Driver weariness is frequently one of the prima causes of traffic accidents. In this concluding twelvemonth undertaking, a computing machine vision attack which exploits the driver ‘s facial look is considered, utilizing a combination of the Viola-Jones face sensing technique and support vector machines to sort facial visual aspect and find the degree of weariness.
Section 1: Description
Statisticss show that driver weariness is frequently one of the prima causes of traffic accidents. Over the past few old ages, a batch of research and attempt has been put forth in planing systems that monitor both driver and driving public presentation. A computing machine vision attack which exploits the driver ‘s facial look is considered in this concluding twelvemonth undertaking. The Viola-Jones real-time object sensing model working on a boosted cascade of Haar ripple characteristics is adopted for face sensing. To find the degree of weariness, multiple characteristic categorization is so performed utilizing support vector machines. The motives for taking to develop the system in this mode are the rapid face sensing times coupled with the simple and inexpensive overall execution, avoiding the demand to put in expensive and complex hardware.
Concise Literature Review
This subdivision gives a wide reappraisal of the literary work related to face sensing in fatigue monitoring systems and engineerings, concentrating peculiarly on what has been done in the field of driver weariness. In subdivision 1.2.1, several statistics of fatigue-related motor vehicle accidents are mentioned and analysed. Section 1.2.2 high spots some of the more successful systems ( both commercial and non-commercial ) that have been implemented in recent old ages. On the other manus, subdivision 1.2.3 nowadayss an enlightening overview of the algorithms and techniques typically used in the development of such systems, particularly those refering to both face and facial characteristic sensing. Representative plants for each of these methods will be included.
Statisticss Related to Driver Fatigue
Driver weariness has been one of the chief causes of route accidents and human deaths in recent old ages, and in this subdivision an effort is made to foreground some of the more of import statistics that demonstrate this negative tendency.
The National Highway Traffic Safety Administration ( NHTSA ) [ 1 ] estimations that 2-23 % of all vehicle clangs can be attributed to driver weariness. Every twelvemonth, around 100,000 traffic accidents and 71,000 hurts related to driver sleepiness are reported in the United States, out of which more than 1,300 are fatal [ 2 ] . The NHTSA [ 3 ] besides reports that in the twelvemonth 2005 entirely, there were about 5,000 route human deaths ( around 8.4 % ) which were caused either by driver inattention ( 5.8 % ) or sleepy and fatigued drive ( 2.6 % ) . Furthermore, 28 % of fatal traffic accidents were due to lane maintaining failure, one of the indirect effects of weariness on drivers, ensuing in the loss of 16,000 lives. Undoubtedly, truck drivers are more capable to tire chiefly because of the long hours travelled on main roads, taking to inevitable humdrum journeys. In fact, a survey by the U.S. National Transportation Safety Board ( NTSB ) [ 4 ] confirmed that weariness was the finding factor in 51 out of 87 instances of truck accidents.
These dismaying statistics pointed to the demand to plan and implement systems capable of tracking and analyzing a driver ‘s facial features or organic structure provinces and giving a warning signal at the first noticeable marks of weariness to seek and forestall the likely happening of an accident. In the following subdivision of this literature reappraisal, a figure of these systems will be presented.
Existing Fatigue Monitoring Systems
Many different attacks for systems undertaking the job of driver fatigue have been studied and implemented over the past few old ages. Earlier devices tended to be instead intrusive, necessitating physical contact to mensurate fatigue characteristics while driving. These characteristics included bosom rate variableness, analysis of encephalon signals every bit good as the driver ‘s physiological province. Other systems studied the relation of driver sleepiness to maneuvering clasp and vehicle motions, with some besides using lane tracking installations. However, the focal point nowadays is more towards independent non-intrusive systems that work in the background without deflecting the driver in any manner, able to observe and track caput and oculus motions by agencies of one or more cameras mounted on the vehicle ‘s splashboard. The bulk of merchandises tracking weariness have been designed for on-road vehicles, such as autos, trucks and engines, and these will be reviewed in the undermentioned subdivision. In Section 184.108.40.206, other types of weariness monitoring systems that have been deployed will be analysed.
On-Road Fatigue Monitoring Systems
Commercially Implemented Systems
In the system presented by Advanced Brain Monitoring Inc. [ 5 ] , a caput mounted device in the signifier of a baseball cap uses the encephalon ‘s EEG ( Electroencephalography ) signals to mensurate weariness. Two electrodes inside the baseball cap are connected to the driver ‘s scalp to capture these signals, directing them via wireless moving ridges to a processing device 20 pess off from the driver. Russian seller Neurocom marketed the Engine Driver Vigilance Telemetric Control System ( EDVTCS ) [ 6 ] for usage within the Russian railroad system. EDVTCS continuously track drivers ‘ physiological province by mensurating alterations in the electro cuticular activity ( EDA ) i.e. alterations in the tegument ‘s opposition to electricity based on the eccrine perspiration secretory organs of the human organic structure, located chiefly on the thenar of our custodies and the colloidal suspensions of our pess.
One of the first non-intrusive driver weariness supervising systems was ASTiD ( Advisory System for Tired Drivers ) [ 7 ] . It consists of an up-to-date knowledge-base theoretical account exposing a 24-hour anticipation form sing the possibility of the driver traveling to kip piece at the wheel, and a guidance wheel detector system capable of placing humdrum driving intervals, such as those in main roads, every bit good as unusual maneuvering motions as a consequence of driver weariness. Lane trailing is another attack taken to place distraction forms while driving. SafeTRAC, by AssistWare Technology [ 8 ] , consists of a picture camera located on the windscreen of the vehicle ( confronting the route ) and a splashboard mounted having device to which it is connected. The camera is able to observe lane markers in roads and issues hearable, ocular or haptic warnings if fickle drive forms, such as changeless impetuss between lanes, are observed.
Sing the issues encountered in earlier systems, more importance now started being given to systems that monitored driver head motions, face and facial characteristics. MINDS ( MicroNod Detection System ) , described in [ 9 ] , paths head place and motion, with caput nodding being the chief weariness characteristic used for observing micro-sleep ( short periods of distraction ) while driving. Head motion is tracked by an array of three capacitance detectors located merely above the driver ‘s cockpit. Yet another attack was taken by David Dinges and Richard Grace [ 10 ] at the Carnegie Mellon Research Institute ( CMRI ) in the development of the PERCLOS proctor, which determines the oculus closing per centum over clip for fatigue sensing. In [ 11 ] , PERCLOS is defined as the proportion of clip the eyes are closed 80 % or more for a specified clip interval. FaceLAB [ 12 ] focal points on both face and oculus trailing, mensurating PERCLOS ( PERcentage of oculus CLOSure over clip ) and analyzing water chickweeds in existent clip ( including wink frequence and wink continuance ) . A important difference from other systems is that the absolute place of the eyelid, instead than the occlusion of the student, is used to mensurate oculus closing, doing it much more accurate.
The 2001 AWAKE undertaking of the European Union [ 13 ] focused specifically on driver weariness, integrating many of the above mentioned steps. The chief end of this undertaking, ( its acronym standing for System for effectual Assessment of driver watchfulness and Warning Harmonizing to traffic hazard Estimation ) , was to supply research on the real-time, non-intrusive monitoring of the driver ‘s current province and driving public presentation. Many spouses were involved in AWAKE, including developers, makers and providers of electronics, research institutes, universities, auto makers and terminal users. The undertaking ‘s initial ends were those of accomplishing over 90 % dependability, a lower than 1 % false dismay rate and a user credence rate transcending 70 % .
Car fabrication companies, such as Toyota, Nissan and DaimlerChrysler [ 9 ] are besides in the procedure of developing their ain weariness supervising systems.
Research Based Systems
Many research documents closely related to driver fatigue monitoring have been published in recent old ages. Assorted attacks have been proposed, among which skin coloring material information has been really popular. Smith [ 14 ] nowadayss a system based on skin coloring material predicates to find weariness from oculus wink rate and caput rotary motion information. Similarly, in the gaze way monitoring system proposed by Wahlstrom et Al. [ 15 ] , coloring material predicates are used to turn up the lip part by finding those pels that match the needed coloring material values. Face extraction by skin coloring material cleavage utilizing the normalized RGB skin coloring material theoretical account is adopted in both [ 16 ] and [ 17 ] . Veeraraghavan and Papanikolopoulos [ 16 ] developed a system to observe forms of micro-sleep by continuously tracking the driver ‘s eyes. PERCLOS is the fatigue characteristic measured in Aryuanto and Limpraptono ‘s system [ 17 ] . Horng and Chen [ 18 ] attempted to utilize the HSI coloring material theoretical account to take the consequence of brightness from the image.
Machine acquisition is another common attack to tire sensing. Yang et Al. [ 19 ] choose to follow a Bayesian Network based “ probabilistic model ” to find the fatigue degree. A Bayesian Network theoretical account is besides constructed in [ 20 ] , where Zhu and Lan track multiple ocular cues, including caput and oculus motions and facial looks via two cameras, one for the face and the other concentrating specifically on the eyes, every bit good as Infra-Red illuminators to illume up the needed countries of the face.
A nervous web attack is adopted by D’Orazio et Al. [ 21 ] and RibariA‡ et Al. [ 22 ] in their proposed systems. In [ 21 ] , the oculus is detected based on the border information of the flag, with its darker coloring material doing it much easier to turn up. A back extension nervous web is trained to sort the province of the eyes ( either unfastened or closed ) . On the other manus, in [ 22 ] , a intercrossed nervous web and a combination of the “ HMAX theoretical account ” and “ Viola-Jones sensor ” together with a Multi-Layer Perceptron ( MLP ) are used to turn up the face. The grade of caput rotary motion, oculus closing and oral cavity openness are the fatigue steps calculated.
To sort driver public presentation informations, Liang et Al. [ 23 ] make usage of Support Vector Machines ( SVMs ) . They focus on cognitive ( mental ) , instead than ocular driver distractions. For fast face and facial characteristic sensing, the method proposed by Viola and Jones affecting a boosted cascade of characteristics based on Haar ripples is adopted in a figure of documents, including [ 24 ] and [ 25 ] . Often, a loanblend of techniques are used to obtain better consequences for driver weariness sensing. Saradadevi and Bajaj [ 26 ] usage Viola-Jones ‘ method for mouth sensing and SVMs to right sort normal and yawning oral cavity cases. On the contrary, the one presented by Narole and Bajaj [ 27 ] combines pixel-based skin coloring material cleavage for face sensing and a mixture of nervous webs and familial algorithms to optimally find the weariness index, with the nervous web being given as initial input values for oculus closing and oscitance rate.
Other Fatigue Monitoring Systems
As with drivers in autos, pilots in aircrafts are obviously capable to tire, chiefly due to the drawn-out flight continuances. NTI Inc. and Science Applications International Corporation ( SAIC ) [ 28 ] designed the Fatigue Avoidance Scheduling Tool ( FAST ) , a system intended to track and foretell weariness degrees for U.S. Air Force pilots, based on the SAFTE ( Sleep, Activity, Fatigue and Task Effectiveness ) theoretical account created by Dr. Steven Hursh. Another application in which weariness monitoring is utile is in the bar of Computer Vision Syndrome [ 29 ] , a status caused by working for drawn-out hours in forepart of show devices, such as computing machine proctors. Matsushita et Al. [ 30 ] besides developed a wearable weariness monitoring system which detects marks of weariness based on caput motions.
The broad assortment of different applications developed to supervise weariness is an grounds of the turning importance of this field. The focal point in the following portion of the literature reappraisal will switch to the weariness analysis attack taken in this thesis: the sensing of faces and their characteristics in images. The implicit in methods and algorithms typically used in this procedure will be discussed.
Reappraisal on Face and Facial Feature Detection Techniques
Detecting faces in knowledge-based techniques involves the encryption of a set of simple regulations specifying the features of the human face, including pixel strengths in the images and the places and correlativities between the different characteristics, since these are common to all human existences.
In a knowledge-based method presented by Yang and Huang [ 31 ] , a hierarchy of grayscale images of different declarations together with three different classs of regulations are used. The images are analysed for possible face campaigners by using regulations that have to make with the cell strength distribution of the human face. An betterment to this multi-resolution method was proposed by Kotropoulos and Pitas [ 32 ] . Alternatively of ciphering the mean pixel strength of each cell, merely those for each image row and column are computed, organizing perpendicular and horizontal profiles severally.
To vouch a high sensing rate, the regulations in knowledge-based methods must neither be excessively general nor excessively specific, and hence, the coevals of regulations for the face must be performed really carefully. Because of the complexness required in coding all possible face constellations, rule-based techniques do non provide for different face airss [ 33 ] , doing them decidedly inappropriate for weariness monitoring applications.
Feature-based attacks to confront sensing differ in a important manner from rule-based techniques in that they foremost attempt to place a individual ‘s facial properties and later find whether the latter are valid plenty to represent a human face, ensuing in the sensing of that face.
The presence of faces in images is frequently determined by trying to observe facial characteristics such as the eyes, nose and mouth. In a method presented by Sirehoy [ 34 ] , the egg-shaped nature of the human face is used as the footing for face sensing in grayscale images with littered backgrounds. Due to the different visual aspects of facial characteristics in images, Leung et Al. [ 35 ] usage a combination of several local characteristic sensors utilizing Gaussian derivative filters together with a statistical theoretical account of the geometrical distances between these characteristics to guarantee accurate face localisation. Han et Al. [ 36 ] , on the other manus, usage morphological operations that focus chiefly on the oculus part in their efforts to observe faces, based on the logical thinking that this is the most consistent facial part in different light conditions. A more robust and flexible feature-based system was presented by Yow and Cipolla [ 37 ] . The theoretical account cognition of the face that is used screens a wider country, including the superciliums, eyes, nose and mouth. A figure of Partial Face Groups ( PFGs ) , tantamount to a subset of these characteristic points ( 4 ) , are used to provide for partial face occlusions.
Another face cue that is used for sensing intents is its textural form, this being specific to worlds and hence easy discriminable from other forms. Manian and Ross [ 38 ] present an algorithm that uses the symmetricalness and uniformity of the facial form as the footing of sensing. Rikert et Al. [ 39 ] tackle texture-based sensing in a different manner, utilizing a statistical method that learns to correctly sort whether an image contains a face or non.
Many plants related to human clamber coloring material as a face sensing cue have been presented in recent old ages. Detection can be either pixel-based or region-based. The former attack is normally taken, in which each pel is analysed and classified as either tegument or non-skin. Two chief picks are made during this procedure: the coloring material infinite and tegument modeling method. Harmonizing to [ 40 ] , the normalized RGB, HSV and YCrCb coloring material infinites are typically used to pattern skin coloring material. Normalized RGB [ 41 – 45 ] is reported to be consistent in different light conditions and face orientations. On the other manus, YCrCb [ 46 – 48 ] and HSV [ 49 – 51 ] are normally chosen since they specifically separate the luminosity and chrominance constituents of the images. In [ 40 ] , several other tegument patterning techniques normally adopted are mentioned.
Template matching methods
Another proposed method for face sensing involves the storage of forms of the face and its characteristics, which are so compared to existent face images and given a correlativity value ( i.e. the degree of similarity between the existent image and the stored form ) . The higher this value, the greater is the opportunity that the image contains a face. Works on templet fiting techniques in recent old ages have focused both on fixed and variable-size ( deformable ) templets.
Fengjun et Al. [ 52 ] and Ping et Al. [ 53 ] usage a combination of skin coloring material cleavage and templet matching for face sensing. Two grayscale templets with predefined sizes – one covering the whole face and the other concentrating merely on the part incorporating the two eyes – are utilised in both systems. Fixed-size templets, although straightforward to implement, miss adaptability to different caput places since sensing is greatly affected by the orientation defined in the templet.
An improved templet matching method is one in which the templet can be altered to better reflect the input images and therefore would be able to place a wider assortment of faces in different airss. Yuille et Al. [ 54 ] propose deformable oculus and mouth templet matching in their work. Initially, the templets are parameterized through pre-processing to bespeak the expected form of both characteristics. The work presented by Lanitis et Al. [ 55 ] besides parameterizes the templets, concentrating on the coevals of flexible molded human face theoretical accounts through the usage of a “ Point Distribution Model ” ( PDM ) [ 56 ] which is trained on a figure of images per individual with characteristic fluctuations within and between faces.
Rather than being based on a set of preset templets, appearance-based face sensing relies on machine larning techniques that identify the presence of faces and their major features after a procedure of developing on existent universe informations. One of the most widely adopted machine larning attacks for face sensing are nervous webs, chiefly because of the success they achieved in other applications affecting pattern acknowledgment. Rowley et Al. [ 57 ] propose a robust multi-layer multi-network nervous web that takes as input pre-processed 20×20 grayscale pel images to which a filter is applied at each pel place, returning a face correlativity value from -1 to 1. The concealed beds of the nervous web are designed to supervise different shaped countries of the human face, such as both eyes utilizing a 20×5 pel window and single eyes and other characteristics with the 5×5 and 10×10 Windowss. The web so outputs another mark finding the presence or otherwise of a face in a peculiar window.
Yang et Al. [ 58 ] establish their system on a Sparse Network of Winnows ( SNoW ) [ 59 ] . Two mark nodes ( “ linear units ” ) patterning face and non-face form characteristics are used in this instance. The active characteristics ( with binary representation ) in an input illustration are first identified and given as input to the web. The mark nodes are “ coupled via leaden borders ” to a subset of the characteristics. To update the weights for farther preparation, the Winnow update regulation method developed by Littlestone [ 60 ] is adopted.
A additive categorization technique in the signifier of Support Vector Machines ( SVMs ) was used to observe faces in an application presented by Osuna et al [ 61 ] in 1997. While the bulk of machine acquisition attacks ( including nervous webs ) effort to take down the “ empirical hazard ” , i.e. the mistake value in the preparation procedure, SVMs attempt to cut down the upper edge of the expected generalisation mistake in a procedure called “ structural hazard minimisation ” .
Viola and Jones [ 62 ] present a rapid object sensing system holding face sensing as its motive. A important difference from other proposed systems is that rectangular characteristics, instead than pels, nowadays in the inputted grayscale images are used as the bases for categorization. This has the consequence of increasing the velocity of the overall procedure. Viola and Jones ‘ method will be discussed in item in the following chapter of this thesis.
Purposes and Aims
Familiarization with the OpenCV tool.
Literature Review about bing systems and methods to be used in this Dissertation.
Fast face sensing utilizing Viola-Jones technique.
Execution of multiple facial characteristics used to find the fatigue degree.
Application of Support Vector Machine classifier to observe unsafe state of affairss such as driver kiping etc.
Real-time execution of the proposed methods within OpenCV.
Viola-Jones technique for face sensing.
Support vector machines to sort facial visual aspect ( e.g. open/closed eye/mouth ) .
Features to be taken into consideration: caput motion, oculus closing and frequence of oral cavity gap ( bespeaking yawning ) .
Eye weariness steps include PERCLOS ( PERcentage Eye CLOSure over clip ) and AECS ( Average Eye Closure Speed ) .
Comparing the developed system to other systems found in literature in footings of preciseness, callback and truth.
Deducing some trial informations on which the algorithms will be tested.
Test topics seeking out the application.
Showing the consequences obtained.
2 page abstract for ICT Final YearA Student Projects Exhibition.
Presentation Slides and Poster.
Spiral and difficult edge transcripts of the Dissertation Report.
C++ application, preparation and testing resources.
Section 2: Work Plan
Work done so far
Collected and read several documents related to bing driver weariness systems and face sensing in general.
Completed the first bill of exchange of the literature reappraisal.
Familiarized myself with the OpenCV environment.
Used a webcam to capture two short cartridge holders inside a auto, one in sunny and the other in cloud-covered conditions.
Collected 2000 positive and 4000 negative images for face sensing.
Positive images: 1500 taken from FERET grayscale face database, the other 500 from the captured cartridge holders.
Negative images: created a C++ application to randomly choice non-relevant countries of the frames of the two captured cartridge holders.
Created another C++ application to be able to harvest the positive images to bespeak merely the needed rectangular countries, bring forthing a text file to be used in the preparation procedure.
Used this information to bring forth a classifier for faces in XML format with OpenCV ‘s Haar preparation public-service corporation.
Compute truth, preciseness and callback values for the face sensing preparation.
Trial with new picture cartridge holders and observing the consequences obtained.
Perform Cross Validation.
Train the classifier for oral cavities, once more utilizing positive and negative images. For oculus sensing, an already generated classifier will be used.
Extract characteristics from face, oculus and mouth sensing.
Integrate and utilize a C++ library for support vector machines, such as libSVM, to sort facial visual aspect.
Write Abstract, Introduction, Methodology, Evaluation, Results, Future Work and Conclusion of the Dissertation Report.
Write Review Report.
Write 2 page abstract for ICT Final YearA Student Projects Exhibition.
Work on Presentation Slides and Poster.
Schedule ( Gantt Chart )
Section 3: Mentions
[ 1 ] D. Dinges, M. Mallis, G. Maislin and J. Powell ( 1998 ) . “ Concluding study: Evaluation of Techniques for Ocular Measurement as an Index of Fatigue and the Basis for Alertness Management ” , U.S. Dept. Transportation, National Highway Traffic Safety Administration, [ online ] , Last accessed on 4th October 2010, Available at: hypertext transfer protocol: //ntl.bts.gov/lib/jpodocs/edlbrow/7d01! .pdf
[ 2 ] National Highway Traffic Safety Administration ( 2005 ) . “ NHTSA Vehicle Safety Rulemaking and Supporting Research Priorities: Calendar Old ages 2005-2009 ” , [ online ] , Last accessed on 4th October 2010, Available at: hypertext transfer protocol: //www.nhtsa.gov/cars/rules/rulings/priorityplan-2005.html
[ 3 ] National Highway Traffic Safety Administration ( 2005 ) . “ Traffic Safety Facts 2005: A Compilation of Motor Vehicle Crash Data from the Fatality Analysis Reporting System and the General Estimates System ” , National Center for Statistics and Analysis, U.S. Dept. Transportation, [ online ] , Last accessed on 4th October 2010, Available at: hypertext transfer protocol: //www-nrd.nhtsa.dot.gov/pubs/tsf2005.pdf
[ 4 ] Hall, Hammerschmidt and Francis ( 1995 ) . “ Safety Recommendation ” , National Transportation Safety Board, [ online ] , Last accessed on 21st December 2010, Available at: hypertext transfer protocol: //www.ntsb.gov/recs/letters/1995/H95_5D.pdf
[ 5 ] J. Cavuoto, “ Alertness Monitoring Devices Emerge from San Diego ” , Neurotech Business Report, [ online ] , Last accessed on 21st September 2010, Available at: hypertext transfer protocol: //www.neurotechreports.com/pages/alertness.html
[ 6 ] J-S Co. Neurocom, “ Engine Driver Vigilance Telemetric Control System EDVTCS ” , [ online ] , Last accessed on 21st September 2010, Available at: hypertext transfer protocol: //www.neurocom.ru/en2/pdf/edvtcs_adv_eng.pdf
[ 7 ] Fatigue Management International, “ ASTiD: Advisory System for Tired Drivers ” , [ online ] , Last accessed on 22nd September 2010, Available at: hypertext transfer protocol: //www.fmig.org/ASTID % 20Information % 20Document.pdf
[ 8 ] AssistWare Technology, “ Tired of Confronting Another Night Entirely? SafeTRAC can assist ” , [ online ] , Last accessed on 22nd September 2010, Available at: hypertext transfer protocol: //www.assistware.com/Downloads/SafeTRAC-Fleet % 20Datasheet.pdf
[ 9 ] European Commission, Information Society Technologies ( 2002 ) . “ System for effectual Assessment of driver watchfulness and Warning Harmonizing to traffic hazard Estimation ” , [ online ] , Last accessed on 21st September 2010, Available at: hypertext transfer protocol: //www.awake-eu.org/pdf/d1_1.pdf
[ 10 ] D. F. Dinges and R. Grace ( 1998 ) . “ PERCLOS: A Valid Psychophysiological Measure of Alertness As Assessed by Psychomotor Vigilance ” , US Department of Transportation, Federal Highway Administration, [ online ] , Last accessed on 21st December 2010, Available at: hypertext transfer protocol: //www.fmcsa.dot.gov/documents/tb98-006.pdf
[ 11 ] W. W. Wierwille ( 1994 ) . “ Overview of Research on Driver Drowsiness Definition and Driver Drowsiness Detection ” , 14th Technical Int. Conf. on Enhanced Safety of Drivers ( ESV ) , Munich, Germany, pp.23-26.
[ 12 ] Sing Machines, “ faceLAB 5 ” , [ online ] , Last accessed on 21st September 2010, Available at: hypertext transfer protocol: //www.seeingmachines.com/pdfs/brochures/faceLAB-5.pdf
[ 13 ] E. Bekiaris ( 2004 ) . “ AWAKE Project Aim and Objectives ” , Road Safety Workshop, Balocco, Italy, [ online ] , Last accessed on 21st September 2010, Available at: hypertext transfer protocol: //www.awake-eu.org/pdf/aim_achievements.pdf
[ 14 ] P. Smith, M. Shah and N. D. V. Lobo ( 2003 ) . “ Determining Driver Visual Attention with One Camera ” , IEEE Transactions on Intelligent Transportation Systems, Vol. 4, No. 4, pp. 205 – 218, [ online ] , Last accessed on 16th August 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.4.842 & A ; rep=rep1 & A ; type=pdf
[ 15 ] E. Wahlstrom, O. Masoud and N. Papanikolopoulos ( 2003 ) . “ Vision Based Methods for Driver Monitoring ” , IEEE Intelligent Transportation Systems Conf, pp. 903 – 908, [ online ] , Last accessed on 28th July 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.3.4434 & A ; rep=rep1 & A ; type=pdf
[ 16 ] H. Veeraraghavan and N. Papanikolopoulos ( 2001 ) . “ Detecting Driver Fatigue Through the Use of Advanced Face Monitoring Techniques ” , ITS Institute, Center for Transportation Studies, University of Minnesota, [ online ] , Last accessed on 28th July 2010, Available at: hypertext transfer protocol: //www.cts.umn.edu/pdf/CTS-01-05.pdf
[ 17 ] Aryuanto and F. Y. Limpraptono ( 2009 ) . “ A Vision Based System for Monitoring Driver Fatigue ” , Department of Electrical Engineering, Institut Teknologi Nasional ( ITN ) Malang, Yogyakarta, Indonesia, [ online ] , Last accessed on 17th June 2010, Available at: hypertext transfer protocol: //aryuanto.files.wordpress.com/2008/10/teknoin09-1.pdf
[ 18 ] W.-B. Horng and C.-Y. Chen ( 2009 ) . “ Improved Driver Fatigue Detection System Based on Eye Tracking and Dynamic Template Matching ” , Department of Computer Science and Information Engineering, Tamkang University, Taipei, Taiwan, [ online ] , Last accessed on 28th July 2010, Available at: hypertext transfer protocol: //dspace.lib.fcu.edu.tw/bitstream/2377/11188/1/ce07ics002008000132.pdf
[ 19 ] J. H. Yang, Z.-H. Mao, L. Tijerina, T. Pilutti, J. F. Coughlin and E. Feron ( 2009 ) . “ Detection of Driver Fatigue Caused by Sleep Deprivation ” , IEEE Transactions on Systems, Man and Cybernetics, Part A: Systems and Humans, Vol. 39, No. 4, pp. 694 – 705, [ online ] , Last accessed on 16th September 2010, Available at: hypertext transfer protocol: //www.engr.pitt.edu/electrical/faculty-staff/mao/home/Papers/YMT09_DriverFatigue.pdf
[ 20 ] Q. Ji, Z. Zhu and P. Lan ( 2004 ) . “ Real-time Nonintrusive Monitoring and Prediction of Driver Fatigue ” , IEEE Transactions on Vehicular Technology, Vol. 53, No. 4, pp. 1052 – 1068, [ online ] , Last accessed on 16th August 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.2.4714 & A ; rep=rep1 & A ; type=pdf
[ 21 ] T. D’Orazio, M. Leo, P. Spagnolo and C. Guaragnella ( 2004 ) . “ A Neural System for Eye Detection in a Driver Vigilance Application ” , Proceedings of the 7th International IEEE Conference on Intelligent Transportation Systems, pp. 320 – 325, [ online ] , Last accessed on 28th July 2010, Available at: hypertext transfer protocol: //pr.radom.net/~pgolabek/its/A nervous system for oculus sensing in a driver watchfulness application.pdf
[ 22 ] S. RibariA‡ , J. LovrencI?icI? and N. PavesI?icI? ( 2010 ) . “ A Neural-Network-Based System for Monitoring Driver Fatigue ” , 15th IEEE Mediterranean Electrotechnical Conference, pp. 1356 – 1361.
[ 23 ] Y. Liang, M. L. Reyes and J. D. Lee ( 2007 ) . “ Real-time Detection of Driver Cognitive Distraction Using Support Vector Machines ” , IEEE Transactions on Intelligent Transportation Systems, Vol. 8, No. 2, pp. 340 – 350.
[ 24 ] H. Ma, Z. Yang, Y. Song and P. Jia ( 2008 ) . “ A Fast Method for Monitoring Driver Fatigue Using Monocular Camera ” , Proceedings of the 11th Joint Conference on Information Sciences, Atlantis Press, [ online ] , Last accessed on 28th July 2010, Available at: hypertext transfer protocol: //www.atlantis-press.com/php/download_paper.php? id=1717
[ 25 ] T. Brandt, R. Stemmer, B. Mertsching and A. Rakotonirainy ( 2004 ) . “ Low-cost Ocular Driver Monitoring System for Fatigue and Monotony ” , 2004 IEEE International Conference on Systems, Man and Cybernetics, Vol. 7, pp. 6451 – 6456, [ online ] , Last accessed on 28th July 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.93.1899 & A ; rep=rep1 & A ; type=pdf
[ 26 ] M. Saradadevi and P. R. Bajaj ( 2008 ) . “ Driver Fatigue Detection utilizing Mouth and Yawning Analysis ” , International Journal of Computer Science and Network Security, Vol. 8, No. 6, pp. 183 – 188, [ online ] , Last accessed on 28th July 2010, Available at: hypertext transfer protocol: //paper.ijcsns.org/07_book/200806/20080624.pdf
[ 27 ] N. G. Narole and P. R. Bajaj ( 2009 ) . “ A Neuro-Genetic System Design for Monitoring Driver ‘s Fatigue ” , International Journal of Computer Science and Network Security, Vol. 9, No. 3, pp. 87 – 91, [ online ] , Last accessed on 28th July 2010, Available at: hypertext transfer protocol: //paper.ijcsns.org/07_book/200903/20090311.pdf
[ 28 ] C. Trautvetter ( 2005 ) . “ Software Scheduling Tool Fights Crewmember Fatigue ” , Aviation International News, [ online ] , Last accessed on 20th September 2010, Available at: www.novasci.com/AIN-JL05.pdf
[ 29 ] M. Divjak and H. Bischof ( 2009 ) . “ Eye Blink Based Fatigue Detection for Prevention of Computer Vision Syndrome ” , IAPR Conference on Machine Vision Applications, Keio University, Hiyoshi, Japan, [ online ] , Last accessed on 20th September 2010, Available at: hypertext transfer protocol: //www.icg.tugraz.at/Members/divjak/prework/MVA_2009_presentation % 20- % 20Divjak.pdf
[ 30 ] S. Matsushita, A. Shiba and K. Nagashima ( 2006 ) . “ A Wearable Fatigue Monitoring System – Application of Human-Computer Interaction Evaluation ” , Proceedings of the seventh Australasian User Interface Conference, Vol. 50, [ online ] , Last accessed on 17th September 2010, Available at: hypertext transfer protocol: //crpit.com/confpapers/CRPITV50Matsushita.pdf
[ 31 ] G. Yang and T. S. Huang ( 1994 ) . “ Human Face Detection in Complex Background ” , Pattern Recognition, Vol. 27, No. 1, pp. 53 – 63.
[ 32 ] C. Kotropoulos and I. Pitas ( 1997 ) . “ Rule-Based Face Detection in Frontal Views ” , Proceedings of the International Conference on Acoustics, Speech and Signal Processing, Vol. 4, pp. 2537 – 2540, [ online ] , Last accessed on 16th October 2010, Available at: hypertext transfer protocol: //poseidon.csd.auth.gr/papers/PUBLISHED/CONFERENCE/pdf/Kotropoulos_ICASSP97.pdf
[ 33 ] M.-H. Yang, D. J. Kriegman and N. Ahuja ( 2002 ) . “ Detecting Faces in Images: A Survey ” , IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 24, No. 1, pp. 34 – 58, [ online ] , Last accessed on 16th August 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.63.7658 & A ; rep=rep1 & A ; type=pdf
[ 34 ] S. A. Sirehoy ( 1993 ) . “ Human Face Segmentation and Identification ” , Computer Vision Laboratory, Center for Automation Research, University of Maryland, [ online ] , Last accessed on 25th October 2010, Available at: hypertext transfer protocol: //drum.lib.umd.edu/bitstream/1903/400/2/CS-TR-3176.pdf
[ 35 ] T. K. Leung, M. C. Burl and P. Perona ( 1995 ) . “ Finding Faces in Cluttered Scenes utilizing Random Labelled Graph Matching ” , Proceedings of the fifth International Conference on Computer Vision, Cambridge, Massachusetts, U.S.A. , [ online ] , Last accessed on 25th October 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.34.8710 & A ; rep=rep1 & A ; type=pdf
[ 36 ] C.-C. Han, H.-Y. M. Liao, K.-C. Yu and L.-H. Chen ( 1996 ) . “ Fast Face Detection via Morphology-based Pre-processing ” , Proceedings of the 9th International Conference on Image Analysis and Processing, Florence, Italy, [ online ] , Last accessed on 25th October 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.29.4448 & A ; rep=rep1 & A ; type=pdf
[ 37 ] K. C. Yow and R. Cipolla ( 1996 ) . “ Feature-Based Human Face Detection ” , Image and Vision Computing, Vol. 15, No. 9, pp. 713 – 735, [ online ] , Last accessed on 24th October 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.28.5815 & A ; rep=rep1 & A ; type=pdf
[ 38 ] V. Manian and A. Ross ( 2004 ) . “ A Texture-based Approach to Face Detection ” , Biometric Consortium Conference ( BCC ) , Crystal City, VA, [ online ] , Last accessed on 26th October 2010, Available at: hypertext transfer protocol: //www.csee.wvu.edu/~ross/pubs/RossFaceTexture_BCC04.pdf
[ 39 ] T. D. Rikert, M. J. Jones and P. Viola ( 1999 ) . “ A Texture-Based Statistical Model for Face Detection ” , Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, [ online ] , Last accessed on 26th October 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.32.8916 & A ; rep=rep1 & A ; type=pdf
[ 40 ] V. Vezhnevets, V. Sazonov and A. Andreeva ( 2003 ) . “ A Survey on Pixel-Based Skin Color Detection Techniques ” , GRAPHICON-2003, pp. 85-92, [ online ] , Last accessed on 24th October 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.5.521 & A ; rep=rep1 & A ; type=pdf
[ 41 ] D. Brown, I. Craw and J. Lewthwaite ( 2001 ) . “ A SOM Based Approach to Skin Detection with Application in Real Time Systems ” , Proceedings of the British Machine Vision Conference, [ online ] , Last accessed on 27th October 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.16.2675 & A ; rep=rep1 & A ; type=pdf
[ 42 ] M. Soriano, B. Martinkauppi, S. Huovinen and M. Laaksonen ( 2000 ) . “ Skin Detection in Video Under Changing Illumination Conditions ” , Proceedings of the fifteenth International Conference on Pattern Recognition, pp. 839 – 842, [ online ] , Last accessed on 13th November 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download ; jsessionid=751F3CF514D95B2D7C8C425A1753714B? doi=10.1.1.16.2582 & A ; rep=rep1 & A ; type=pdf
[ 43 ] N. Oliver, A. P. Pentland and F. Berard ( 1997 ) . “ LAFTER: Lips and Face Real Time Tracker ” , Proceedings of the 1997 Conference on Computer Vision and Pattern Recognition, pp. 123 – 129, [ online ] , Last accessed on 13th November 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.50.9491 & A ; rep=rep1 & A ; type=pdf
[ 44 ] J. Yang, W. Lu and A. Waibel ( 1998 ) . “ Skin Color Modelling and Adaptation ” , Proceedings of the Asian Conference on Computer Vision, pp. 687 – 694, [ online ] , Last accessed on 13th November 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.44.8168 & A ; rep=rep1 & A ; type=pdf
[ 45 ] L. Mostafa and S. Abdelazeem ( 2005 ) . “ Face Detection Based on Skin Color Using Neural Networks ” , Proceedings of the 1st International Conference on Graphics, Vision and Image Processing, Cairo, Egypt, pp. 53 – 58, [ online ] , Last accessed on 24th October 2010, Available at: hypertext transfer protocol: //www.icgst.com/GVIP05/papers/P1150535113.pdf
[ 46 ] R.-L. Hsu, M. Abdel-Mottaleb and A. K. Jain ( 2002 ) . “ Face Detection in Color Images ” , IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 24, No. 5, pp. 696 – 706, [ online ] , Last accessed on 30th August 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.33.4990 & A ; rep=rep1 & A ; type=pdf
[ 47 ] J. Ahlberg ( 1999 ) . “ A System for Face Localization and Facial Feature Extraction ” , Technical Report, no. LiTH-ISY-R-2172, Linkoping University, [ online ] , Last accessed on 13th November 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.43.7504 & A ; rep=rep1 & A ; type=pdf
[ 48 ] D. Chai and A. Bouzerdoum ( 2000 ) . “ A Bayesian Approach to Skin Color Classification in YCbCr Color Space ” , IEEE TENCON 2000, Vol. 2, pp. 421 – 424, [ online ] , Last accessed on 13th November 2010, Available at: www.se.ecu.edu.au/~dchai/public/papers/tencon2000.pdf
[ 49 ] S. J. McKenna, S. Gong and Y. Raja ( 1998 ) . “ Modeling Facial Colour and Identity with Gaussian Mixtures ” , Proceedings of Pattern Recognition, pp. 1883 – 1892, [ online ] , Last accessed on 13th November 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.34.902 & A ; rep=rep1 & A ; type=pdf
[ 50 ] L. Sigal, S. Sclaroff and V. Athitsos ( 2000 ) . “ Estimation and Prediction of Evolving Color Distributions for Skin Segmentation Under Changing Illumination ” , Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 152 – 159, [ online ] , Last accessed on 13th November 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.1.9735 & A ; rep=rep1 & A ; type=pdf
[ 51 ] L. Jordao, M. Perrone and J. P. Costeira ( 1999 ) . “ Active Face and Feature Tracking ” , Proceedings of the tenth International Conference on Image Analysis and Processing, pp. 572 – 576, [ online ] , Last accessed on 13th November 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.33.893 & A ; rep=rep1 & A ; type=pdf
[ 52 ] L. Fengjun, A. Haizhou, L. Luhong and X. Guangyou ( 2000 ) . “ Face Detection Based on Skin Color and Template Matching ” , Proceedings of the 1st International Conference on Image and Graphics, [ online ] , Last accessed on 14th November 2010, Available at: hypertext transfer protocol: //220.127.116.11:8080/44/course/chap03/sourse/colorfacedetect.pdf
[ 53 ] S. T. Y. Ping, C. H. Weng and B. Lau, “ Face Detection Through Template Matching and Color Segmentation ” , Stanford University, [ online ] , Last accessed on 14th November 2010, Available at: hypertext transfer protocol: //www.stanford.edu/class/ee368/Project_03/Project/reports/ee368group04.pdf
[ 54 ] A. L. Yuille, P. W. Hallinan and D. S. Cohen ( 1992 ) . “ Feature Extraction from Faces utilizing Deformable Templates ” , International Journal of Computer Vision, Vol. 8, No. 2, pp. 99 – 111, [ online ] , Last accessed on 14th November 2010, Available at: hypertext transfer protocol: //www.ittc.ku.edu/~potetz/EECS_741/SuggestedReadings/Lecture_14_Yuille_DeformableTemplates_IJCV92.pdf
[ 55 ] A. Lanitis, C. J. Taylor and T. F. Cootes ( 1995 ) . “ An Automatic Face Identification System Using Flexible Appearance Models ” , Image and Vision Computing, Vol. 13, No. 5, pp. 393 – 401, [ online ] , Last accessed on 14th November 2010, Available at: hypertext transfer protocol: //www.bmva.org/bmvc/1994/bmvc-94-006.pdf
[ 56 ] T. F. Cootes, A. Hill, C. J. Taylor and J. Haslam ( 1994 ) . “ The Use of Active Shape Models For Locating Structures in Medical Images ” , Image and Vision Computing, Vol. 12, No. 6, pp. 355 – 366, [ online ] , Last accessed on 15th November 2010, Available at: hypertext transfer protocol: //www.sci.utah.edu/~gerig/CS7960-S2010/handouts/ivc95.pdf
[ 57 ] H. A. Rowley, S. Baluja and T. Kanade ( 1998 ) . “ Neural Network Based Face Detection ” , IEEE Transactions On Pattern Analysis and Machine intelligence, Vol. 20, No. 1, pp. 23 – 38, [ online ] , Last accessed on 4th December 2010, Available at: hypertext transfer protocol: //citeseer.ist.psu.edu/viewdoc/download? doi=10.1.1.110.5546 & A ; rep=rep1 & A ; type=pdf
[ 58 ] M.-H. Yang, D. Roth and N. Ahuja ( 2000 ) . “ A SNoW-Based Face Detector ” , Advances in Neural Information Processing Systems 12, MIT Press, pp. 855 – 861, [ online ] , Last accessed on 4th December 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.41.152 & A ; rep=rep1 & A ; type=pdf
[ 59 ] N. Rizzolo ( 2005 ) . “ SNoW: Sparse Network of Winnows ” , Cognitive Computation Group, Department of Computer Science, University of Illinois at Urbana-Champaign, 2005, [ online – presentation ] , Last accessed on 5th December 2010, Available at: hypertext transfer protocol: //cogcomp.cs.illinois.edu/tutorial/SNoW.pdf
[ 60 ] N. Littlestone ( 1988 ) . “ Learning Quickly when Irrelevant Attributes Abound. A New Linear-threshold Algorithm ” , Machine Learning 2, Kluwer Academic Publishers, pp. 285 – 318, [ online ] , Last accessed on 5th December 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.130.9013 & A ; rep=rep1 & A ; type=pdf
[ 61 ] E. Osuna, R. Freund and F. Girosi ( 1997 ) . “ Training Support Vector Machines: An Application to Face Detection ” , Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 130 – 136, [ online ] , Last accessed on 5th December 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.9.6021 & A ; rep=rep1 & A ; type=pdf
[ 62 ] P. Viola and M. Jones ( 2001 ) . “ Rapid Object Detection utilizing a Boosted Cascade of Simple Features ” , Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 511 – 518, [ online ] , Last accessed on 5th December 2010, Available at: hypertext transfer protocol: //citeseerx.ist.psu.edu/viewdoc/download? doi=10.1.1.137.9386 & A ; rep=rep1 & A ; type=pdf