep learningtechniqueswereemployed to analyze photos and extract information about a driver's facial expressions, eye movements, and head position. A camera records human images, and research is being done to see how that information can be applied to raise driving safety. This method uses a dataset of actual driving situations to show how well it can identify drowsy drivers. Gather photos from a live camera feed, run a machine learning algorithm on the image to check whether or the driver is sleepy or not. The input image is classified as drowsy or not using the machine learning technique. Finally, this study will handle several difficulties at once based on a variety of characteristics and will give a thorough method for anticipating driver fatigue.
Key Words: Driver drowsiness, Accident, Facial Expression, Fatigue detection, Eye and mouth tracking.
1.INTRODUCTION
Due to driver drowsiness many accidents were happening.Driverdrowsinessincreasestheriskofaccidents and collisions, as fatigued drivers experience impaired reaction times and compromised judgment, making them morepronetocollisions.Whendriversaretired,theycan't reactquickly,makebaddecisions,andevenfallasleepbriefly while driving. This can lead to accidents and makes our roadslesssafe.Additionally,sleepydrivingputsthesafetyof other road users, pedestrians, and passengers at risk in addition to the driver. To avoid these issues and maintain everyone's safety on the roadways, it is imperative that everyone obtain enough rest before getting behind the wheel.
Driverdrowsinessdetectionisatermusedtodescribea deviceorsystemthattracksadriver'slevelofalertnessand looksforindicatorsofexhaustionordrowsinesswhilethey aredriving.Thedriver'sbehaviourisoftenstudiedusinga variety of sensors and algorithms, including steering patterns, eye movements, facial expressions, and occasionallyphysiologicalindicationslikeheartrate.These systems' main objective is to warn the driver when they exhibit signs of being too fatigued to operate the vehicle safely,henceloweringthepossibilityofaccidentsbroughton by sleepy driving. Driver Drowsy driving can be just as deadly as drunk driving, hence drowsiness monitoring devicesareacrucialsafetyelement.Theycanwarnfatigued drivers about the need for rest, which can help prevent accidentsandsavelives.
Thecreationofasleepinessdetectionsystemmakesuse of a camera that captures a live video of the driver's face. This technology continuously scans the driver's eyes and facialfeaturesforindicatorsoftiredness.Itspecificallylooks toseeifthemotoristhasopenorclosedeyes.Ifdrowsiness isdetected,thedriverreceivesawarningsignal.Thesystem calculatestheproportionoftimetheeyesremainclosedfora specificduration.Thetechnologydeterminesthatthedriver is dozing off if the cumulative eye closure time exceeds a certainthresholdandsoundsanalarmtowarnthem.
2.RELATED WORK
Existing approaches for detecting driver fatigue frequentlyconcentrateexclusivelyononeortwoaspectsof drivingbehaviour,suchaseyelidclosureorsteeringwheel movements,accordingtoV. Uma Maheswarietal.[1].For detectingdriverdrowsiness,anumberoftechnologiesare used,suchasPERCLOS,speechprocessingdata,andlinear regression. While these techniques have shown some promiseintermsofidentifyingsleepiness,theirapplicability and accuracy are frequently constrained. The suggested methodanalysesseveralcharacteristicsofdriverbehaviour andenvironmentusingimageprocessingtechniques.There are ways to use hybrid machine learning to identify and detect driver tiredness at an early stage, according to a report[2].Theproposedprocessisdiscussed,whichcallsfor acameratorecordthedriver'sfootageandseparateitinto
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 11 Issue: 03 | Mar 2024 www.irjet.net p-ISSN: 2395-0072
pictures.Fourstepsmakeuptheprocess:preparation,eyestate assessment, warning, and warning stadium. The technology described in the file uses a camera to record footage of the driver and divide it into images, which are then analysed using machine learning algorithms to find indicatorsofintoxication.
ThepaperwaswrittenbyHananeLamaaziandothers[3], It provides an overview of the literature on the subject, outlining the datasets used, the features deduced, and the classification algorithms used. Additionally, it emphasizes theresearchgapsanddifficultiesinthisarea,includingthe demand for more realistic and varied datasets, the incorporationofnumeroussensors,andthecreationofinthe-moment and edge-based solutions. The proposed frameworkemploysaconvolutionalneuralnetwork(CNN) toanalyzemulti-sensordatafromasmartphonemountedon a car's dashboard. A sizable dataset of labelled driving sessionsthatincludesbothaggressiveandpassivedriving behavioursisused totraintheCNN model. Thesuggested methodology uses a sleepiness dataset that is processed throughanumberofprocessestodetectdrivertirednessand preventaccidents,accordingtoasurveybyV.VijayPriyaet al. [4]. The Multi-Scale Convolutional Neural Network (MCNN) is utilized for video conversion, facial feature detection, pre-processing, feature optimization, and classification The method leverages the Flamingo Search Algorithm (FSA), a revolutionary method for feature optimization in sleepiness detection, to optimize the extractedfeatures.Auniquemethodforfeatureextractionin sleepinessdetectionisusedinthemethod,whichcombines the Walsh-Hadamard transform with a hybrid dual-tree complexwavelettransformtoextractfeaturesfromthepreprocessedimages.
A real-time driver drowsiness monitoring system called Dri Care employs facial features to detect the new facetrackingalgorithm,andstudy[5]claimsthatthesuggested strategyusesadetectionmethodbasedon68keypointsto determinedriver'sstate.DriCarecanadvisethemotoristof theirsleepinessbycombiningtheirmouthandeyefeatures. Thesuggestedapproachisarealisticandeconomicalmethod for reducing traffic accidents since it can identify driver tirednesswithouttheneedofanygadgets.
The paper'sauthorsareYasarBeceriklietal.[6],a realtime driver fatigue detection system that tracks driver exhaustionbasedoneyeandmouthfeaturesusingaMultiTask CNNmodel.The methodattemptstomaketheroads saferandavoidaccidentsbroughtonbydriverweariness. Usingeyeandmouthfeaturestotrackdrivingbehaviour,the Multi-TaskCNNmodelcanidentifydriverweariness.There are three categories for fatigue levels: highly fatigued, beginning to feel fatigued, and normal. For the purpose of determiningfatigueparameters,thesystemistrainedusing Multi-TaskCNNmodels.
A survey by Z. Halim et al [7] suggests that a proposed system includes a feature selection technique for profile prediction, unsupervised learning clustering methods for pattern extraction from labelled data, and a specially designed hardwaresystem forsituation creation and data recording. The system employs a deep neural network to learndriver'sLTVdrivingstylepatternsandidentifyrisky behaviors. It's evaluated on real-world data and demonstrateditseffectivenessinidentifyingrisks.Thepaper comparesitwithrelatedworksanddiscussesitspotential applicationsincriticalinfrastructureprotection.
The survey [8] reveals a method using a multi-task convolutional neural network (MTCNN) to extract facial features, including distance between eyebrows and chin, eyelidandlipaspectratios,andlipheightratios.Themethod also considers six other features in calculating change curves,includingeyeaspectratio,mouthaspectratio,nose aspectratio,eyeheightratio,mouthheightratio,andnose height ratio. The paper introduces a new method for detectingdriverfatigueusingdeeplearningtechniques.The methodoutperformstraditionalfatiguedetectionalgorithms. Thepaperalsointroducestheexperimentalenvironmentof drivinganddrivingsimulationplatforms,andpresentsdata fromthreetypicalsubjects.
Fengyouandothers[9],Themethodoutlinesareal-time algorithm for detecting driver weariness using a deep learning model, considering personal variations and facial landmarkextraction.Theprogramcomprisestwomodules: offline training and online monitoring. On both publicly available and custom-built datasets, the suggested methodologywasputtothetestandfoundtobereasonable andmoreaccuratethancompetingapproaches.Futurestudy directions are also mentioned in the method, such as investigating multi-feature fusion techniques and doing nighttime driving sleepiness detection studies. The GuangdongNatural Science Foundationandotherfunding helpedtofundtheproject.
FedericoGuede-Fernandezetal.'sstudy[10]indicatesthat thismodelsuggestsafreshapproachforidentifyingdriver tirednessbasedonvariationsintheirrespiratorysignal.The system examines the respiratory rate variability (RRV) to identifysleepinessindrivers.Theproposedmodelexplains the steps involved in data gathering, the suggested algorithm,andalgorithmvalidation.Thetechniqueendsby discussingtheresultsofthisstudyandthepossibilitiesfor enhanced driver assistance systems to reduce accidents broughtonbyfatigueddriving.
AccordingtoSeyedKianMousavikiaetal.[11],Thestudy proposes a driver drowsiness detection system using a modified RiscV processor on an FPGA, which achieved an 81.07% accuracy using a Convolutional Neural Network trained to classify four primary driver expressions. The proposed model describes the hardware implementation, hardwareenhancements,CNNsoftwarearchitecture,dataset
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 11 Issue: 03 | Mar 2024 www.irjet.net p-ISSN: 2395-0072
for system training and validation, and hardware implementationandhardwareoptimizations.Themodel's outcomesandaccomplishmentsarehighlighted,alongwith possibleusesforthisdriverdrowsinessdetectionsystem.
Acutting-edgetechniquefordetectingdrivingtiredness while protecting the drivers' privacy was developed by LinlinZhangetal.[12].Thesuggestedmethodusestransfer learning and federated learning to increase drowsiness detectionaccuracywhileloweringcommunicationcosts.The confidentialityofthedrivers'dataisprotectedviaaCKKSbased privacy-preserving mechanism. A flowchart of the suggested PFTL-DDD technique is included, as well as a thread model of the federated learning system. The suggestedapproachalsoofferscitationstorelevantstudies onfederatedlearninganditsapplications.
RiadAlharbey et al.[13],Thisarticleisaboutastudyon detectingfatigueinpeoplewhentheyaredriving.Regardless ofthedatasetutilizedinpriorstudies,thestudy'sgoalisto show the superiority of the suggested methodologies for detectingdriverwearinessanddrowsiness.Alongwiththe suggestedways,thestudycomparesanumberofalgorithms, includingthosedevelopedbysomeothers.Thestudyuses EEG-basedandvideostreamingalgorithmstoidentifydriver weariness,comparingtheproposeddeeplearningapproach withvarious algorithms. Thestudyalsocoversthetypical indicationsofdriverwearinessanddrowsiness.
This is a system for forecasting driver fatigue levels using faceandheadbehaviourinformation,accordingtoHaiderA. Kassem et al. [14]. The study employed a dataset from NationalTsuingHuaUniversitythatwasmadeavailableto thepublicandincluded36participantsfromvariousethnic backgrounds,includingmenandwomen,whowererecorded in a variety of simulated driving situations, including barefaced, night-bareface, sunglasses, glasses, and nightglasses.Ineachscenario,theparticipantswererecordedin boththeirasleepandawakephases,yieldingatotalof200 movies.
The study discovered that eye, mouth, and head movement features can predict fatigue levels determined fromsubjectiverateswithexcellentaccuracy.CNNmodels canforecastthefeaturesoftheeyes,mouth,andheadusing faciallandmarksasinputfeatures.Eachfeature'sprediction accuracyinthestudywasgreaterthan97%.Therefore,there isalotofroomforadvancementintheuseoflow-cost,nonintrusivein-carcamerasforthemonitoringandalertingof driver weariness. Overall, employing non-physical touch sensorinput,thestudyoffersalow-costframeworkforearly driverfatiguedetection.Theexperiment'sfindingsindicate that forecasting driver fatigue levels can be done with reasonableaccuracy.
Thisisavision-basedreal-timemeasurementsystemfor trackingdriverweariness,accordingtoYin-Chengetal.[15]. The technique suggests using a remote
photoplethysmography (rPPG) signal to detect driver wearinesswithoutmakinganyphysicaltouch.Thismethod enhancesaccuracyandgets ridofthe nuisanceassociated with conventional contact-based physiological tiredness detection systems by monitoring both motional and physiological data using a single image sensor. The model emphasizestheimportanceofdetectingdrowsydrivingto reduce accident costs. A remote photoplethysmography (rPPG) signal has several advantages over conventional contact-basedphysiologicaltirednessmonitoringsystems, which are covered in the model as well. The system takes intoaccountenvironmentalnoisethat,inactualapplications, mayhaveanimpactonthemeasuredsignal.
The method provides a thorough analysis of earlier research on driver state identification, including intrusive andnon-intrusivesystems,accordingtoKhadidjaetal.[16]. The authors suggest a fresh approach to feature selection thatcombinesvisualandsignal-basedsensors,andtheygive experimental evidence to support it. Using a dataset of drivingdatagatheredfromvariousdrivers,theapproachis assessed.Themodelcomestoaclosewithadiscussionofthe findingsandpotentialnextlinesofinquiry.
The model analysis focuses on the development of advanced driver-assistance systems and intelligent safety features, specifically yawn detection and seatbelt state detection, according to Paul Kielty et al. [17]. The authors explainhowtoaltereventdatatomeetavarietyofneedsfor various activities, and they also provide neuromorphic event-based algorithms for identifying these traits. The technique outlines the network construction, gathering of videodatasets,creationoffabricatedevents,andspecialized event data preprocessing for various purposes. The suggested model also discusses relevant literature and contrastsitsfindingswiththoseofotherresearchersinthe area. The technique comes to a close with the authors' overallfindingsandtheirramifications.
Thistechniqueproposesanewalgorithmforidentifying fatigueddrivingbasedonthemergingofmanyfacefeatures, asproposedbyKeningLietal.[18].Thealgorithmhasthree modules: driver mouth state classifier, driving fatigue assessment, and identity entry. The experimental analysis evaluates the algorithm's accuracy and real-time performance.Theconclusionsummarizesthepaper'smain points, analyzes the system's flaws, and suggests future optimization paths and algorithmic prospects. The background of the system and its research relevance are presentedintheintroduction,alongwiththecurrentstateof domestic and international fatigue driving detection research.
By Anna Li et al. [19], A model representation augmentation module was developed to enhance adaptability to multi-scale features, improve channel attention strategy, and enhance spatial encoding abilities, whilepromotingaGPS-basedcollaborativedecision-making
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 11 Issue: 03 | Mar 2024 www.irjet.net p-ISSN: 2395-0072
approach between humans and machines. The technique usesanactor-critic-basedreinforcementlearningalgorithm andbuildsanenhancedactornetworkusinganiLQR-based GPS. The idea proposes an adaptable module that adjusts vehicle decisions based on driver fatigue, utilizing experimentalresultsfromMujoco'sreinforcementlearning platformandanenhancedguidedpolicysearchalgorithm. The model analyses the experimental data and rates the efficiencyofthesuggestedapproach.
Thestudyprovidesamethodforassessingdriverfatigue through the examination of intelligent facial expressions, according to Sajid Ali Khan et al. [20]. Entropy analysis is usedbytheframeworktoselectinformativechunkswithina picture. DWT generates four sub-band pictures from the inputimage,followedbydividingtheLLimageintoeight8blockblocksandmulti-scalingsimultaneously.
3.RESULT ANALYSIS
TABLE I: Thetableshowstheresultsofastudyonthe accuracyandchallengesofthedifferentmethods.
SNO MODEL PARAMETERS CHALLENGES
1. CNN EAR,MAR,FAR
Singlecamerais tocapture images
2. SVMand Image Processing Clustering PERCLOS,Eye status Difficultyin accurately detecting drowsiness
3. Random Forest,SVM, SE,FCN
4. MCNN,FSA
Camera,Frame Rate Resolution, Numberof frames Variabilityin lighting conditionsand cameraquality
1000video sequencesof driverswith andwithout drowsiness
5. CNN, MTCNN
Learningrate, Regularization parameter,and kerneltype
6. MTCNN PERCLOS, Frequencyof mouthopening (FOM)
Toimprovethe accuracyand efficiency
Needfora robustface tracking algorithm
Needforalarge datasetfor detectingthe face
7. NaïveBayes, KNN,SVM Crashdata, Trafficdata, Weatherdata Gathering additionaldata formore accuracy
8. MTCNN EAR,MAR, NAR Highquality camera
9. CNN,SVM EAR,PERCLOS Problemin detectingthe faceregion
10. TEDD WLR,WLQ, NBC,QuaTh
11. CNN
Rangeof Gaussiannoise, Numberof layers
12. SVM Eyeblinking, EAR,MAR
Noprior approachesfor respiratory signalanalysis
Datasetusedfor trainingand validationis small
Privacyofthe driver’sdataisa majorconcern
13. SVM,CNN Head Movement, Framerate Resolution Systemcan process largeamountof data
14. CNN Eyeclosure duration, mouthopening duration Accuracyofeye position
15. MRSPT HRT,PRT
16. SVM, DBSCAN Head movementand eyetracking
17. CNN Drivereyes, Yawnratio
Limiteddataset size
Datasetused wasrelatively small
Difficultto identifyaccurate blinkdetection
18. algorithm onmerging ofmultiple facefeature EyeFeature Vectorand MouthFeature Vector Differentaspects areusedto detecteyesand mouth
19. DDPG Real-lifevideo data Difficultyin detecting accuratedriver
SVM=Support Vector Machine, CNN=Convolutional Neural Networks, MTCNN=Multi-Task Cascaded Convolutional NeuralNetwork,FSA=FlamingoSearchAlgorithm,KNN=KNearestNeighbor,FCN=FullyConvolutionalNetwork
Volume: 11 Issue: 03 | Mar 2024 www.irjet.net
Table-2: ComparisonAnalysis
A.Accuracy
TheaccuracyformulacomprisesTruePositives(correctly identified positives), True Negatives (correctly identified negatives), False Positives (incorrectly classified as positives), and False Negatives (incorrectly classified as negatives).Itgaugestheoverallcorrectnessinclassification or data evaluation by calculating the ratio of correct classificationstothetotalnumberofinstances.
B.Precision
Precisioninthecontextofmeasurementordataanalysis representsthelevelofexactnessandaccuracyinobtaining results.Itquantifieshowcloselyindividualmeasurementsor datapointsmatcheachother.Ahighprecisionindicateslow variability and consistency in data, while low precision suggestsgreatervariabilityandpotentialerror.
Intheformula,TruePositives,whicharethecorrectpositive predictions made. False Positives, which are incorrect positivepredictions.Precisionmeasurestheproportionof truepositivepredictionsrelativetoallpositivepredictions (truepositivesandfalsepositives),providingametricforthe accuracyofpositivepredictions.
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 11 Issue: 03 | Mar 2024 www.irjet.net p-ISSN: 2395-0072
C.Recall
Recall, whether in information retrieval or machine learning,assessesthesystem'scapabilitytofindandinclude all relevant items from a dataset, essentially gauging its abilitytoavoidmissingimportantresults.Itisexpressedas theproportionofrelevantitemsretrievedcomparedtothe totalnumberofrelevantitems.
D.F1Score
F1 score is a metric that combines precision and recall evaluatingtheaccuracyofaclassificationmodel.Itprovides abalancebetweenthesetwomeasures,makingitusefulfor assessingamodel'soverallperformanceintaskslikebinary classification.
4.CHALLENGES AND GAPS
The method for anticipating driver drowsiness's limitations.Thedynamicvarietyingesturesmadebyvarious peopleisoneoftheconstraints.Theauthorscontendthatas aresult,thesystem'saccuracymaysuffersinceitmightnot beabletoreliablyidentifytirednessinallusers.
Anotherdrawback isthatthemodel canonlybetrained usingimagestakenwithasinglecamerathatispositionedon acar'sdashboard.Accordingtothescientists,thismightnot be enough to account for all the differences in driver behaviour,whichwouldreducethesystem'saccuracy.
The scientists also pointoutthatthe properties derived fromthephotosmaynotbepertinentinothercontexts,such as an office or home environment, making the suggested method unsuitable for identifying drowsiness in those settings.
Finally,theypoint outthatbecausepoorlightingorbad weathermayinfluencethequalityofthephotostaken,the suggested method may not be appropriate for identifying tirednessinalldrivingsituations.
Overall,theyacceptthattheirsuggestedmethodhaslimits andrecommendthatfuturestudyconcentrateonaddressing theseconstraintstoincreasethesystem'saccuracy.
5.CONCLUSION
The conclusion of the research is that the proposed strategy for detecting driver drowsiness based on many criteriaandusingimageprocessingtechniqueshasshown promising resultsin terms ofaccuracyand efficiency. The
suggestedmethodcanbeutilizedinreal-timesystems,such self-driving automobiles, to stop accidents brought on by driver inattention. Future research can concentrate on enhancingthesystem'sperformanceundervariouslighting situationsandwithvariouscamerastances.
REFERENCES
[1]Maheswari, V. Uma, Rajanikanth Aluvalu, MVV Prasad Kantipudi,KrishnaKeerthiChennam,KetanKotecha,and Jatinderkumar R. Saini. "Driver drowsiness prediction based on multiple aspects using image processing techniques." IEEE Access 10(2022):54980-54990.
[2]Altameem, Ayman, Ankit Kumar, Ramesh Chandra Poonia, Sandeep Kumar, and Abdul Khader Jilani Saudagar. "Early identification and detection of driver drowsinessbyhybridmachinelearning." IEEE Access 9 (2021):162805-162819.
[3]Lamaazi, Hanane, Aisha Alqassab, Ruba Ali Fadul, and Rabeb Mizouni. "Smart Edge-Based Driver Drowsiness Detection in Mobile Crowdsourcing." IEEE Access 11 (2023):21863-21872.
[4]Vijaypriya, V., and M. Uma. "Facial Feature-Based Drowsiness Detection with Multi-Scale Convolutional NeuralNetwork."IEEEAccess(2023).
[5]Deng, Wanghua, and Ruoxue Wu. "Real-time driverdrowsinessdetectionsystemusingfacialfeatures."Ieee Access7(2019):118727-118738.
[6]Savaş,BurcuKir,andYaşarBecerikli."Realtimedriver fatigue detection system based on multi-task ConNN." IeeeAccess8(2020):12491-12498.
[7]Halim,Zahid,MuhammadSulaiman,MuhammadWaqas, and Doğan Aydın. "Deep neural network-based identificationofdrivingriskutilizingdriverdependent vehicle driving features: A scheme for critical infrastructure protection." Journal of Ambient IntelligenceandHumanizedComputing14,no.9(2023): 11747-11765.
[8]Li,Dan,XinZhang,XiaofanLiu,ZhichengMa,andBaolong Zhang."Driverfatiguedetectionbasedoncomprehensive facialfeaturesandgatedrecurrentunit."JournalofRealTimeImageProcessing20,no.2(2023):19.
[9]You,Feng,XiaolongLi,YunboGong,HailweiWang,and Hongyi Li. "A real-time driving drowsiness detection algorithmwithindividualdifferencesconsideration."Ieee Access7(2019):179396-179408.
[10] Guede-Fernandez, Federico, Mireya FernandezChimeno, Juan Ramos-Castro, and Miguel A. Garcia-
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 11 Issue: 03 | Mar 2024 www.irjet.net p-ISSN: 2395-0072
Gonzalez. "Driver drowsiness detection based on respiratory signal analysis." IEEE access 7 (2019): 81826-81838.
[11] Mousavikia, Seyed Kian, Erfan Gholizadehazari, Morteza Mousazadeh, and Siddika Berna Ors Yalcin. "InstructionSetExtensionofaRiscVBasedSoCforDriver DrowsinessDetection."IEEEAccess10(2022):5815158162.
[12] Zhang,Linlin,HideoSaito,LiangYang,andJiajieWu. "Privacy-preserving federated transfer learning for driver drowsiness detection." IEEE Access 10 (2022): 80565-80574.
[13] Alharbey, Riad, Mohamed M. Dessouky, Ahmed Sedik,AliI.Siam,andMohamedA.Elaskily."Fatiguestate detection for tired persons in presence of driving periods."IEEEAccess10(2022):79403-79418.
[14] Kassem,HaiderA.,MorshedChowdhury,andJemal H.Abawajy."Driversfatiguelevelpredictionusingfacial, andheadbehaviorinformation."IEEEAccess9(2021): 121686-121697.
[15] Tsai,Yin-Cheng,Peng-WenLai,Po-WeiHuang,TzuMin Lin, and Bing-Fei Wu. "Vision-based instant measurementsystemfordriverfatiguemonitoring."IEEE Access8(2020):67342-67353.
[16] Henni, Khadidja, Neila Mezghani, Charles GouinVallerand,PerrineRuer,YoussefOuakrim,andÉvelyne Vallières. "Feature selection for driving fatigue characterizationanddetectionusingvisual-andsignalbasedsensors."InAppliedInformatics,vol.5,no.1,pp.115.SpringerOpen,2018.
[17] Kielty, Paul, Mehdi Sefidgar Dilmaghani, Waseem Shariff, Cian Ryan, Joe Lemley, and Peter Corcoran. "NeuromorphicDriverMonitoringSystems:AProof-ofConceptforYawnDetectionandSeatbeltStateDetection usinganEventCamera."IEEEAccess(2023).
[18] Li,Kening,YunboGong,andZiliangRen."Afatigue drivingdetectionalgorithmbasedonfacialmulti-feature fusion."IEEEAccess8(2020):101244-101259.
[19] Li,Anna,XinnanMa,JiaxinGuo,JingyueZhang,Jing Wang,KaiZhao,andYaochenLi."Driverfatiguedetection and human-machine cooperative decision-making for road scenarios." Multimedia Tools and Applications (2023):1-32.
2024, IRJET | Impact Factor value: 8.226 | ISO