05-05-2011, 04:21 PM
Abstract
A human-computer interface (HCI) system designed for useby people with severe disabilities is presented. People thatare severely paralyzed or afflicted with diseases such asALS (Lou Gehrig’s disease) or multiple sclerosis are unableto move or control any parts of their bodies except fortheir eyes. The system presented here detects the user’s eyeblinks and analyzes the pattern and duration of the blinks,using them to provide input to the computer in the form ofa mouse click. After the automatic initialization of the systemoccurs from the processing of the user’s involuntary eyeblinks in the first few seconds of use, the eye is tracked inreal time using correlation with an online template. If theuser’s depth changes significantly or rapid head movementoccurs, the system is automatically reinitialized. There areno lighting requirements nor offline templates needed forthe proper functioning of the system. The system works withinexpensive USB cameras and runs at a frame rate of 30frames per second. Extensive experiments were conductedto determine both the system’s accuracy in classifying voluntaryand involuntary blinks, as well as the system’s fitnessin varying environment conditions, such as alternative cameraplacements and different lighting conditions. These experimentson eight test subjects yielded an overall detectionaccuracy of 95.3%.
1 Introduction
A great deal of computer vision research is dedicated tothe implementation of systems designed to detect usermovements and facial gestures [1, 2, 4, 5, 6, 15, 16]. Inmany cases, such systems are created with the specific goalof providing a way for people with disabilities or limitedmotor skills to be able to use computer systems, albeit inmuch simpler applications [1, 15, 16]. The motivation forthe system proposed here is to provide an inexpensive,unobtrusive means for disabled people to interact withsimple computer applications in a meaningful way thatrequires minimal effort.This goal is accomplished using a robust algorithmbased on the work by Grauman et al. [11, 12]. Some ofthese methods are implemented here, while some have beenenhanced or modified to the end of simplified initializationand more efficient maintenance of the real time tracking.The automatic initialization phase is triggered by theanalysis of the involuntary blinking of the current user ofthe system, which creates an online template of the eyeto be used for tracking. This phase occurs each time thecurrent correlation score of the tracked eye falls below adefined threshold in order to allow the system to recoverand regain its accuracy in detecting the blinks. This systemcan be utilized by users that are able to voluntarily blinkand have a use for applications that require mouse clicks asinput (e.g. switch and scanning programs/games
Download full report
http://citeseerx.ist.psu.edu/viewdoc/dow...1&type=pdf
A human-computer interface (HCI) system designed for useby people with severe disabilities is presented. People thatare severely paralyzed or afflicted with diseases such asALS (Lou Gehrig’s disease) or multiple sclerosis are unableto move or control any parts of their bodies except fortheir eyes. The system presented here detects the user’s eyeblinks and analyzes the pattern and duration of the blinks,using them to provide input to the computer in the form ofa mouse click. After the automatic initialization of the systemoccurs from the processing of the user’s involuntary eyeblinks in the first few seconds of use, the eye is tracked inreal time using correlation with an online template. If theuser’s depth changes significantly or rapid head movementoccurs, the system is automatically reinitialized. There areno lighting requirements nor offline templates needed forthe proper functioning of the system. The system works withinexpensive USB cameras and runs at a frame rate of 30frames per second. Extensive experiments were conductedto determine both the system’s accuracy in classifying voluntaryand involuntary blinks, as well as the system’s fitnessin varying environment conditions, such as alternative cameraplacements and different lighting conditions. These experimentson eight test subjects yielded an overall detectionaccuracy of 95.3%.
1 Introduction
A great deal of computer vision research is dedicated tothe implementation of systems designed to detect usermovements and facial gestures [1, 2, 4, 5, 6, 15, 16]. Inmany cases, such systems are created with the specific goalof providing a way for people with disabilities or limitedmotor skills to be able to use computer systems, albeit inmuch simpler applications [1, 15, 16]. The motivation forthe system proposed here is to provide an inexpensive,unobtrusive means for disabled people to interact withsimple computer applications in a meaningful way thatrequires minimal effort.This goal is accomplished using a robust algorithmbased on the work by Grauman et al. [11, 12]. Some ofthese methods are implemented here, while some have beenenhanced or modified to the end of simplified initializationand more efficient maintenance of the real time tracking.The automatic initialization phase is triggered by theanalysis of the involuntary blinking of the current user ofthe system, which creates an online template of the eyeto be used for tracking. This phase occurs each time thecurrent correlation score of the tracked eye falls below adefined threshold in order to allow the system to recoverand regain its accuracy in detecting the blinks. This systemcan be utilized by users that are able to voluntarily blinkand have a use for applications that require mouse clicks asinput (e.g. switch and scanning programs/games
Download full report
http://citeseerx.ist.psu.edu/viewdoc/dow...1&type=pdf